DWQ-2024-008259

Utah Lake Water Quality Study
Atmospheric Deposition Decision Support Document

Draft: February 17, 2023
Revised: February 27, 2023, March 29, 2024

Developed by: the ULWQS Atmospheric Deposition Subgroup (Dr. Mike Brett, University of
Washington; Dr. Mitch Hogsett; Dr. Theron Miller, Wasatch Front Water Quality Council; and
Dr. Hans Paerl, University of North Carolina Chapel Hill)

Technical Support: Tetra Tech, Inc.
Facilitation Support: Peak Facilitation Group

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
i

Contents

Tables ..................................................................................................................................................................... iii
Figures.................................................................................................................................................................... iv
Definitions.............................................................................................................................................................. vi
1 Executive Summary ........................................................................................................................................ 1
2 Background ..................................................................................................................................................... 7
3 Analysis Plan Development ............................................................................................................................ 8
4 Review and Summarize Data .......................................................................................................................... 9
4.1 Methods .................................................................................................................................................... 9
4.1.1 Data Compilation and Organization ................................................................................................. 9
4.1.2 Result Units ..................................................................................................................................... 11
4.1.3 Outlier Identification ....................................................................................................................... 15
4.2 Results .................................................................................................................................................... 16
4.2.1 Williams Dataset ............................................................................................................................. 16
4.2.2 W. Miller Dataset ............................................................................................................................ 26
5 Evaluate outlier samples for potential explanations ..................................................................................... 30
5.1 Review of previous Science Panel and third-party recommendations ................................................... 30
5.2 Evaluate outlier samples for potential explanations ............................................................................... 32
5.2.1 Methods........................................................................................................................................... 32
5.2.2 Results ............................................................................................................................................. 34
5.3 Imputing flux estimates .......................................................................................................................... 40
5.3.1 Methods........................................................................................................................................... 40
5.3.2 Results: Williams Dataset ............................................................................................................... 41
5.3.3 Results: W. Miller Dataset .............................................................................................................. 46
5.4 Comparing samples between studies ..................................................................................................... 48
6 Evaluate spatial interpolation among sites and attenuation of fluxes ........................................................... 56
6.1 Bird Island Sampler Results ................................................................................................................... 56
6.2 Attenuation ............................................................................................................................................. 63
6.2.1 Methods........................................................................................................................................... 63
6.2.2 Results ............................................................................................................................................. 66
7 Determine loading to Utah Lake for including in the ULNM ...................................................................... 68
7.1 Methods .................................................................................................................................................. 68
7.2 Results .................................................................................................................................................... 72

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
ii

8 Evaluate chemical speciation ........................................................................................................................ 75
9 References ..................................................................................................................................................... 77
Appendix A: AD Subgroup Analysis Plan ........................................................................................................... 79
Appendix B: TP and DIN Constituents ................................................................................................................ 86
Appendix C: Diverging Perspectives Memo and Reference Material .................................................................. 92

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
iii

Tables

Table 1. Subgroup recommendation of atmospheric nutrient loading to Utah Lake. ............................................. 6
Table 2. Summary of available atmospheric deposition datasets. ........................................................................ 11
Table 3. Weather stations located closest to each atmospheric deposition sampler. The primary weather station
was closest to the sampler and the secondary weather station was next closest to the sampler. When there
were gaps in the weather data for the primary, weather station, data from the secondary weather station
were applied. ................................................................................................................................................. 14
Table 4. Details about the data available for each dataset. .................................................................................. 34
Table 5. Multiple regression results to predict atmospheric deposition fluxes from weather conditions. ........... 43
Table 6. Cumulative annual fluxes for 2020 from the Williams dataset, with missing sampling dates imputed via
weather regression relationship..................................................................................................................... 44
Table 7. Cumulative fluxes of nutrients across sites for the W. Miller and Williams datasets. ........................... 55
Table 8. Attenuation scenarios based on information in VanCuren et al. 2012a and Wilson and Serre 2007. .... 66
Table 9. Atmospheric deposition loading estimates for Utah Lake as a result of this analysis (rows 1-4 in blue)
compared to constraining analyses (rows 5-9 in white) and other published studies (rows 10-12 in blue). 73
Table 10. Proportions of chemical constituents in DIN and TP across sites and compared to other studies. ...... 75

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
iv

Figures

Figure 1. Time series of TP fluxes for the Williams dataset. Note the differences in y-axis range for each year.
....................................................................................................................................................................... 17
Figure 2. Time series of TP fluxes for the Williams dataset. Non-outlier samples are indicated by “False” and
shown in purple. Outlier samples are indicated by “True” and shown in orange. ...................................... 18
Figure 3. Boxplots of TP fluxes grouped by year for the Williams dataset.......................................................... 19
Figure 4. Boxplots of TP fluxes with discrete values grouped by sampler location for the Williams dataset.
Non-outlier samples are indicated by “False” and shown in purple. Outlier samples are indicated by
“True” and shown in orange. ....................................................................................................................... 20
Figure 5. Time series of DIN fluxes for the Williams dataset. Note the difference in y-axis range for each year.
....................................................................................................................................................................... 21
Figure 6. Time series of DIN fluxes for the Williams dataset. Non-outlier samples are indicated by “False”
and shown in purple. Outlier samples are indicated by “True” and shown in orange. ............................... 22
Figure 7. Boxplots of DIN fluxes for the Williams dataset grouped by year and sampler location. ..................... 23
Figure 8. Boxplots of DIN fluxes with discrete values grouped by sampler location for the Williams dataset.
Non-outlier samples are indicated by “False” and shown in purple. Outlier samples are indicated by
“True” and shown in orange. ....................................................................................................................... 24
Figure 9. Relationship between TP flux and sampling interval (in days) for the Williams dataset. .................... 25
Figure 10. Relationship between DIN flux and sampling interval (in days) for the Williams dataset. ................ 25
Figure 11. Time series of TP fluxes for the W. Miller dataset. ............................................................................ 26
Figure 12. Boxplots of TP fluxes grouped by year for the W. Miller dataset. ..................................................... 27
Figure 13. Boxplots of TP fluxes with discrete values grouped by sampler location for the W. Miller dataset. . 27
Figure 14. Time series of TN fluxes for the W. Miller dataset. ............................................................................ 28
Figure 15. Boxplots of TN fluxes Grouped by Year and Sampler Location for the W. Miller dataset. ............... 29
Figure 16. Boxplots of TN fluxes Grouped by Year and Sampler Location for the W. Miller dataset. ............... 29
Figure 17. Counts of contaminated, uncontaminated, and unknown TP samples in the Williams dataset, grouped
by whether or not the sample was an outlier (true or false). ......................................................................... 35
Figure 18. Boxplots of TP flux in the Williams dataset, divided by samples that were contaminated,
uncontaminated, and unknown. .................................................................................................................... 35
Figure 19. Counts of contaminated, uncontaminated, and unknown DIN samples in the Williams dataset,
grouped by whether or not the sample was an outlier. ................................................................................. 36
Figure 20. Boxplots of DIN flux in the Williams dataset, divided by samples that were contaminated,
uncontaminated, and unknown. .................................................................................................................... 36
Figure 21. Time series of uncontaminated TP samples in the Williams dataset. ................................................. 37
Figure 22. Time series of uncontaminated DIN samples in the Williams dataset. ............................................... 38
Figure 23. Boxplots of uncontaminated TP samples in the Williams dataset. ..................................................... 38
Figure 24. Boxplots of uncontaminated DIN samples in the Williams dataset. ................................................... 39
Figure 25. Flux, precipitation, and average daily wind speed at the weather stations associated with the
atmospheric deposition sampling sites in the Williams dataset. TP fluxes for the same time period are
displayed for additional context. ................................................................................................................... 42
Figure 26. Cumulative TP flux for the Williams dataset, with gaps in sampling dates imputed by linear
interpolation (solid lines) and weather regression (dotted lines). Note that the two imputation approaches
were equivalent for Orem and Pump Station, so the solid and dotted lines overlap. ................................... 44
Figure 27. Cumulative DIN flux for the Williams uncontaminated dataset, with gaps in sampling dates imputed
by linear interpolation (solid lines) and weather regression (dotted lines). Note that the two imputation

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
v

approaches were equivalent for Orem and Pump Station, so the solid and dotted lines overlap. ............... 45
Figure 28. Counts of the ratio of net precipitation (cumulative precipitation minus evaporation) to cumulative
precipitation for sampling events in the W. Miller dataset. .......................................................................... 47
Figure 29. Log scale boxplots of TP flux from the W. Miller and Williams datasets. The star on each plot
represents the mean of the data and the median is displayed as the solid horizontal line in the box. .......... 49
Figure 30. Log scale boxplots of TP flux across sites from the W. Miller and Williams datasets. The star on
each plot represents the mean of the data and the median is displayed as the solid horizontal line in the box.
Stations are organized in clockwise order across the lake. ........................................................................... 50
Figure 31. Log scale boxplots of DIN and TN flux from the W. Miller and Williams datasets. The star on each
plot represents the mean of the data and the median is displayed as the solid horizontal line in the box. ... 50
Figure 32. Log scale boxplots of DIN and TN flux across sites from the W. Miller and Williams datasets. The
star on each plot represents the mean of the data and the median is displayed as the solid horizontal line in
the box. Stations are organized in clockwise order across the lake. ............................................................. 51
Figure 33. Time series of TP fluxes in the Williams dataset, with Bird Island fluxes highlighted (purple)
compared to other sites (gray)....................................................................................................................... 57
Figure 34. Time series of DIN fluxes in the Williams dataset, with Bird Island fluxes highlighted (purple)
compared to other sites (gray)....................................................................................................................... 58
Figure 35. Cumulative TP fluxes for the Williams dataset, starting on the date when the Bird Island sampler was
installed. ........................................................................................................................................................ 59
Figure 36. Cumulative DIN fluxes for the Williams dataset, starting on the date when the Bird Island sampler
was installed. ................................................................................................................................................. 60
Figure 37. Wind rose data for seven weather stations located around Utah Lake. ............................................... 61
Figure 38. Illustration of attenuation from VanCuren et al. 2012a (relevant particle size: 10-25 µm) and Wilson
and Serre 2007. Reproduced from Figure 11 and Figure 3 of the respective references. ............................. 65
Figure 39. Inverse distance weighted (IDW) spatial interpolation of shoreline fluxes of TP (left) and DIN (right)
based on observations at the four sampling sites in the Williams dataset for 2020. Sampling sites
(clockwise starting on the east side of the lake) were Orem, Lakeshore, Mosida, and Pump Station. ........ 70
Figure 40. Display of the estimate of TP (left) and DIN (right) loading to Utah Lake, which incorporates
shoreline fluxes (local and regional atmospheric deposition sources) at the edge of the lake that attenuate to
a regional flux moving toward the middle of the lake. ................................................................................. 71
Figure 41. Inverse distance weighted (IDW) spatial interpolation of shoreline fluxes of TP (left) and DIN (right)
across the lake, thus representing a “no attenuation” scenario. The flux values are based on observations at
the four sampling sites in the Williams dataset for 2020. Sampling sites (clockwise starting on the east side
of the lake) were Orem, Lakeshore, Mosida, and Pump Station. ................................................................. 72
Figure 42. Time series of SRP samples from the raw Williams dataset. .............................................................. 86
Figure 43. Time series of SRP samples from the processed Williams dataset, excluding any samples that
included insect contamination or did not have metadata available............................................................... 87
Figure 44. SRP to TP ratios from the 2020 Williams dataset. Values >1 area a functional impossibility because
SRP is a component of TP. ........................................................................................................................... 87
Figure 45. Time series of nitrate samples from the raw Williams dataset ............................................................ 88
Figure 46. Time series of ammonium samples from the raw Williams dataset .................................................... 89
Figure 47. Nitrate to DIN ratios from the 2020 Williams dataset. ....................................................................... 90
Figure 48. Ammonium to DIN ratios from the 2020 Williams dataset. ............................................................... 91

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
vi

Definitions

 Bulk deposition – Deposition of atmospheric materials via wet and dry pathways
 Wet deposition – Deposition of atmospheric gases and aerosols that have been absorbed and/or
intercepted by precipitation.
 Dry deposition – Deposition of atmospheric particles and aerosols directly onto a surface, driven by
winds and gravity
 Flux - Flux is a flow of mass across (or upon) an area and as such, it is a rate, expressed herein as
mass/area/time. Measured as mass deposited on a known area over a known amount of time.
 Outlier – Observation that lies an abnormal distance from other values in a population
 Local – In the context of this study, deposition that occurs near the source of the material. Local scale
was considered as within the Utah Lake catchment or smaller.
 Regional – In the context of this study, deposition whose source is outside the Utah Lake catchment.
 Attenuation – Reduction of effect or value. For this study, attenuation refers to a potential reduction
in atmospheric deposition moving away in space from the source of the material being deposited.
 ULNM – Utah Lake Nutrient Model. The Utah Lake Nutrient Model is a suite of two linked models
to address lake hydrodynamics and sediment transport (Environmental Fluids Dynamic Code (EFDC))
and water quality (Water Quality Analysis Simulation Program (WASP)).
 N - Nitrogen
 P - Phosphorus
 TP - Total phosphorus
 DIN - Dissolved inorganic nitrogen
 SRP - Soluble reactive phosphorus

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
1

1 Executive Summary
Mass balances of the nitrogen (N) and phosphorus (P) inputs to the lake must be defined to
inform the construction and calibration of in-lake water quality models. Nutrient inputs to the
lake include surface water, groundwater, and atmospheric deposition. As part of the effort to
establish mass balances for N and P in Utah Lake, the Science Panel formed a Subgroup to
determine N and P mass loading rates from atmospheric deposition. The work of the Subgroup
consisted of reviewing, analyzing, and resolving information and data from both the scientific
literature and Utah Lake specific studies on nutrients and related topics involving atmospheric
deposition. Through these efforts, the Subgroup derived recommendations for these important
loading rates.
The Atmospheric Deposition Subgroup work was approached with these objectives:
 Analyze available information and data to improve understanding of nutrient loadings
by atmospheric deposition to Utah Lake;
 Work collaboratively toward a recommendation for N and P atmospheric loading rates;
and
 Document the Science Panel’s decision-making process for analyzing and evaluating
evidence resulting in their recommendation.
The Subgroup developed and agreed to the Atmospheric Deposition Subgroup Analysis Plan
(Appendix A), establishing a process to accomplish these objectives and derive an atmospheric
deposition loading recommendation. The analysis plan guided the Subgroup throughout the
process following these steps:
1. Review and summarize available data.
2. Evaluate potential explanations of high magnitude results.
3. Evaluate fluxes among shoreline sampling sites and potential attenuation of fluxes
moving into Utah Lake.
4. Evaluate the chemical speciation of atmospheric deposition nutrient loads.
5. Compare direct estimates of atmospheric deposition to other constraining analyses.
6. Determine atmospheric deposition loading estimates to Utah Lake.
Each of these six steps will be summarized in this section. Details of how each step was
executed and Subgroup decisions are detailed in full in the following report.
Analysis Plan steps 1-2: Review and summarize available data and evaluate high magnitude
results.
Data were obtained from four separate atmospheric deposition studies that were led by Dr. Gus
Williams and Dr. Wood Miller. These studies were conducted from 2017 through 2020,
employed multiple collection methodologies, and resulted in a range of loading estimates.
Nutrient data from these studies included total phosphorus (TP), dissolved inorganic nitrogen
(DIN), and the nutrient constituents of soluble reactive phosphorus (SRP), nitrate, and

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
2

ammonium.
The raw data from these studies were compiled, reviewed and summarized. Calculations were
used to convert raw data to an equal basis (area flux rate) for comparison and application. Data
quality reviews and analyses were used to assess individual data points that warranted further
investigation (called “outliers” herein). Outliers were investigated for insect and debris
contamination, weather events, and local atmospheric deposition sources. Data gaps were
resolved using a best fit regression analysis.
As a result of the data quality review, key decisions were made toward processing the data for
further analysis:
 Sample concentrations were converted to area-based flux rates for the W. Miller dataset
to facilitate comparison to the Williams dataset. Concentration data were paired with the
corresponding precipitation data from the nearest precipitation gauge.
 Insects were acknowledged as a potential nutrient import (e.g., terrestrial insects entering
the lake) and export (e.g., aquatic insects leaving the lake) as well as a source of recycled
nutrients (e.g., aquatic insects leaving and reentering the lake). The Subgroup decided
that: 1) insects and atmospheric deposition represent two independent sources of
nutrients to Utah Lake and should be considered separately; and 2) an intentionally
designed study or analysis would need to be completed to quantify the influence of
insects on the Utah Lake nutrient budget. The subgroup decided that insects were
considered contamination to an atmospheric deposition sample and samples containing
insects were removed from consideration.
 No data points that were flagged as outliers were removed from the analysis unless
insect contamination was present or metadata was not available to demonstrate insect
presence or absence.
 A regression model using local weather data was used to fill data gaps and generate a
complete time series.
 Following analysis and comparison of the Williams and W. Miller datasets, the Williams
dataset was used for estimating lake-wide atmospheric deposition loadings. The W.
Miller dataset was used comparatively but not quantitatively in estimating nutrient fluxes
or loading rates.
Analysis Plan step 3: Evaluate fluxes among shoreline sampling sites and potential
attenuation of fluxes moving into Utah Lake.
Cumulative annual N and P fluxes were computed for each sampling site by integration of
individual shoreline sampler results over time. To determine how shoreline sampling site fluxes
translate into loadings across the full area of the lake, several methods were considered: 1)
simple averaging of fluxes; 2) “raster” estimates wherein fluxes from each site were applied to a
proportional area of the lake; 3) load attenuation of shoreline fluxes; and 4) load attenuation of
regional sources. Attenuation in the context of this report is the decline of atmospheric
deposition flux as distance increases from the shoreline to the center of the lake or as distance

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
3

increases from the primary source.
Previous Utah Lake atmospheric deposition studies modeled significant attenuation moving
away from the lake shoreline (Olsen et al. 2018, Reidhead 2019, Brahney 2019). However,
these studies did not directly measure attenuation or deposition rates in the interior of the lake.
A sampler was installed on Bird Island in May 2020 to provide direct measurements of interior
deposition rates. Barrus et al. (2021) analyzed results from the Bird Island sampler and
concluded that attenuation was not occurring. This contradiction between the Olsen et al.
(2018), Reidhead (2019), and Brahney (2019) results (which predicted offshore attenuation) and
Barrus et al.’s (2021) results (which did not detect offshore attenuation) required additional
literature review and Subgroup analysis.
A literature review of similar studies reinforced the scientific expectation that offshore fluxes at
Bird Island would be less than or equal to fluxes on the shoreline (Jassby et. al 1994 and
VanCuren et al. 2012a). Jassby et. al (1994) sampled wet and dry deposition of nitrogen and
phosphorus in and around Lake Tahoe and found that deposition decreases significantly moving
toward the middle of the Lake. A later study of Lake Tahoe by VanCuren et al. (2012a and
2012b) found that deposition rates for regional sources were consistent across the lake, but
deposition rates for local sources (such as urban and near-shore dust) decreased exponentially
moving away from the source origin. VanCuren et al. also found that the exponential decrease
of local source deposition was due to the material particle size: larger local particles decayed
more rapidly than smaller regionally sourced particles.
The Subgroup also analyzed raw Bird Island data. In contrast to Barrus et al. (2021), they found
that Bird Island daily and cumulative N and P fluxes were higher than at the shoreline samplers
for the corresponding sampling period. Since these findings contradicted the results of the
literature review and the findings of Barrus et al. (2021), two additional hypotheses were
evaluated:
 The first hypothesis was that shoreline fluxes from an unknown area southeast of the
lake contributed to higher fluxes at Bird Island. Atmospheric deposition samplers are not
currently installed in this region to identify the origin of the source, quantify the flux
magnitude, or to inform the potential transport mechanisms that would explain the
higher Bird Island fluxes. Ultimately, this hypothesis was evaluated using local weather
station wind data and the potential source of additional nutrient deposition was not
identified.
 The second hypothesis was that Bird Island collects nutrients from an in-lake source
such as bird droppings, volatilized nutrients from the island, or lake water spray during
wind events. Quality assurance data and field metadata were requested to evaluate these
mechanisms as a potential explanation; however, this information was not available and
the Subgroup could not rule out this hypothesis.
The Subgroup found that the higher fluxes observed at Bird Island could not be sufficiently
explained, and thus concluded the Bird Island sampler would not be used to evaluate attenuation

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
4

of fluxes across the lake. Rather, information from the atmospheric deposition literature were
used to estimate the potential attenuation rates.
Two types of atmospheric deposition sources were calculated in this analysis: 1) regional
sources originating from the Sevier Dry Lake, Great Lake Salt Lake, and West Desert playas;
and 2) local sources originating from areas adjacent to Utah Lake. The shoreline samplers used
by Williams and W. Miller collected both local and regional deposition so it was not possible to
resolve the measured amounts of these sources. Instead, literature from the Utah Lake basin
(Goodman et al. 2022, Putman et. al 2022) was used to estimate regional deposition. Pollutants
from local sources of atmospheric deposition are expected to attenuate as they move across the
lake, and potential attenuation distances of 100-2,000 m were applied to Subgroup calculations.
Pollutants from regional sources are not expected to attenuate and were calculated as evenly
distributed across Utah Lake (Wilson and Serre 2007, VanCuren et al. 2012, Jassby et al. 1994,
Goodman et al. 2022).

Analysis Plan steps 4-6: Evaluate chemical speciation, compare estimates of atmospheric deposition,
and calculate loading estimates
The chemical speciation of TP and DIN was calculated from direct measurements in the
Williams dataset, representing an average of 37.5% of TP as SRP, 20.25% of DIN as nitrate,
and 69.75% of DIN as ammonium. These measured constituents did not match completely with
the input data needed to characterize the Utah Lake Nutrient Model (including organic N and
P). The Subgroup recommended that the specifics on implementing the observed proportions
were to be determined by the modeling team.
Atmospheric deposition loads to Utah Lake were calculated using Geographic Information
Systems (GIS) to: 1) determine the proportional contribution of local and regional fluxes for
shoreline and open water locations; 2) calculate the total N and P flux across the lake by
summing flux rates across the GIS grid layer; and 3) calculate an additional scenario assuming
no attenuation.
The majority of Subgroup members recommended that atmospheric deposition loading to Utah
Lake be 32 metric tons/yr TP and 220 metric tons/yr DIN, with a potential range of 31-45
metric tons/yr TP and 218-249 metric tons/yr DIN that could be evaluated as part of a model
sensitivity analysis (

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
5

Table 1). These ranges reflected different attenuation distances.
One Subgroup member did not support the Subgroup’s recommendation and was invited to
provide a memo with their diverging perspectives. The memo and reference material are
included in Appendix C of this report.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
6

Table 1. Subgroup recommendation of atmospheric nutrient loading to Utah Lake.
Recommendation Attenuation
Distance
(m)
TP
(metric
tons/
yr)
SRP:
TP
DIN
(metric
tons/yr)
Nitrate/
DIN
Ammonium
/DIN
Organic
N/DIN
Low 100 31 37.5% 218 30.25% 69.75% Unknown
Primary 200 32 37.5% 220 30.25% 69.75% Unknown
High 2000 45 37.5% 249 30.25% 69.75% Unknown

In summary, understanding the role of atmospheric deposition in the nutrient balance of Utah
Lake is important for successful modeling and nutrient criteria development. In recognition of
this complex issue, a Subgroup of subject matter experts was organized to engage in analysis of
the existing data and relevant scientific literature. This process was guided by an analysis plan
detailed in the following report. Where uncertainty existed, the Subgroup relied on objective
peer-reviewed publications to assure data integrity and draw appropriate conclusions. The
results of the Subgroup analysis are a range of recommended N and P atmospheric deposition
values for Utah Lake that represent the most scientifically rigorous conclusions possible from
the available data.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
7
2 Background
Characterizing the mass balance of nitrogen (N) and phosphorus (P) entering and exiting Utah
Lake is fundamental to the Science Panel’s understanding of nutrient processing within the lake.
It is essential for constructing and calibrating in-lake water quality models, and deriving N and P
water quality criteria. Nutrient inputs to Utah Lake come from groundwater inflow, tributary and
overland inflows, precipitation, and atmospheric deposition. A recent analysis commissioned by
the Science Panel (Tetra Tech, 2021) computed a mass balance for the groundwater and tributary
sources.
Studies to characterize wet and dry atmospheric deposition of nutrients to Utah Lake were
initiated by the Wasatch Front Water Quality Council (WFWQC) in conjunction with Brigham
Young University in 2018, with initial results presented to the Science Panel in spring of 2019.
Throughout 2019 and the first half of 2020, the Science Panel worked with the WFWQC,
primarily through discussions with Science Panel member Dr. Theron Miller, to guide ongoing
and future atmospheric deposition monitoring. That effort resulted in several work products: 1) a
preliminary atmospheric deposition load estimate; 2) an atmospheric deposition monitoring plan;
and 3) a set of recommendations from the Science Panel to the WFWQC to guide the
atmospheric deposition monitoring program (ULWQS Science Panel, 2020a). A chronological
accounting of these discussions and resulting products was provided to the Steering Committee
on May 28, 2020 (ULWQS Science Panel, 2020b).
In 2022, the Science Panel revisited their atmospheric recommendation with new data presented
by WFWQC. On March 3, 2022, Dr. Theron Miller, Dr. Wood Miller, and Dr. Gus Williams
presented the results from their studies to the Science Panel. Over several meetings, Science
Panel members reviewed atmospheric deposition studies and attempted to establish atmospheric
deposition values for P and N for calibration of the Utah Lake in-lake water quality model. At
the August 3, 2022 Science Panel meeting, Science Panel members agreed to form a subgroup
that would meet regularly to review assumptions, aggregate and analyze available atmospheric
deposition data, and recommend atmospheric deposition N and P loading rates for Utah Lake.
The ULWQS Atmospheric Deposition Subgroup members included Dr. Mike Brett, Dr. Mitch
Hogsett, Dr. Theron Miller, and Dr. Hans Paerl. The Subgroup met from August 18, 2022
through February 23, 2023. In total, Subgroup members attended 19 meetings to review, discuss,
and analyze atmospheric deposition literature, data, and reports. This report documents the
Subgroup’s process, findings, and recommendations for calculating nutrient atmospheric
deposition loading rates to Utah Lake. This report also documents when Subgroup members
were able to reach consensus on analysis decisions, when there were diverging perspectives
within the Subgroup, and the reasons why.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
8
3 Analysis Plan Development
The Subgroup was tasked with the following objectives:
1. Analyze available information and data to improve understanding of atmospheric
deposition to Utah Lake.
2. Work collaboratively toward a recommendation for atmospheric loadings, ideally
achieved through consensus.
3. Document the Science Panel’s decision-making process for analyzing and evaluating
evidence and working toward atmospheric deposition recommendations.
In order to achieve these objectives, the Subgroup developed and agreed to a rigorous analysis
plan. The purpose of the analysis plan was to establish the process by which atmospheric loading
recommendations would be generated, prior to beginning the work. The detailed analysis plan is
included in Appendix A. To summarize, the analysis plan included the following steps:
1. Review and summarize data from atmospheric deposition samplers around Utah Lake
2. Evaluate “outlier” samples for potential explanations such as collection methodology,
contamination, weather events, and local sources. Review and discuss previous Science
Panel and third-party recommendations for interpreting atmospheric deposition data
3. Evaluate spatial interpolation among shoreline sampling sites and evaluate potential
attenuation of fluxes moving into Utah Lake
4. Evaluate the chemical speciation of total nutrient atmospheric deposition loads
5. Compare direct estimates of atmospheric deposition to other constraining analyses
6. Determine atmospheric deposition loading estimates to Utah Lake
The following sections of this report will detail the methods and results of each step of the above
analysis plan.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
9
4 Review and Summarize Data
4.1 Methods
4.1.1 Data Compilation and Organization
Atmospheric deposition data from four independent studies were acquired from Drs. Gus
Williams and Wood Miller. The data set from Dr. Williams included data from Olsen et al.
(2018), Reidhead (2019), and Barrus et al. (2021). Samples from these studies were collected
between 2018 and 2020. The Williams data set was collected using a sampler design that
measured wet and dry deposition independently. However, Dr. Williams reported that due to
sampler error, only bulk deposition data was available to share with the Subgroup. Most
samplers in the data set provided by Dr. Williams were unscreened. Screening a sampler is a
technique used to prevent insects and other materials from entering the sampler. Screens were
installed on the Williams samplers starting on 21 May 2020.

The Miller data was collected using a bulk deposition precipitation sampler design that did not
differentiate between wet and dry deposition. All samplers in the Miller dataset were
unscreened.
The two datasets were evaluated separately and are referred to as the “Williams” and “W.
Miller” datasets herein. Analyses in this document focus on areas around Utah Lake and thus
include all sites from the W. Miller dataset and all sites from the Williams dataset except Central
Davis Low and Ambassador, which were located near Salt Lake City.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
10
Table 2 presents a summary of the data used for this analysis.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
11
Table 2. Summary of available atmospheric deposition datasets.
Data
Provider
Study Period of
Record
Sampler
Design
Screened
(Y/N)
Data type
Provided
Metadata
Provided
(Y/N)
G. Williams Olsen et. al
2018
5/11/2017 to
12/4/2017
Wet/Dry1 N Bulk2 Y
G. Williams Reidhead
2018
4/6/2018 to
11/17/2018
Wet/Dry1 N Bulk2 N
G. Williams Reidhead,
unpublished
11/30/2018 to
5/18/2019
Wet/Dry1 N Bulk2 N
G. Williams Barrus et. al
2021
5/10/2019 to
12/16/2020
Wet/Dry1 Partial Bulk2 Y
W. Miller W. Miller 1/2/2017 to
12/17/2020
Bulk2 N Bulk2 N
1. Wet/Dry sample collectors independently sample precipitation derived atmospheric deposition (i.e., wet deposition) and dry
weather particulate deposition (i.e., dry deposition). Bulk samplers collect a combined sample and do not differentiate between
wet and dry deposition.
2. Bulk deposition data is the sum of wet and dry deposition. The wet/dry sample collectors noted in the table did not produce
independent measurements of wet and dry deposition, but rather combined the two components into one bulk deposition
measurement. Therefore, the data available to this analysis cannot be used to accurately estimate the wet and dry deposition
components of a bulk sample.
Data were organized and analyzed in the R statistical programming language. The data were
unchanged from the raw data in the excel spreadsheets received, with the exception of compiling
data from different sites into a single spreadsheet, adding columns for date information, and
conversions to flux (mg/m2/day). The Williams dataset included measurements listed as 0
mg/m2, which were associated with non- detect concentrations of nutrients. The Williams
research team confirmed there was no method to convert a detection limit-based concentration to
an area-based flux, so the values were retained as-is. The W. Miller dataset contained detection
limit information, and non-detects were set at ½ the detection limit.

Decision Point: Assigning non-detect values
All Subgroup members agreed with assigning non-detect values at 0 mg/m2. Subgroup members
discussed that there were relatively few results in the Williams dataset reported at 0 mg/m2, and
retaining the values as 0 mg/m2 would not significantly affect the calculated area-based flux for
each site.
4.1.2 Result Units
The nutrient deposition units for the Williams dataset were reported in mg/m2, which were
converted to mg/m2/d and mg/m2/wk. Most Williams samples were collected on weekly
timescales, but this was not always the case. Thus, total accumulation was calculated by dividing
Williams’ nutrient deposition results by the time interval of each sampling period (daily flux) or
the fractional week (weekly flux). Flux is a mass flow rate across (or upon) an area and as such,

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
12
it is a rate, expressed herein as mass/area/time.
W. Miller reported sample results units in mg/L and calculated flux using paired precipitation
data from a single weather station located in Lehi, UT. Sample volumes were not reported for the
W. Miller dataset. For the Subgroup analysis, W. Miller dataset fluxes were recalculated using
proximate precipitation stations. The Science Panel Subgroup identified several precipitation
samplers in the area surrounding Utah Lake and paired sampling stations with the nearest
precipitation sampler (

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
13
Table 3). The depth of precipitation accumulated over the course of a sampling event was
calculated, and this value was used to convert volume to area for the flux values. Conversions
were calculated using the relevant diameters of the sampler (collector: 20 in diameter; container:
4 in diameter). This adjustment aligned units between the Williams and W. Miller datasets.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
14
Table 3. Weather stations located closest to each atmospheric deposition sampler. The
primary weather station was closest to the sampler and the secondary weather station was
next closest to the sampler. When there were gaps in the weather data for the primary,
weather station, data from the secondary weather station were applied.
AD Station Dataset Primary
Weather
Station
Primary
Station ID
Secondary
Weather
Station
Secondary
Station ID
Orem W. Miller I-15 at Orem UTORM Provo
Municipal
Airport
KVPU
Orem Williams I-15 at Orem UTORM Provo
Municipal
Airport
KVPU
BYU W. Miller Provo, BYU USC00427064 Eyring
Science
Center
EYSC
Spanish
Fork

W. Miller
Spanish Fork
Power House

USC00428119 EW2355
Spanish Fork

UKBKB
Lake Shore Williams Lincoln Point FG015
Lincoln
Point W. Miller Lincoln Point FG015
Bird Island Williams Lincoln Point FG015
Genola W. Miller Genola South FG019 Genola FG004
Elberta W. Miller Genola South FG019 Goshen FG014
Mosida W. Miller SR-68 at MP
16
Mosida
UTLAK Genola FG004
Mosida Williams SR-68 at MP
16
Mosida
UTLAK Genola FG004
Pelican
Point W. Miller SR-68 at MP
16
Mosida
UTLAK Utah Lake,
Lehi USC00428973
Saratoga
Springs Williams SR-68 at MP
16
Mosida
UTLAK Utah Lake,
Lehi USC00428973
Lehi W. Miller Utah Lake,
Lehi USC00428973 Pioneer
Crossing
UTPCR
Pump
Station Williams Utah Lake,
Lehi USC00428973 Pioneer
Crossing
UTPCR

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
15
Decision Point: Converting W. Miller volume-based deposition(mg/L) to area-based fluxes
(mg/m2/day)
W. Miller (2021) estimated flux values using the precipitation measurements from a single
precipitation gauge. All Subgroup members agreed that calculating area-based flux values for the
W. Miller dataset would be more representative of local conditions when the sampling stations
were paired with the data from the nearest possible precipitation sampler. Subgroup members
used a map to identify the primary and secondary weather stations associated with each W. Miller
sampling station to generate precipitation values to calculate an area-based flux, hereafter
referred to as the final W. Miller dataset.
4.1.3 Outlier Identification
An outlier analysis was performed for each chemical constituent in the Williams dataset to
identify deposition events for further evaluation. The interquartile range (IQR) is defined as the
span between the 25th and 75th percentile. Upper limit outliers were flagged when a result
exceeded the 75th percentile + 1.5*IQR from the whole dataset. Lower limit outliers were
flagged when a result was less than the 25th percentile – 1.5*IQR.
IQR analysis was applied to the entire Williams dataset within yearly sampling intervals. The
application of this IQR approach to identify outliers yielded a low outlier range that was entirely
negative values. The IQR approach did not result in the identification of any low outliers. An
upper outlier range was determined and used to identify and explore unusually high data points
Flagging outliers in this way was used as an exploratory step and not a firm rule that dictated an
action or decision. Additional analyses to determine potential explanations of outliers (e.g., wind
and precipitation events, local source contributions, and insect presence) are presented in
subsequent sections of this report. An outlier analysis was not performed on the W. Miller
dataset.
Decision Point: Identifying outliers
The Subgroup members considered several potential methods for identifying outliers, including
the IQR approach, an extreme value analysis, and the assignment of a specific flux as a cutoff for
outliers. All Subgroup members agreed to identify outliers using the defined IQR approach. The
rationale behind the decision was: 1) the Williams dataset is not normally distributed; 2) the
episodic nature of atmospheric deposition events skewed the distribution; and 3) the Science Panel
Subgroup concluded that the IQR approach was most statistically robust for identifying outliers.
All Subgroup members also agreed to apply the IQR approach to the entire Williams dataset rather
than on a site-by-site basis. This decision was made with the understanding that Subgroup
members could later apply the IQR approach to identify outliers in site-specific datasets if there
was interest in exploring specific locations more in-depth.
The decision on the approach to identify outliers was not a decision on whether to include or
exclude outliers within the dataset. This decision was an exploratory step to identify extreme
values for more detailed consideration.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
16
4.2 Results
4.2.1 Williams Dataset
Figure 1 through Figure 8 show all total phosphorus (TP) and dissolved inorganic nitrogen (DIN)
results for the Williams dataset. Large flux events for TP and DIN were typically associated with
individual stations, usually Saratoga Springs, Lakeshore, and Mosida. At some times, high flux
events occurred at all sites, while at other times they occurred at only one site.
TP outliers were considered for samples exceeding 2 to 3 mg/m2/d, depending on the sampling
interval (Figure 2 and Figure 4). DIN outliers were considered for samples exceeding 15
mg/m2/d, with variability depending on the sampling interval (Figure 9 and Figure 10). Further
investigation of outlier samples is presented in subsequent sections of this report.
Data were also available and analyzed for nitrate, ammonium, and soluble reactive P (SRP) for
the Williams dataset, but this report focuses on the total constituents. Graphs of SRP, nitrate, and
ammonium are provided in Appendix B.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
17
Figure 1. Time series of TP fluxes for the Williams dataset. Note the differences in y-axis range for each
year.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
18

Figure 2. Time series of TP fluxes for the Williams dataset. Non-outlier samples are indicated by
“False” and shown in purple. Outlier samples are indicated by “True” and shown in orange.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
19

Figure 3. Boxplots of TP fluxes grouped by year for the Williams dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
20

Figure 4. Boxplots of TP fluxes with discrete values grouped by sampler location for the
Williams dataset. Non-outlier samples are indicated by “False” and shown in purple. Outlier samples are indicated by
“True” and shown in orange.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
21

Figure 5. Time series of DIN fluxes for the Williams dataset. Note the difference in y-axis range for each
year.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
22

Figure 6. Time series of DIN fluxes for the Williams dataset. Non-outlier samples are indicated by
“False” and shown in purple. Outlier samples are indicated by “True” and shown in orange.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
23

Figure 7. Boxplots of DIN fluxes for the Williams dataset grouped by year and sampler
location.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
24

Figure 8. Boxplots of DIN fluxes with discrete values grouped by sampler location for the
Williams dataset. Non-outlier samples are indicated by “False” and shown in purple. Outlier samples are indicated by
“True” and shown in orange.
Sampling protocol for the Williams dataset was for a weekly sampling interval, however, weekly
sampling did not always occur. The Science Panel Subgroup evaluated the influence of sampling
interval on flux magnitude and determined that time interval between sampling events did not
appear to impact flux. Specifically, samples taken on an interval longer than one week were not
associated with systematically higher or lower fluxes than weekly samples (Figure 9, Figure 10).
This observation may be due to the importance of event driven fluxes and influence of local
sources as discussed later in this report.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
25

Figure 9. Relationship between TP flux and sampling interval (in days) for the Williams
dataset.

Figure 10. Relationship between DIN flux and sampling interval (in days) for the Williams
dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
26
4.2.2 W. Miller Dataset
Figure 11 through Figure 16 show all total phosphorus (TP) and dissolved total nitrogen (TN)
results for the W. Miller dataset. Large flux events for TP and TN were typically associated with
individual stations rather than being consistent between stations. Several sampling events were
associated with large time gaps. Metadata was not available for more detailed assessment of the
gaps.

Figure 11. Time series of TP fluxes for the W. Miller dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
27

Figure 12. Boxplots of TP fluxes grouped by year for the W. Miller dataset.

Figure 13. Boxplots of TP fluxes with discrete values grouped by sampler location for the
W. Miller dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
28

Figure 14. Time series of TN fluxes for the W. Miller dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
29

Figure 15. Boxplots of TN fluxes Grouped by Year and Sampler Location for the W. Miller
dataset.

Figure 16. Boxplots of TN fluxes Grouped by Year and Sampler Location for the W. Miller
dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
30
5 Evaluate outlier samples for potential explanations
Outliers identified by the IQR analysis described in 4.1.3 were examined for data integrity as
described in the following sections. Subgroup members determined that high outlier flux values
could occur for several reasons: (a) presence of materials such as insects, (b) influence of local
nutrient sources, or (c) influence of high deposition events such as wind or rain storms. Of
particular concern to calculating accurate atmospheric deposition rates, was the presence of
external materials.
5.1 Review of previous Science Panel and third-party recommendations
Previous studies (Olsen et al. 2018, Barrus et al. 2021), Science Panel recommendations (Utah
Lake Science Panel 2020a, Utah Lake Science Panel 2020b), and external reviews of available
Utah Lake atmospheric deposition studies and reports from Dr. David Gay (Gay 2019a, Gay
2019b) addressed item (a), the presence of materials such as insects.
Olsen et al. (2018) labeled a sample as “contaminated” if samplers had visible contamination
(bird droppings, insects, plant matter, and algal growth). The authors described this
contamination as follows:
“The largest dry deposition rates occurred at Saratoga Springs during the summer
months with rates significantly higher than any of the other sites (see Figure 4). We
attribute some of these high values to a terrestrial bee, Halictidae Lasioglossum. During
the summer period, sample buckets had numerous bee bodies in the water. As noted
above, these bodies were removed before sample analysis, but having been present in the
water during the week, they significantly raise the amount of nutrients in the samples.”

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
31

Olsen et al. (2018) did not employ sampler screens to reduce insect contamination and found that
loading (ton/yr) associated with contaminated samples were 44 and 10 times higher for TP and
DIN loads, respectively, than those associated with uncontaminated samples.
Barrus et al. (2021) also recorded a high prevalence of insects in unscreened samplers,
particularly at the Mosida location in the summer months. The majority of outlier samples in
Barrus et al. (2021) were associated with large numbers of visible insects in the samples.
The sampler collectors used by Barrus et al. (2021) were screened starting on May 21, 2020. TP
and DIN fluxes were significantly higher in samplers that were not equipped with screens than in
samplers equipped with 500 µm screens (paired t-test; TP: avg. difference in means = 0.36
mg/m2/d, p < 0.0018, DIN: avg. difference in means = 0.26 mg/m2/d, p < 0.0116). From a subset
of the Barrus data, Richards (2022) described a significant difference between the atmospheric
depositions measured with screened and unscreened samplers at Orem, i.e., lower atmospheric
depositions were measured when screens were used (mixed effects negative binomial regression;
TP: difference in means = 8.43 mg/m2, p < 0.01; DIN: difference in means = 22.1 mg/m2, p <
0.001).
In Science Panel Comments Regarding Wasatch Front Water Quality Council’s Atmospheric
Deposition Study (Utah Lake Science Panel 2019), the Science Panel recommended that midge
biomass and other insect biomass not be considered as atmospheric deposition flux to Utah Lake.
Though insect biomass moving into or out of the lake may represent an import or export of
nutrients, it was recommended that any insect flux should be considered separate from rates of
atmospheric deposition.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
32
Decision Point: Considering insects as contamination
Subgroup members discussed whether insects or insect parts in sampling buckets should be
considered contamination. The subgroup did not reach a consensus on this decision point. The
majority of subgroup members considered insects as contamination. However, One Subgroup
member did not agree with the majority of the Subgroup. Their perspective was that insects should
not be considered contamination in the sampling buckets. Their rationale was that insects
contribute to the nutrient budget of Utah Lake and should be considered a legitimate nutrient
source in the analysis. The majority of Subgroup members agreed that insect or insect parts in
sampling buckets should be considered contamination. The Subgroup members that agreed with
this perspective acknowledged that terrestrial insects do fall on lakes, but concluded that the
nutrient content from insects or insect parts in these samplers is not representative of the insect
contribution to the lake’s nutrient loading.
Additionally, the Subgroup members in the majority stated that there is uncertainty about whether
insects in the sampling buckets are terrestrial or aquatic in origin. Terrestrial insects are not part
of the Utah Lake system, so their parts in a sampling bucket would be considered a net import of
nutrients to the lake. Since aquatic insects are a part of the Utah Lake system, their parts in the
sampling bucket would represent a net export of nutrients from the lake. If insects are considered
important to the nutrient budget of Utah Lake, the Subgroup members suggested there be a study
intentionally designed to provide a better estimate of the influx and efflux of insects to Utah Lake.

5.2 Evaluate outlier samples for potential explanations
5.2.1 Methods
As stated above, high outlier flux values could occur for several reasons: (a) presence of
materials such as insects, (b) influence of local nutrient sources, or (c) influence of high
deposition events such as wind or rain storms. To separate the potential causes of high outlier
fluxes, field metadata were reviewed for the Williams dataset. Metadata were provided for the
2017 (Olsen et al. 2018) and 2020 (Barrus et al. 2021) study periods (see

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
33
Table 4). The metadata identified insect and plant matter presence in samples. If a sample
contained insect matter, it was considered contaminated per Science Panel Subgroup
recommendations. Metadata were not available for the 2018 and 2019 datasets. The Science
Panel Subgroup recommended those data not be used because contamination could not be ruled
out in the unscreened samplers. The influences of high deposition events such wind or rain are
evaluated in section 5.32. The influence of local sources (b) was not evaluated due to limited
information available to describe local sources.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
34
Table 4. Details about the data available for each dataset.
Study Year(s) Number
of
stations
Constituents Sampler
Design
Result
type
Metadata
availability
Williams
– Olsen
et a. 2018
2017 5 TP, DIN,
nitrate,
ammonium,
SRP
Wet/dry Bulk Yes
Williams
–
Reidhead
2019
2018 5 TP, DIN,
nitrate,
ammonium,
SRP
Wet/dry Bulk No
Williams
– Barrus
et al.
2021
2019 5 TP, DIN,
nitrate,
ammonium,
SRP
Wet/dry Bulk No
Williams
– Barrus
et al.
2021
2020 5 TP, DIN,
nitrate,
ammonium,
SRP
Wet/dry Bulk Yes
W. Miller 2017-2020 9 TP, TN,
orthophosphate
Bulk Bulk No
5.2.2 Results
Following the identification of uncontaminated, contaminated, and unknown samples, the
Williams dataset was evaluated for potential explanations of outlier samples. Figure 17 through
Figure 20 show the count of outlier samples and flux box plots for the contaminated,
uncontaminated, and unknown condition for TP and DIN, respectively. The figures show that the
majority of outliers were contaminated, and the majority of non-outliers were not contaminated,
neglecting unknown samples where metadata were absent (Figure 17, Figure 18). When
comparing contaminated and uncontaminated samples, TP flux was significantly higher in
contaminated samples (ANOVA, p < 0.0001, df = 275; Figure 18). Among uncontaminated TP
samples, there was no significant difference among sites (ANOVA, p = 0.33, df = 169). DIN flux
was significantly higher in contaminated samples than uncontaminated samples (ANOVA, p <
0.0001, df = 275; Figure 20). Among uncontaminated samples, there was no significant
difference among sites (ANOVA, p = 0.25, df = 169).

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
35

Figure 17. Counts of contaminated, uncontaminated, and unknown TP samples in the
Williams dataset, grouped by whether or not the sample was an outlier (true or false).

Figure 18. Boxplots of TP flux in the Williams dataset, divided by samples that were
contaminated, uncontaminated, and unknown.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
36

Figure 19. Counts of contaminated, uncontaminated, and unknown DIN samples in the
Williams dataset, grouped by whether or not the sample was an outlier.

Figure 20. Boxplots of DIN flux in the Williams dataset, divided by samples that were
contaminated, uncontaminated, and unknown.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
37
Figure 21 through Figure 24 show the nutrient flux values for all of the uncontaminated samples,
including outliers, that were identified in Figure 18 and Figure 20 above (purple fields). These
sample results were either confirmed through metadata to have no contamination or were from
samples collected after screen installation (May 21, 2020). Most outliers were associated with
contaminated samples, but some occur within the uncontaminated subset of 2020 (open circles in
Figure 21 and Figure 22). These outliers were thought to be associated with local sources and/or
deposition events and were kept in the dataset for Subgroup calculations of atmospheric
deposition.

Figure 21. Time series of uncontaminated TP samples in the Williams dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
38

Figure 22. Time series of uncontaminated DIN samples in the Williams dataset.

Figure 23. Boxplots of uncontaminated TP samples in the Williams dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
39

Figure 24. Boxplots of uncontaminated DIN samples in the Williams dataset.
Decision Point: Metadata
Subgroup members requested metadata on the atmospheric deposition data from the Williams
dataset. Dr. Theron Miller and Dr. Gus Williams searched their records and reached out to the
graduate students who collected data from the samplers. They provided metadata for the 2017
(Olsen et al. 2018) dataset through November 2017 and the 2020 dataset (Barrus et al. 2021). The
metadata for 2018 (Reidhead 2019) and 2019 was not available.
Subgroup members discussed how to incorporate bulk sampler data from the Williams (2017-
2020) dataset if metadata were not available. All Subgroup members supported using these
atmospheric deposition data if one of the following conditions were true:
• The atmospheric deposition data were collected from a sampler with a screen installed
• The atmospheric deposition data were collected from a sampler without a screen installed and
metadata were available to indicate insect or insect parts were not in the sample
The rationale for this approach is that screens on the samplers keep insects out of the sample.
Metadata, when available, gives Subgroup members confidence that insects or insect parts were
not included in the samples. Keeping data that met either of these conditions – versus only screened
samples – also retained multiple years and seasons of data for robust flux analysis. For samples
without metadata or sampler screens, there remained uncertainty on the extent that insect or insect
parts affected the magnitude of measured N and P values in the samples.
One Subgroup member suggested that the Subgroup only use screened data for the analysis and
exclude any unscreened data with or without metadata. Subgroup members indicated that they had
confidence in the screened data, so an approach that only incorporates screened data would

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
40
simplify the analysis. Subgroup members discussed that one challenge with this approach is that
there is only screened data for half of one year. It would be difficult to extrapolate an annual flux
based on the data, given that seasonal impacts on atmospheric deposition may not be captured
with only a half-year worth of data.
Ultimately, Subgroup members elected to use a dataset that included 1) data from screened
samplers; 2) data from unscreened samplers where the metadata indicates that there were no
insect parts in the sample; and 3) all data that were identified as outliers that met the first two
conditions. All subsequent discussions and analyses use only the data meeting these conditions,
hereafter referred to as the accepted dataset.

5.3 Imputing flux estimates
5.3.1 Methods
When contaminated and unknown samples were removed from the Williams dataset in
development of the accepted dataset, gaps between sampling events occurred in the time series.
These gaps limited the ways that a cumulative annual flux could be calculated. To fill these gaps,
two methods were employed:
1. Impute via linear interpolation. If a missing sampling date is located equidistant between
two other sampling dates that have data, the missing date would be assigned as the midpoint
between the values of the two sampled dates. If the missing date was closer in time to one
of the sampling dates, the imputed value would be proportionally closer to the closer date
than to the farther date. This method assumes that missing data fall within the range of
existing data, and values within a time series are related in time.
2. Impute via relationships with weather. Develop statistical relationships with weather
patterns expected to drive atmospheric deposition, such as precipitation and wind, and use
these relationships to estimate likely flux values for the missing sampling dates under the
weather conditions at that time.
As part of (2), in order to develop the statistical relationships between nutrient flux values and
weather parameters, data were compiled for:
● Average daily precipitation throughout sampling period
● Average & maximum PM2.5 throughout sampling period
● Average & max PM10 throughout sampling period
● Average of daily average wind speed throughout sampling period
● Max of daily average wind speed throughout sampling period
● Average peak daily wind gust throughout sampling period
● Max of peak daily wind gust throughout sampling period
● Month (as factor)
The weather stations located nearest the sampling stations were used, per Table 3. Weather
stations included precipitation and wind data. Additional data on PM2.5 and PM10 were
obtained from the Purple Air website at the West Mountain Ranch sampling location

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
41
(https://www2.purpleair.com/). Stepwise model selection for multiple linear regression was run
to determine the best subset of the potential predictor variables.
The methods and results for calculating flux for the W. Miller dataset are discussed in Sections
4.1.2 and 4.2.1, respectively. With these results, the Subgroup evaluated evaporation, sampler
overflow, dust blow off, and sampler cleaning protocols for their potential to influence flux
results. Weather data from proximal weather stations to the W. Miller samplers (Section 4.1.2)
were used to calculate cumulative precipitation and evaporation for each sampling interval.
Calculated precipitation volume was compared to the sample collector volume to assess potential
for sampler overflow during precipitation events. Evaporation volume was calculated for each
sampling interval to assess potential influence on sample volume and concentration.

5.3.2 Results: Williams Dataset
Weather tended to be fairly episodic and was often inconsistent among sites (Figure 25). Though
episodic events were common across the time series, the period from July through August had
fewer higher precipitation and wind events than other months in 2020. High deposition events
were often preceded by high precipitation and/or wind events.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
42

Figure 25. Flux, precipitation, and average daily wind speed at the weather stations
associated with the atmospheric deposition sampling sites in the Williams dataset. TP fluxes
for the same time period are displayed for additional context.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
43
For TP fluxes, average peak daily wind gust and maximum of peak daily wind gust were the best
subset of weather regression predictors (Table 5). For DIN fluxes, the average daily
precipitation, average PM2.5 and PM10, average peak daily wind gust, and maximum of peak
daily wind gust were the best subset of predictors. These statistical relationships supported the
hypothesis that weather events such as high wind and precipitation drive high atmospheric
deposition events (Table 5). Note that month of year was not a significant predictor for TP or
DIN. The linear regressions did not fully explain the variability in flux, suggesting that (a) the
sampling periods (1+ weeks) could not resolve episodes with shorter timespans, and/or (b)
weather data alone do not illuminate the full context of the drivers of atmospheric deposition,
and other factors such as local sources and wind direction and temporal pattern may also
contribute.
Table 5. Multiple regression results to predict atmospheric deposition fluxes from weather
conditions.
Response
Variable

df

R2
Coefficient: Avg.
peak daily wind
gust
Coeffici
ent:
Max.
wind
gust
Coefficient
: Avg. daily
precip.
Coefficient
: Avg.
PM2.5
Coefficient:
Avg. PM10

TP

106

0.12

0.176

-0.045
No statistical
significance
No statistical
significance
No
statistical
significance

DIN

93

0.40

0.188

-0.088

0.296

0.657

-0.604
When gaps in sampling dates were filled via the weather regression, the cumulative nutrient flux
values were calculated for four shoreline sampler locations. These flux values are summarized in
Table 6.
Cumulative flux was lower than that estimated from linear interpolation for the Mosida and
Lakeshore sites and was equivalent for the Pump Station and Orem sites (Table 6, Figure 26,
Figure 27). The explanation for the cumulative flux being lower for the weather regression
method than for linear interpolation was that the sampling periods associated with gaps tended to
be “calm” periods of weather (i.e., fairly low wind and precipitation) that resulted in lower flux
estimates than those generated from linear interpolation between sampling periods with
relatively high observed fluxes. Linear interpolation that incorporates event driven fluxes would
result in overestimating gaps that occurred during calm non-event periods.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
44
Table 6. Cumulative annual fluxes for 2020 from the Williams dataset, with missing
sampling dates imputed via weather regression relationship.
Dataset Site TP Cumulative Flux
(mg/m2/y)
DIN Cumulative Flux
(mg/m2/y)

Williams
Lakeshore 150.0 740.8
Mosida 444.5 1,624.5
Orem 203.5 735.5
Pump Station 235.5 764.2

\

Figure 26. Cumulative TP flux for the Williams dataset, with gaps in sampling dates
imputed by linear interpolation (solid lines) and weather regression (dotted lines). Note that
the two imputation approaches were equivalent for Orem and Pump Station, so the solid and
dotted lines overlap.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
45

Figure 27. Cumulative DIN flux for the Williams uncontaminated dataset, with gaps in
sampling dates imputed by linear interpolation (solid lines) and weather regression (dotted
lines). Note that the two imputation approaches were equivalent for Orem and Pump Station, so
the solid and dotted lines overlap.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
46
Decision Point: Imputing flux values for sampling dates removed due to contamination
To help evaluate which method for imputing flux values is most appropriate for the dataset, all
Subgroup members requested that Tetra Tech calculate the annual cumulative flux using a linear
interpolation approach and the weather regression analysis to impute values for missing data.
After calculating the annual cumulative load using both approaches, Subgroup members discussed
which approach is better suited to interpolate data between sampling events. As part of the
discussion, they identified the benefits and drawbacks of linear interpolation. One of the benefits
of linear interpolation is that it is a simple method for imputing data. One of the drawbacks of
linear interpolation is that it assumes consistent and predictable patterns between sampling events,
so it is not an effective method to capture patterns for episodic time series (i.e., events occurring
at irregular intervals). Linear interpolation is also not an effective method for addressing episodic
events that are short in duration with wide gaps between episodes.
Subgroup members also discussed the approach for imputing data via weather regression analysis.
With support from the Subgroup, Tetra Tech conducted a weather regression analysis using
weather parameters (e.g., precipitation, wind speed, PM2.5, and PM10) and uncontaminated
atmospheric deposition values from the Williams dataset. Subgroup members also examined the
relationship between precipitation and average wind speed and specific outliers at Mosida to
assess if weather events could explain high nutrient fluxes.
After reviewing the results of all the analyses conducted by Tetra Tech, all Subgroup members
agreed to use the results of the weather regression analysis to impute missing values within the
dataset. Weather regression was preferred over the linear interpolation approach, which assumes
a consistent and predictable pattern of deposition between sampling events. Since the atmospheric
deposition time series is episodic, linear interpolation is not a good method for imputing values.
Since the weather regression analysis showed a relationship between weather variables and
atmospheric deposition values, all Subgroup members supported using the results of the weather
regression to estimate missing values between sampling events present in the accepted dataset.
The resulting dataset is applied to all subsequent analyses and is hereafter referred to as the final
Williams dataset.
5.3.3 Results: W. Miller Dataset
The W. Miller bulk deposition sampler employed a precipitation collector to intercept
precipitation and funnel it into a collection vessel. The rain volume captured was not measured.
Nutrient concentrations were measured in the rain water collected and these concentrations were
converted to flux using precipitation data from a single weather station in Lehi. The Subgroup
discussed potential limitations of this sample design that could influence the calculated flux value
including: 1) the potential for evaporation to occur if the sample was not collected immediately
following a precipitation event; 2) the potential for overflow of the sample collector when
precipitation volume exceeded the collector volume; 3) The potential for dry deposition occurring
between precipitation events to be blown off of the precipitation collection plate; and 4) the
influence of sampler cleaning protocols. The W. Miller dataset was not evaluated for outliers and
contamination.
The limitations of the W. Miller study are outlined below:

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
47
Evaporation: The intent for the sampling design was to immediately sample following
precipitation events, but this was not always accomplished. If precipitation occurred and was
followed by evaporation out of the sampler before the sample was collected, the nutrient flux
would be overestimated because the sample would be more concentrated because of the
evaporative loss. Out of 434 samples, 48 had no precipitation except on the sampling day.
386 samples (89%) were not collected immediately following a precipitation event with
sample volumes potentially reduced by evaporation resulting in overestimated flux.
The potential for evaporation was explored for the situations when precipitation occurred a
day or more prior to sampling. The depth of precipitation in each sampler was calculated
daily, and daily evaporation rates as measured at the BYU weather station (the only station
from the originally identified weather stations with evaporation data) were applied if the
precipitation depth was nonzero. If the cumulative precipitation exceeded the depth of the
sampler, the precipitation depth was set to the depth of the sampler. On the date of each
sampling event, the precipitation depth was reset to zero. Though evaporation sometimes
drew down the depth of water in the sampler, sampling events usually occurred close enough
to precipitation events that substantial evaporation impacts were relatively rare (Figure 28).

Figure 28. Counts of the ratio of net precipitation (cumulative precipitation minus
evaporation) to cumulative precipitation for sampling events in the W. Miller dataset.
Overflow: Given the relative areas of the sampler collector (20 in diameter) and the sampler
container (4 in diameter), it was possible that precipitation events of a certain intensity would
cause the sampler to overflow. Depending on how evenly atmospheric deposition is
distributed across a precipitation event and how homogenized the sample would be inside the
container, the potential impact of sampler overflow on the flux could be overestimation or
underestimation. Given the dimensions of the sampler, a precipitation event of >0.48 inches
would exceed the sampler volume. This magnitude of precipitation was observed in 0.1-3.2%

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
48
of single days across weather stations and in 48.9-76.7% of sampling events (from
cumulative precipitation during the sampling interval) across sampling sites. The latter
estimate assumes no evaporation, which would reduce the volume over time.
Loss of Dry Deposition: The sampler was a shallow black pan that funnels into a collection
tube. If dry deposition fell onto the collector without falling into the collection tube and was
subsequently blown off, the flux would be underestimated. However, the collection tubes
contained wet and dry deposition because of the possibility of dry deposition falling into or
being washed into the collector.
Sampler Cleaning Between Events: In W. Miller’s response to a review by David Gay, he
reported that the samplers at BYU, Spanish Fork, and Lehi were cleaned “quite well”
between sampling events, whereas the other samplers that were collected by National
Weather Service observers were cleaned “now and then.” If samplers were not cleaned
between each sampling event using a method that would remove nutrients (e.g., acid
washing), it is possible that flux estimates would be overestimated due to nutrient residue on
the sampler. The consistency and effectiveness of sampler cleaning could not be established.
Due to absent metadata and insufficient information to constrain these sources of error, it was
decided that the W. Miller dataset would be used for comparisons with the Williams dataset but
would not be primary data used toward developing the comprehensive load estimate for the lake.
Decision Point: Interpreting the W. Miller dataset
Subgroup members discussed potential sources of error in the results of the W. Miller dataset.
They primarily focused on four ideas: a) evaporation, b) overflow, c) loss of dry deposition, and
d) sampler cleaning between events. Subgroup members had different perspectives on the potential
magnitude and impact of each source of error on the W. Miller study results. They recommended
several analyses to evaluate the impact of each potential source of error on the W. Miller dataset.
The analyses provided useful insight into how potential sources of errors impacted the results, but
Subgroup members concluded that it did not provide conclusive evidence on the exact degree of
impact. Subgroup members concluded that the W. Miller data could be useful for comparison with
the Williams dataset to help resolve outlier or otherwise anomalous atmospheric deposition
values. Given the uncertainties in the W. Miller data, all Subgroup members recommended using
the Williams uncontaminated dataset as primary data in calculating the cumulative annual flux
and loading to Utah Lake.

5.4 Comparing samples between studies
TP fluxes in the final W. Miller dataset were significantly lower than in the final Williams
uncontaminated dataset (ANOVA; p < 0.01, F = 19.6, df = 428) (Figure 29, Figure 30).
Similarly, TN fluxes in the W. Miller dataset were significantly lower than the DIN fluxes in the
Williams dataset (ANOVA; p < 0.01, F = 18.56, df = 424) (Figure 31, Figure 32). Note that
while W. Miller (2021) reported that his N fluxes were measured as TN, which was verbally
confirmed by Dr. Williams during the Science Panel Subgroup meetings, the TN measurements
were computed as the sum of nitrate and ammonium fluxes, thus representing DIN rather than

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
49
TN. Thus, the nitrogen measurements used in the two studies are directly comparable. Several
sampling sites were common between the two datasets, enabling direct comparison.

Figure 29. Log scale boxplots of TP flux from the W. Miller and Williams datasets. The
star on each plot represents the mean of the data and the median is displayed as the solid
horizontal line in the box.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
50

Figure 30. Log scale boxplots of TP flux across sites from the W. Miller and Williams
datasets. The star on each plot represents the mean of the data and the median is displayed
as the solid horizontal line in the box. Stations are organized in clockwise order across the
lake.

Figure 31. Log scale boxplots of DIN and TN flux from the W. Miller and Williams datasets.
The star on each plot represents the mean of the data and the median is displayed as the solid
horizontal line in the box.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
51

Figure 32. Log scale boxplots of DIN and TN flux across sites from the W. Miller and
Williams datasets. The star on each plot represents the mean of the data and the median is
displayed as the solid horizontal line in the box. Stations are organized in clockwise order
across the lake.

Lower daily flux values in the final W. Miller dataset resulted in lower calculated cumulative fluxes as well (

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
52
Table 7). Cumulative annual fluxes were calculated for the W. Miller dataset from 2017-2020 and

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
53
Table 7 shows the range for these annual fluxes. Only the 2020 dataset from Williams are shown in

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
54
Table 7 so there is no range of values.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
55
Table 7. Cumulative fluxes of nutrients across sites for the W. Miller and Williams
datasets.

Dataset

Site
TP Cumulative Flux
(mg/m2/y)
(min and max)
DIN Cumulative Flux
(mg/m2/y)
(min and max)
Williams Orem 203.5 735.5
W. Miller Orem 37.8 101.9 223.1 571.6
W. Miller BYU 14.4 48.8 479.5 966.2
W. Miller Spanish Fork 40.1 80.2 467.0 768.7
Williams Lakeshore 150.0 740.8
W. Miller Lincoln 246.5 415.9 571.7 1,852.2
W. Miller Genola 86.1 504.5 274.0 550.7
W. Miller Elberta 55.6 129.4 319.6 630.8
Williams Mosida 444.5 1,624.5
W. Miller Mosida 105.8 318.7 495.5 4,385.3
W. Miller Saratoga Springs 41.0 170.7 365.0 628.3
Williams Pump Station 235.5 764.2
W. Miller Pump Station 80.6 168.6 407.2 650.5

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
56
6 Evaluate spatial interpolation among sites and attenuation of fluxes
Two types of atmospheric deposition sources were considered in this analysis: 1) regional
sources of dust and nutrients originating from the Sevier Dry Lake, Great Lake Salt Lake, and
the West Desert playas; and 2) local sources of dust and nutrients originating from areas adjacent
to and in proximity of Utah Lake. Previous studies in other systems demonstrated attenuation of
atmospheric deposition fluxes moving away from the source, particularly for locally-derived
sources (Wilson and Serre 2007, VanCuren et al. 2012a, 2012b). It follows that for Utah Lake,
atmospheric deposition from local sources is expected to decrease moving over the lake and
away from shore. Regional fluxes are expected to not attenuate and be more equally distributed
across the lake.
Previous investigations of Utah Lake atmospheric deposition all included some level of nutrient
flux attenuation in their estimates. Brahney (2019) used a first-order decay rate to describe
attenuation of local dry fluxes moving away from the shoreline, considering 200, 400, and 600
meter intervals of termination. In addition to this, Brahney (2019) calculated a uniform dry and
wet regional flux over the entire lake. Olsen et al. (2018) assumed “background” regional fluxes
at five interior points in the lake and used kriging to spatially interpolate between the in-lake
points and the shoreline sampling sites. Reidhead (2019) used a linear fall-off of shoreline fluxes
to a point of zero deposition at the center of the lake.

6.1 Bird Island Sampler Results
To attempt to quantify fluxes in the center of the lake and assess the magnitude of attenuation,
Barrus et al. (2021) installed a wet/dry atmospheric deposition sampler at Bird Island in mid-
2020. Results were reported as bulk deposition and not as individual wet/dry components.
Measuring atmospheric deposition at a mid-lake location allowed for the testing of two
hypotheses:
Hypothesis 1: Attenuation of atmospheric deposition fluxes occurs as distance increases from
land-based local sources. If fluxes are lower at Bird Island than at shoreline sites, this
hypothesis would be supported.
Hypothesis 2: Attenuation of atmospheric deposition does not occur, and fluxes at the center
of the lake are similar to land-based fluxes. If fluxes are equal in magnitude at Bird Island and
shoreline sites and temporal patterns are consistent, this hypothesis would be supported.
The daily and cumulative fluxes at Bird Island were higher than at other sites (Figure 33 through
Figure 36).

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
57

Figure 33. Time series of TP fluxes in the Williams dataset, with Bird Island fluxes
highlighted (purple) compared to other sites (gray).

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
58

Figure 34. Time series of DIN fluxes in the Williams dataset, with Bird Island fluxes
highlighted (purple) compared to other sites (gray).

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document
59

Figure 35. Cumulative TP fluxes for the Williams dataset, starting on the date when the
Bird Island sampler was installed.
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

60

Figure 36. Cumulative DIN fluxes for the Williams dataset, starting on the date when the
Bird Island sampler was installed.
This observation was not consistent with either hypothesis 1 or 2, thus pointing to other
potential hypotheses:
Hypothesis 3: Attenuation of atmospheric deposition does not occur, and there is a land-based
local source of higher atmospheric deposition nearshore of Bird Island that is not captured by
the current array of samplers.
Hypothesis 4: The fluxes observed at Bird Island represent a lake-based or island-based local
source of nutrient flux. Possibilities for a lake-based or island-based source could include
contamination from bird droppings, volatilized material from the island, and spray from lake
water.
T. Miller (2022) described support for hypothesis 3, stating

The windrose... shows that Bird Island would be most influenced by shoreline rates from the
northwest shore of Utah Lake and the area north of the Mosida sampling site. Neither of these
areas have a shoreline sampler. The northwest shore area does not have much agriculture but
is experiencing urban expansion in the cities of Lehi and Eagle Mountain. We are exploring
the possibility of placing a sampler in this area for future collections.
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

61

The wind roses around Utah Lake could provide information about the potential prevailing wind
patterns moving over the lake (Figure 37). However, wind direction alone cannot fully support
the hypothesis of an unsampled high shoreline flux that would explain the magnitudes observed
at Bird Island. David Gay, in his review of the (T. Miller, 2022) report, stated
One way you might be able to show that this is a real signal goes something like this. The
Lakeshore sampling site is not capturing the urban “plume” moving over the lake (plume is
to the north). So put another shore line sampler north of Lakeshore where it would capture
these high samples.
David Gay also provided feedback (for T. Miller, 2022) on the potential for hypothesis 4, either
to demonstrate support or rule it out. His review states
I would expect criticism will come on these observations, such as ‘Can you prove that there is no
contamination going on in the lake that is not representative of the lake surface?’ Condensation
into the bucket because the sampler is colder than the water, for example? Mist/droplets from
waves being added to the sample? Do the wet only samples also show this difference? Is the
difference in the dry side? Bird poop in the dry side? Are the birds using it as a resting place
(although then you get into the argument of bird feces as a source)? I would again recommend
beefing up the QA [quality assurance] information for the Bird Island sampler. Prove to the
reader that you have QA info that shows these samples are valid.

Figure 37. Wind rose data for seven weather stations located around Utah Lake.
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

62

Decision Point: Bird Island Data
Subgroup members discussed whether, and if so, how data collected from samplers on Bird Island
could inform the atmospheric deposition loading analysis. The Subgroup members did not reach
a consensus on their recommendation on whether to incorporate Bird Island data into the
atmospheric deposition analysis.
The majority of the Subgroup members supported retaining the Bird Island data as a
representation of local lake-based or island-based source nutrient input but not using it to estimate
external atmospheric deposition influx into Utah Lake.
• The Subgroup members who supported this decision stated that they were concerned that birds
could have deposited droppings into the samplers due to the number of birds visiting the Island.
These droppings would increase the N and P values in the sampler.
• They acknowledged that bird droppings are a nutrient source to Utah Lake but that the fluxes
calculated at Bird Island may not be representative of atmospheric deposition inputs across
all of Utah Lake since the sampler is stationed at a bird rookery. Furthermore, since birds may
be eating organisms from Utah Lake, their droppings may not necessarily represent a net influx
of nutrients to Utah Lake.
• Subgroup members expressed concerns about other potential influences on the samplers,
including the aerosolization of bird materials from the island and spray from the lake water.
• Additionally, Subgroup members considered the potential for an unidentified land-based
source reaching Bird Island, but ultimately concluded that additional studies as recommended
by Dr. David Gay’s would be needed to evaluate hypothesis 3.
One Subgroup member did not support this decision. They stated that the Bird Island samplers
should be used to estimate the annual atmospheric deposition nutrient load to Utah Lake and
that the data from the Bird Island sampler indicates that atmospheric deposition is not
attenuating across Utah lake. They also stated that the samplers did not show evidence that bird
droppings got into the samples. They shared that perching birds do not travel to Bird Island due
to its distance from the shore, and the webbed-footed birds that travel to the Island would be
unable to perch on the sampler. Additionally, they collected samples of bird droppings around
Utah Lake. They measured that the nutrient content of those droppings was five to ten times
higher than the nutrient concentration values found in the Bird Island sampler, suggesting that
bird droppings did not influence the data collected at Bird Island.
Other Subgroup members stated that this evidence is inconclusive in determining whether bird
droppings influence Bird Island sampler data, as droppings could have partially been deposited
into a sampler.
As an alternative explanation for why atmospheric deposition flux values were higher at Bird
Island than at the shoreline samplers, the Subgroup member in the minority hypothesized that
the southeastern winds and eastern winds from the canyons could transport and deposit dust and
aerosols to the Bird Island sampler in the evening and early morning. Southwestern prevailing
winds could transport and deposit dust and aerosol particles to the Bird Island sampler in the
afternoon. The southeastern and southwestern winds would converge over Utah Lake and settle
dust and aerosols near the Bird Island sampler, which is why the values from the Bird Island
sampler are higher than the shoreline samplers. Additionally, they stated that there is an
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

63

inversion nearly every day over Utah Lake, which results in the deposition of aerosols from
urban zones into Utah Lake, including the area near the Bird Island sampler.
6.2 Attenuation
6.2.1 Methods
Instead of using Bird Island to characterize the degree of attenuation of atmospheric deposition,
observations from the literature were explored. Jassby et al. (1994) measured nutrient fluxes in
atmospheric deposition across Lake Tahoe, noting that dry deposition of DIN increased moving
toward mid-lake (potentially due to canopy uptake of DIN in shoreline forested sites) and dry
deposition of SRP decreased moving toward mid-lake. Jassby et al. (1994) also found that wet
deposition, for both DIN and SRP, decreased moving toward mid-lake, likely due to precipitation
patterns in the basin. Wet deposition made up the majority of nutrient deposition, suggesting that
in total, atmospheric deposition decreased moving toward mid-lake. The mechanistic drivers of
atmospheric deposition in the forested, snowpack-dominated Lake Tahoe basin may be different
than the drivers in Utah Lake.
In a later study in Lake Tahoe, VanCuren et al. (2012a and 2012b) observed aerosol size and
concentrations. The findings from these two related studies highlighted that regional sources of
dry deposition tended to be fairly steady across the lake, whereas local sources such as urban
areas tended to be highly localized, with fluxes dropping off moving away from shore. While all
particles followed an exponential decay rate moving away from local sources, larger particles
tended to attenuate more rapidly than smaller particles.
The attenuation of local sources of atmospheric deposition was also illustrated in a study
conducted in terrestrial systems located near concentrated animal feeding operations (CAFOs) in
North Carolina (Wilson and Serre 2007). This study focused specifically on ammonium and
noted that ammonium concentrations decreased with distance from CAFOs. The steepest
decrease in ammonium concentrations was between the 0.0-0.5 km to 0.5-1.0 km distance bins.
A dominance of regional sources was noted beyond distances of 2 km.
Goodman et al. (2019) sampled bulk dust grain sizes in the region surrounding Utah Lake. Grain
sizes were similar between fine playa, snow, and urban dust. The most common grain sizes were
10 µm for regional playa dust and 20 µm for urban and snow dust. Fine playa dust is distributed
more widely than might be anticipated given its grain size and widespread occurrence across
samplers in the region (Goodman et al. 2019).
The grain size of dust particles was considered to assess potential attenuation rates for local
sources of dust. If grain sizes observed by Goodman et al. (2019) are applied to the observations
from VanCuren et al. (2012a) (Figure 38), attenuation would be anticipated to be rapid moving
away from the source, with an exponential decay rate and a range of ~100 m.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

64

Decision Point: Attenuation of Fluxes
Subgroup members discussed how to apply attenuation rates to the calculation of atmospheric
deposition loading to Utah lake. After reviewing and discussing the Jassby et al. (1994), VanCuren
et al. (2012a and 2012b), Goodman et al. (2019), and Wilson and Serre (2007) papers, Subgroup
members agreed that dust and aerosols attenuate as a function of distance. They did not agree on
the attenuation distance or rate for shoreline samplers at Utah Lake.
Subgroup members discussed that factors like wind speed, particle size, and particle shape affect
attenuation. They considered several different methods to identify an attenuation rate.
• Use standard attenuation rates based on NADP models: One suggestion was to use a standard
attenuation rate based on National Atmospheric Deposition Program (NADP) models. One
challenge with this approach is that the NADP models only have a standard attenuation rate
for wet deposition and do not have a standard attenuation rate for dry deposition. Since all of
the samples collected on Utah Lake were bulk samples, the attenuation rates from the NADP
models cannot be applied to the Utah Lake samples.
• Use the Goodman et al. (2019) grain size and the VanCuren (2012a) attenuation rates by grain
size: This methodology would involve cross-referencing the Goodman et al. (2019) grain size
distribution with VanCuren (2012a) attenuation rates by-grain size distribution to establish an
attenuation rate. One advantage of this methodology is it uses grain size information from
areas around Utah Lake. One disadvantage of this methodology is it assumes grain size
distribution equates to N and P fluxes. Cross-referencing Goodman et al. (2019) and
VanCuren (2012a) with this methodology resulted in an attenuation distance of 100 meters.
• The majority of Subgroup members also supported using the attenuation rate identified by
Wilson and Serre (2007) paper as an attenuation maximum. Wilson and Serre (2007) measured
the attenuation rates of local sources over land. The study’s focus on the attenuation rate of
local sources is a particular advantage of this study. One disadvantage of this study is that it
only analyzed ammonia and no other constituents, so applying the study’s attenuation rate to
the Utah Lake atmospheric deposition data would assume that all constituents attenuate at the
same rate as ammonia. This study also focuses on ammonia from hog farms, which is not the
specific local source around Utah Lake. The potential attenuation distance based on the results
of this study is two kilometers.
The majority of Subgroup members supported applying an attenuation rate to the shoreline fluxes
based on the Goodman et al. (2022) grain size and the VanCuren (2012a) attenuation rates by
grain size (potential attenuation range of 100 meters).
One Subgroup member did not support the decision to apply an attenuation rate to the shoreline
fluxes based on the Goodman et al. (2022) grain size and the VanCuren (2012b) attenuation rates
by grain size. Their perspective was that the shoreline fluxes do not attenuate to the center of Utah
Lake.
All Subgroup members agreed to have Tetra Tech calculate multiple atmospheric loading
estimates using different attenuation rates. They planned to use different loading estimates to
select a primary loading value to calibrate the Utah Lake in-lake model and a low and high loading
value to be used in a sensitivity analysis.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

65

Several attenuation scenarios were considered, whereby nutrient concentrations from local
sources were anticipated to decrease moving away from shore. Three scenarios were defined,
representing attenuation distances of 100 m (observed in VanCuren et al. 2012a), 200 m
(VanCuren et al. 2012a plus a buffer distance to account for uncertainty), and 2000 m (observed
in Wilson and Serre 2007). The exponential decay pattern observed in VanCuren et al. (2012a)
for particles 10-25 µm was used (Figure 38), consistent with local deposition in the Utah Lake
basin that had average grain sizes of 20 µm (Goodman et al. 2019).
The bulk deposition samplers used by Williams did not distinguish between regional and local
types of sources, but presumably collected deposits from both types at the Utah Lake shoreline.
Thus, starting with the Williams data set, fluxes could be adjusted with distance across the lake
based on Table 8 with the atmospheric deposition loads attenuating to that of a regional-source
only moving away from the shoreline.

Figure 38. Illustration of attenuation from VanCuren et al. 2012a (relevant particle size:
10-25 µm) and Wilson and Serre 2007. Reproduced from Figure 11 and Figure 3 of the
respective references.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

66

Table 8. Attenuation scenarios based on information in VanCuren et al. 2012a and Wilson
and Serre 2007.
Shoreline flux proportion Regional flux proportion 100 m scenario 200 m scenario 2000 m scenario
1.00 0.00 0 m 0 m 0 m
0.30 0.70 20 m 40 m 400 m
0.045 0.955 50 m 100 m 1000 m
0.026 0.974 100 m 200 m 2000 m
6.2.2 Results
To apply the attenuation scenarios in Table 8, estimating a regional flux was necessary. To
estimate regional fluxes of TP, data from Goodman et al. (2019) and Carling (2022) were used.
Urban dust fluxes in municipalities around Utah Lake had an average of 30.5 g/m2/yr (range:
24.7-34.9 g/m2/yr). Goodman et al. 2019 found that 91% of urban dust was regional in nature,
leading to a regional dust flux estimate of 27.8 g/m2/yr. Putman et al. (2022) also collected bulk
dust deposition, this time near Lehi, and obtained regional dust deposition estimates of 14.6-36.6
g/m2/yr, which was consistent with the estimates derived from Goodman et al. (2019).
Converting regional dust mass to TP, the phosphorus content of regional dust measured by
Carling (2022) was applied. Carling found the P content in regional dust to be between 1,344 and
4,340 mg/kg. Together, the regional source TP was estimated to be 79 mg TP/m2/yr on average
and with a range of 37.4-120.7 mg TP/m2/yr.
Regional TP flux = (Bulk flux * Proportion regional in bulk) * Regional TP content
79 mg TP/m2/yr = 30.5 g/m2/yr * 0.91 * 2,842 mg/kg
The annual TP regional flux estimated above is higher than Brahney (2019) who estimated 5.6
mg TP/m2/yr (range of 0.8-17.9 mg TP/m2/yr). Brahney used a lower estimate of regional dust
deposition that was derived from mountain regions sourced east of the Colorado Plateau.
To estimate regional sources of DIN, values from Brahney (2019) were used, which were
derived from the CMAQ model (which includes data from CASTNET, NADP, AirMoN, and
NADP NTN). The DIN deposition estimate derived from CMAQ was 575 mg DIN/m2/yr (range:
400-750).
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

67

Decision Point: Estimate of regional fluxes
Subgroup members supported calculating a regional flux to be applied across the entire surface
of Utah Lake. This decision means that the shoreline fluxes will attenuate to a baseline regional
flux instead of to zero.
The rationale for this decision was that the evidence from Carling (2022) and Goodman et al.
(2019) suggested playa dust contributes to a regional flux that would not attenuate over Utah
Lake. The Subgroup members supported using the data from Carling (2022) and Goodman et al.
(2019) to identify a regional flux for TP to Utah Lake. The average value of the TP flux
calculated from the Carling (2022) and Goodman et al. (2019) data was used as the regional
bulk sample and applied across Utah Lake. Subgroup members supported using the CMAQ
modeling from Brahney (2019), which included data from CSTNET, NADP, AirMoN, and NADP
NTN, to estimate the regional flux of DIN to Utah Lake.
After calculating the regional flux of DIN, one Subgroup member noted that the estimate (440-
750 mg DIN/m2/year) fits within the N flux range calculated using the Williams dataset.
This Subgroup member stated that he did not understand why the N flux range calculated from
the Williams dataset is acceptable, but the calculated P flux range from the same studies is not.
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

68

7 Determine loading to Utah Lake for including in the ULNM
7.1 Methods
To calculate the loads of TP and DIN to Utah Lake, rates of cumulative annual fluxes from the Williams dataset
(
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

69

Table 7) were combined with attenuation scenarios and regional flux estimates (Table 8). Total
load to 380 km2 Utah Lake was calculated in several steps:
1. Create a raster layer of shoreline fluxes around the edge of Utah Lake. Four shoreline
sampling sites from the Williams dataset were used (Orem, Lakeshore, Mosida, and Pump
Station). Spatial interpolation via inverse distance weighted interpolation generated flux
estimates for all locations around Utah Lake (Figure 39).
2. Assign the decay rate of shoreline fluxes moving from shoreline to offshore in Utah Lake.
See Section 6for details of the three scenarios (Figure 40).
3. Assign the regional flux in areas of Utah Lake that are beyond the shoreline decay
distance. Fluxes of 79.0 mg TP/m2/yr and 575 mg DIN/m2/yr were applied as estimates of
regional flux (Figure 40).
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

70

Figure 39. Inverse distance weighted (IDW) spatial interpolation of shoreline fluxes of TP
(left) and DIN (right) based on observations at the four sampling sites in the Williams
dataset for 2020. Sampling sites (clockwise starting on the east side of the lake) were Orem, Lakeshore, Mosida, and
Pump Station.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

71

Figure 40. Display of the estimate of TP (left) and DIN (right) loading to Utah Lake, which
incorporates shoreline fluxes (local and regional atmospheric deposition sources) at the
edge of the lake that attenuate to a regional flux moving toward the middle of the lake.
The width of each band represents the distances assigned based on the attenuation scenario (the 2000 m scenario is
displayed as an example). Note that areas of higher shoreline flux, namely in the southwest portion of the lake, have
higher nearshore fluxes than areas with lower shoreline flux.
In addition, a fourth scenario was run, which assumed no attenuation but simply applied the
inverse distance-weighted interpolation of shoreline fluxes across the lake (Figure 41). This
scenario represented a maximum possible load (i.e., assuming no attenuation) based on available
information.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

72

Figure 41. Inverse distance weighted (IDW) spatial interpolation of shoreline fluxes of TP
(left) and DIN (right) across the lake, thus representing a “no attenuation” scenario. The flux
values are based on observations at the four sampling sites in the Williams dataset for 2020. Sampling sites (clockwise starting on
the east side of the lake) were Orem, Lakeshore, Mosida, and Pump Station.
7.2 Results
The attenuation scenarios resulted in total load estimates to Utah Lake of 31-45 metric tons
TP/yr and 218-249 metric tons DIN/yr (Table 9). The scenario assuming no attenuation resulted
in total load estimates of 93 metric tons TP/yr and 351 metric tons DIN/yr. These scenarios fell
in the range of other constraining analyses and were lower than previous published estimates that
included results from samples that were identified as contaminated in this report.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

73

Table 9. Atmospheric deposition loading estimates for Utah Lake as a result of this analysis
(rows 1-4 in blue) compared to constraining analyses (rows 5-9 in white) and other
published studies (rows 10-12 in blue).
Scenario TP
(metric tons/yr)
DIN
(metric tons/yr)
Notes
100 m attenuation 31 218
200 m attenuation 32 220
2000 m attenuation 45 249
No attenuation 93 351
Carling 2022 57.5 Dust conversion, assumes no
attenuation.

Brahney 2019

2-21

153-288
Assumes attenuation of local
sources and no attenuation of
regional sources at 200, 400, and
600 m distance.
Brahney 33 Mass balance
Brett 60 Mass balance

W. Miller 2021

50-104

257-409
Assumes no attenuation
Loads reported by W. Miller, not
calculated in this report.

Olsen et al. 2018

12-525

61-613
Low loads include uncontaminated
samples only, and high loads
include contaminated samples. Included attenuation by assigning background fluxes at the lake interior and interpolating via kriging.

Reidhead 2019

193

637
Unscreened samplers, could
include contamination. Included
attenuation by assigning a linear
fall-off of deposition to a point of
zero at the lake center.

Barrus et al. 2021

133-262

482-1,052
Low loads represent partially screened 2020 samples, and high loads represent unscreened 2019 samples with some contamination. Assumes no attenuation by incorporating fluxes at Bird Island.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

74

Decision Point: Loading recommendations
Subgroup members evaluated atmospheric loading estimates for TP and DIN based on four
attenuation scenarios:
• Shoreline fluxes attenuate at 100 meters off the shoreline (based on the VanCuren (2012a)
attenuation rate by grain size and Goodman et al. (2019) grain size data)
• Shoreline fluxes attenuate at 200 meters off the shoreline (based on VanCuren (2012a) 100
meters attenuation range by grain size and Goodman et al. (2019) grain size data plus a 100
meters buffer to account for uncertainty)
• Shoreline fluxes attenuate at 2,000 meters off the shoreline (based on Wilson and Serre (2007)
• Shoreline fluxes do not attenuate across Utah Lake
All attenuation scenarios included a regional flux of 79.0 TP mg/m2/yr based on Carling (2022)
and Goodman et al. (2019) and 575 DIN mg/m2/yr based on CMAQ modeling from Brahney
(2019), meaning the shoreline fluxes attenuated to the regional flux value rather than zero.
The Utah Lake modeling team requested that the Subgroup provide a primary recommendation to
calibrate the model and a low and high recommendation for a sensitivity analysis once the model
is calibrated.
The majority of Subgroup members supported using the atmospheric deposition loading values
based on the 200-meter attenuation rate as the primary recommendation for model calibration
(TP: 32 metric tons/year; DIN: 220 metric tons/year). They also supported using the atmospheric
deposition loading values based on the 100-meter attenuation rate as the low recommendation
(TP: 31 metric tons/year; DIN: 218 metric tons/year) and the 2,000-meter attenuation rate as the
high recommendation (TP: 45 metric tons/year; DIN: 249 metric tons/year) for the sensitivity
analysis. They stated that the VanCuren (2012a) attenuation rates by grain size and the Goodman
et al. (2019) grain size data suggest that local sources attenuate across Utah Lake.
One Subgroup member did not support the majority Subgroup recommendation for atmospheric
deposition loading values. They recommended that the primary TP loading recommendation for
model calibration is 150 metric tons/year, the low TP loading recommendation is 93 metric
tons/year (the Subgroup analysis value that assumes no attenuation), and the high TP loading
recommendation is 200 metric tons/year. These values are based on calculations of the Williams
dataset in its entirety (i.e., no contaminated samples removed). Additionally, they cited the Bird
Island data as evidence that suggests no attenuation of local sources occurs across Utah Lake.
Accordingly, their low recommendation is based on this conclusion from the Bird Island dataset.
The detailed reasoning for their TP and DIN loading value recommendations is in Appendix C.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

75

8 Evaluate chemical speciation
The proportions of N constituents compared to DIN, on average across sites, were 30.25% for
nitrate and 69.75% for ammonium (Table 10). The proportion of SRP compared to TP was
37.5%, on average across sites. Proportions of N constituents tended to be fairly consistent
across sites except Mosida, suggesting a possible local influence that has a higher proportional
abundance of ammonium relative to nitrate. Proportions of SRP relative to TP ranged from 24-
48% across sites and was more consistent with regional dust proportions than urban dust
proportions (Brahney 2019).
Table 10. Proportions of chemical constituents in DIN and TP across sites and compared to
other studies.
Study Site Nitrate/DIN Ammonium/DIN SRP/TP

Williams data
2020
Orem 0.35 0.65 0.46
Lakeshore 0.37 0.63 0.48
Mosida 0.10 0.90 0.24
Pump Station 0.39 0.61 0.27

Brahney 2019 Urban dust 0.75
Regional dust 0.34
Reidhead 2019 Utah Lake
shoreline sites
0.37
W. Miller 2021 Utah Lake shoreline sites 0.32
The Utah Lake Nutrient Model specifies specific constituents needed as inputs for atmospheric
deposition. These constituents are organic N, nitrate, and ammonium for N, and orthophosphate
and organic P for P. Apportioning the DIN load as 30.25% nitrate and 69.75% ammonium, with
an unknown amount apportioned to organic N was determined as the recommendation for N.
Apportioning the P load was a more complicated recommendation to derive, as much of the TP
load to Utah Lake could be expected to be sediment-bound P that is neither organic nor
orthophosphate. The recommendation was to assign 37.5% of the TP load as orthophosphate and
to allow the modeling team to determine the best course of action to characterize the reactivity
and chemical behavior of the remainder of the TP load. The temporal and spatial distribution of
the load was also recommended to be determined by the modeling team.
Decision Point: Chemical speciation
The Utah Lake modeling team requested that the Subgroup provide organic N, nitrate, and
ammonium as the N constituents. All Subgroup members supported apportioning the N load as an
unknown proportion organic, 30.25% of DIN as nitrate, and 69.75% DIN as ammonium based on
the proportions of chemical constituents measured in the Williams dataset.
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

76

Additionally, the Utah Lake modeling team requested that the Subgroup provide orthophosphate
and organic P as the P constituents to input into the model. One challenge with apportioning P in
the model is that the datasets available to the Subgroup contained information on orthophosphate
but did not contain information on organic P. Subgroup members discussed how to apportion P,
given the limitations in the data.
One suggestion from a Subgroup member was to determine the fraction of TP that is calcium-
bound. Assuming that calcium-bound P is not bioavailable, the modeling team could subtract the
fraction of calcium-bound P from TP, generating a value of bioavailable phosphorus for primary
production. There is data on the amount of calcium-bound P in the Utah Lake sediment, but this
evidence is not representative of the amount of calcium-bound P coming from atmospheric
deposition since the calcium-bound P in the sediment comes from multiple sources (e.g., rivers,
in-lake precipitation, etc.). Additionally, Subgroup members noted that magnesium and iron-
bound P could contribute bioavailable P to Utah Lake if these bounded particles encounter anoxic
conditions at the Utah Lake sediment-water interface. Subgroup members noted studies that may
contain information on P speciation of the fine playa dust (i.e., the regional dust) and be useful in
apportioning total P, including Brahney (2019), Dr. Josh LeMonte's forthcoming study on P-
binding, and Goodman et al. (2019). Given the limitations in the available data, Subgroup
members supported having the Tetra Tech modeling team develop an approach to characterize P
speciation in the model for review.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

77

9 References
Barrus SM, Williams GP, Miller AW, Borup MB, Merritt LB, Richards DC, and Miller T. 2021.
Nutrient atmospheric deposition on Utah Lake: A comparison of sampling and analytical
methods. Hydrology 8: 123. DOI: 10.3390/hydrology8030123
Brahney J. 2019. Estimating total and bioavailable nutrient loading to Utah Lake from the
atmosphere. Prepared for the Utah Lake Science Panel and the Utah Division of Water Quality.
Carling G. 2022. Atmospheric deposition of total phosphorus to Utah Lake.
Gay DA. 2019a. Wasatch Front Water Quality Council external review – Task 1 review of Utah
Lake atmospheric deposition documents. https://documents.deq.utah.gov/water-
quality/watershed-protection/utah-lake/DWQ-2020-012538.pdf
Gay DA. 2019b. Wasatch Front Water Quality Council external review – Task 2 review and
recommendations. https://documents.deq.utah.gov/water-quality/watershed-protection/utah-
lake/DWQ-2020-012558.pdf
Goodman MM, Carling GT, Fernandez DP, Rey KA, Hale CA, Bickmore BR, Nelson ST, and
Munroe JS. 2019. Trace element chemistry of atmospheric deposition along the Wasatch Front
(Utah, USA) reflects regional playa dust and local urban aerosols. Chemical Geology 530:
119317. DOI: 10.1016/j.chemgeo.2019.119317
Jassby AD, Reuter JE, Axler RP, Goldman CR, and Hackley SH. 1994. Atmospheric deposition
of nitrogen and phosphorus in the annual nutrient load of Lake Tahoe (California-Nevada).
Water Resources Research 30(7): 2207-2216.
Miller TG. 2022. Review, chronology and summary of the atmospheric deposition program
sponsored by the Wasatch Front Water Quality Council.
Miller AW. 2021. Atmospheric bulk deposition of nutrients. Progress Report and Reviews.
Olsen JM, Williams GP, Miller AW, and Merritt LB. 2018. Measuring and calculating current
atmospheric phosphorus and nitrogen loadings to Utah Lake using field samples and
geostatistical analysis. Hydrology 5: 45. DOI: 10.3390/hydrology5030045
Putman AL, Jones DK, Blakowski MA, DiViesti D, Hynek SA, Fernandez DP, and Mendoza D.
2022. Industrial particulate pollution and historical land use contribute metals of concern to dust
deposited in neighborhoods along the Wasatch Front, UT, USA. GeoHealth 6: e2022GH000671.
DOI: 10.1029/2022GH000671
Reidhead JG. 2019. Significance of the rates of atmospheric deposition around Utah Lake and
phosphorus-fractionation of local soils. Masters Thesis submitted to the faculty of Brigham
Young University.
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

78

Richards DC. 2022. Nutrient atmospheric deposition on Utah Lake and Great Salt Lake locations
2020, including effects of sampler type: Statistical analyses and results. Prepared for Wasatch
Front Water Quality Council.
Utah Lake Science Panel. 2019. Utah Lake Science Panel comments regarding Wasatch Front
Water Quality Council’s atmospheric deposition study. https://documents.deq.utah.gov/water-
quality/watershed-protection/utah-lake/DWQ-2020-012560.pdf. August 2, 2019.
Utah Lake Science Panel. 2020a. Science Panel Recommendations for the Wasatch Front Water
Quality Council Atmospheric Deposition Sampling Program.
https://documents.deq.utah.gov/water-quality/watershed-protection/utah-lake/DWQ-2020-
012554.pdf
Utah Lake Science Panel. 2020a. Science Panel Recommendations for the Wasatch Front Water
Quality Council Atmospheric Deposition Sampling Program.
https://documents.deq.utah.gov/water-quality/watershed-protection/utah-lake/DWQ-2020-
012554.pdf
Utah Lake Science Panel. 2020b. Science Panel Update to Steering Committee Regarding
Engaging with All Potential Sources of Information to Address Initial Charge Questions.
Background on Atmospheric Deposition Engagements. https://documents.deq.utah.gov/water-
quality/locations/utah-lake/DWQ-2021-007224.pdf
VanCuren R, Pederson J, Lashgari A, Dolislager L, and McCauley E. 2012a. Air pollution in the
shore zone of a Large Alpine Lake – 1 – Road dust and urban aerosols at Lake Tahoe,
California–Nevada. Atmospheric Environment 46: 607-617. DOI:
10.1016/j.atmosenv.2009.12.001
VanCuren R, Pederson J, Lashgari A, Dolislager L, and McCauley E. 2012b. Aerosol generation
and circulation in the shore zone of a Large Alpine lake – 2 – Aerosol distributions over Lake
Tahoe, CA. Atmospheric Environment 46: 631-644. DOI: 10.1016/j.atmosenv.2009.08.049.
Wilson SM and Serre ML. 2007. Examination of atmospheric ammonia levels near hog CAFOs,
homes, and schools in Eastern North Carolina. Atmospheric Environment 41(23): 4977-4987.
DOI: 10.1016/j.atmosenv.2006.12.055Appendix A: AD Subgroup Analysis Plan

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

79

Appendix A: AD Subgroup Analysis Plan
Objectives
1. Analyze available information and data to improve understanding of atmospheric
deposition to Utah Lake
2. Work collaboratively toward a recommendation for atmospheric loading, ideally
achieved through consensus
3. Document the SP’s decision-making process for analyzing and evaluating evidence and
working toward an atmospheric deposition recommendation

Tasks
1. Review and summarize raw data from G. Williams (Olsen 2018, Reidhead 2019, and
Barrus 2021) and W. Miller datasets (SP with Tt support)
Purpose and goals
 Process directly sampled data using transparent and reproducible methodology
 Evaluate data QA/QC and distributions
 Compare spatial and temporal variability across sampling sites
Data needs
 Raw data from Olsen, Reidhead, and Barrus (partially acquired, verify that
information on wet/dry samplers, paired sampler height experiment, and
screened/unscreened sampler experiment data are available and designated)
 Raw data from W. Miller (acquired)
 Surface area of W. Miller sampler (for converting concentration and volume to areal
flux)
 Raw 2021 data

a. Data processing
i. Impute nondetects
ii. For each site and sampling date, convert raw data to areal flux
iii. Flag outliers
b. Data exploration
i. Description of sampling locations, sample size, and period of record for each
dataset and site
ii. Summary statistics for each site
iii. Summarize location, date, and constituent for flagged outliers
c. Visualization
i. Boxplots of flux at each site
ii. Time series plots of flux at each site
iii. Cumulative flux plots at each site
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

80

Output (see details in Analysis section above)
 Summary of data processing steps
 Summary statistics
 Visualizations
 Outlier tables
Decision points
 How to impute nondetects
 How to deal with missing data (interpolation, etc.)
 How to identify outliers

2. Evaluate outlier samples for potential explanations (i.e., collection methodology including
table height and screens; sources of contamination including insects, leaves, algae;
influence of local weather; and potential localized sources) (SP with Tt support)

2.1. Review and discuss previous Science Panel and third-party recommendations for
atmospheric deposition (SP with Tt support)
Purpose and goals
 Identify and review existing and relevant SP and David Gay recommendations to
inform the forthcoming analytical approach
Data needs
 Previously developed SP recommendations and summaries (acquired)
 David Gay review of reports and products (acquired)
Topics
 Sampler screening
 Sampler data QA/QC
 Excluding insects and other materials (e.g., algae, leaves) as sources of atmospheric
deposition
 Sampler height
 Wet/dry vs. bulk sampler design
 David Gay document review
 Attenuation
 Nutrient speciation
 Consideration of multiple lines of evidence: direct sampling, local & regional dust
modeling, mass balance, sediment accumulation, global reviews
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

81

Output
 Summaries of SP decisions/recommendations/discussions to date for each topic
 A list of specific topics for further investigation and proposed analytical
approaches for each.
Decision points
 SP recommendations for proceeding with data analysis, analytical approach, and
weighing lines of evidence. This recommendation may result in modifications to
subsequent tasks 2.2 through 6.
2.2. Evaluate outlier samples for potential explanations
Purpose and goals
 Statistically compare sampler design approaches (wet/dry and bulk samplers,
sampler height, screening)
 Investigate potential causes for outlier samples
 Recommend treatment of outliers to calculate fluxes
Data needs
 Date of screen installation
 Field notes and metadata for outlier samples for Barrus and W. Miller data,
namely information documenting presence of insects, leaves, algae
 Raw data from Barrus for paired screen/no screen samplers from Orem WWTP
and GSL locations
 Precipitation and wind data from local stations
Analysis
1. From the flagged outlier data in item 1, cross-check against sampling metadata to
determine if outliers are associated with (a) sampling methodology and/or (b)
sources of contamination and/or (c) other mechanistic explanations (i.e., weather,
localized sources)
2. From the flagged outlier data in item 1, evaluate whether outliers co-occur for
different nutrient constituents
3. Summary statistics and visualizations (e.g., boxplots) for groups of samples:
a. Boxplots of flux for sampler design experiments
b. Boxplots comparing groups of samples: non-outliers, outliers with
contamination identified, outliers with no contamination identified, and outliers
with no sampling metadata available
c. Visualizations for samples associated with weather events

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

82

4. Statistics
a. Paired statistical tests for comparing sampler design (e.g., paired t-test or
similar depending on appropriate assumptions)
b. Statistically compare sampler design approaches (wet/dry and bulk samplers,
sampler height, screening)
Output
 Tables of outlier samples joined with sampling metadata
 Statistical comparisons of sampler design and among-site fluxes
 Summary statistics and visualizations (see Analysis above)
Decision points
● Inclusion of data from various sampler designs
o Bulk and wet/dry samplers
o Sampler height
o Screened/unscreened samplers
● Type of materials to be considered contaminated vs. not contaminated (e.g., insects,
leaves, algae)
● Treatment (inclusion/exclusion, weighting relative to other stations) of outliers for
the following scenarios:
o If sources of contamination are confirmed to be present in the sample
o If sources of contamination are confirmed to be absent from the sample
o If no metadata are available
● Recommendation for the processed dataset with data subset that are approved for
inclusion in analysis
● Recommendation of areal flux rates
3. Evaluate spatial interpolation among sites and attenuation of fluxes (SP with Tt
support)
Purpose and goals
 Estimate the degree of attenuation of atmospheric fluxes moving from shoreline to
mid-lake
Data needs
 Literature evidence for attenuation over lakes (acquired)
 Bird Island data (acquired)
 David Richards analysis of spatial interpolation vs. mathematical averaging
(report acquired, additional data may be needed)
 Barrus 2021 analysis of Bird Island data

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

83

Analysis
1. Evaluate temporal aggregation across sites, with special attention to the order of
operations between spatial and temporal aggregation (Note: analysis will not include a
comparison of the approaches but rather an incorporation of the SP-recommended
approach following discussion)
a. Aggregate spatially, then temporally
b. Aggregate temporally, then spatially
2. Evaluate spatial interpolation across sites, with special attention to how to deal with
sites that differ from others in areal flux (local sources) (Note: analysis will not include
a comparison of the approaches but rather an incorporation of the SP- recommended
approach following discussion)
a. Spatial interpolation vs. mathematical averaging
b. Aggregation of central tendency (mean, geomean, median) among sites
c. Interpret fluxes at Bird Island relative to shoreline samplers
3. Summary statistics and boxplots
4. Time series plots
5. Statistical analysis of fluxes at Bird Island vs. shoreline samplers (paired statistical
test consistent with relevant assumptions)
6. Evaluate David Gay review of Bird Island data and implications
7. Analysis of the impact of assumptions for no attenuation and rapid attenuation of
loads moving from shoreline to mid-lake (Note: analysis of spatial interpolation
for each individual sampling date will represent an additional level of effort)
Output
 Summary statistics
 Summaries of flux-to-load conversions based on SP decisions
Decision points
 Order of temporal and spatial aggregation
 Method for spatial interpolation
 Combining or keeping separate G. Williams and W. Miller data
 Recommendation of how to handle Bird Island data
 Recommendation for attenuation, including a central estimate as well as upper and
lower bounds
4. Evaluate speciation (SP with Tt support)
Purpose and goals
 Loads of individual chemical species of N and P are needed as inputs to the
ULNM
 Identify relative proportions and absolute amounts of N and P constituents
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

84

making up total atmospheric loads
Data needs
 Previously developed studies and literature that estimate N and P species in
atmospheric deposition (acquired)
 Processed data from items 3 and 5 for individual constituents (e.g., nitrate and
ammonium in addition to total N loads)
Analysis
1. Summarize proportion of N and P loads made up of individual constituents for
literature-based estimates
2. Calculate proportion of loads with directly sampled data, where available
Output
 Summary tables of proportional and absolute loads of individual chemical
constituents
Decision points
 Identify if and when chemical constituents in directly sampled data are appropriate
to use to assign speciation. Considerations:
o Holding times
o Concentrations of individual constituents exceeding total concentrations for an
element
o Comparability of constituents among sampling approaches (e.g., SRP vs.
orthophosphate vs. bioavailable P)
5. Constraining Analysis Evaluation
Purpose and goals
 Compile atmospheric deposition estimates from constraining analyses to compare
to direct estimates
 Evaluate confidence and uncertainty of constraining analyses compared to direct
estimates
Data Needs
● Mike Brett updated mass balance, including updated sediment accumulation rates
and carp removal
● Reports and memos of previously completed constraining analyses (acquired)
Analysis
1. Compilation of flux and load estimates from constraining analyses

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

85

Output and Decision Points
● Summary of constraining analysis approaches, flux and load estimates, and
confidence/uncertainty

6. Determine loading for including in the ULNM
Purpose and goals
● Summarize SP Subgroup recommendations from items 1-6, including a
documentation of consensus-derived output and any dissenting perspectives from
non-consensus-derived output
● Summarize information from items 1-5 to recommend an estimate of atmospheric
loading of N and P to Utah Lake. Includes:
o Fluxes
o Attenuation
o Total load
o Speciation of chemical constituents
Data needs
● None

Analysis
● Synthesis of items 1-5

Output and Decision Points
 Recommendations memo detailing decision points from items 1-5 and a
recommendation for an estimate of atmospheric loading to Utah Lake

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

86

Appendix B: TP and DIN Constituents

Figure 42. Time series of SRP samples from the raw Williams dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

87

Figure 43. Time series of SRP samples from the processed Williams dataset, excluding any
samples that included insect contamination or did not have metadata available.

Figure 44. SRP to TP ratios from the 2020 Williams dataset. Values >1 area a functional
impossibility because SRP is a component of TP.
Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

88

Figure 45. Time series of nitrate samples from the raw Williams dataset

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

89

Figure 46. Time series of ammonium samples from the raw Williams dataset

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

90

Figure 47. Nitrate to DIN ratios from the 2020 Williams dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

91

Figure 48. Ammonium to DIN ratios from the 2020 Williams dataset.

Utah Lake Water Quality Study Atmospheric Deposition Decision Support Document

92

Appendix C: Diverging Perspectives Memo and Reference Material

A Critical Evaluation of Discussions, Presentations, and Recommendations by the Utah Lake SP Subcommittee
on Atmospheric Deposition

by

Theron Miller

Science Panel and Subcommittee Member

February 2023

1

A Critical Evaluation of Discussions, Presentations, and Recommendations by the
Utah Lake SP Subcommittee on Atmospheric Deposition

by

Theron Miller

Science Panel and Subcommittee Member

February 2023
2

Following are clarifications, additional evidence, and deliberations, including those based on
published data that refute the recent discussions, presentations and conclusions of the DWQ
Utah Lake Science Panel Subcommittee on atmospheric deposition. Many of these points I have
made for years, others are more recent.
The paper by Wilson and Serre (2007) reported on transport of NH3, not P. But even as such,
the decline in NH3 plateaued at about 5 mg/cubic meter at 2000 m. The SP subcommittee
needs to acknowledge that our Utah Lake N values are within the range of Wilson and Serre
(2007), Brahney (2019), and USGS data surrounding GSL.
However, there is still an important weakness in the Wilson and Serre study. In short, these
authors tracked ammonia along a linear transect downwind from several individual hog farms.
While measurements tracked attenuation, it was not entirely from settling. Wilson and Serre
(2007) sample design ignored the principle of radial jet theory, a published well-known
mathematical computation describing the lateral dispersion of point source discharges
radiating in a circular pattern outward, such as in a lake. This radial pattern serves to
dilute/disperse laterally the concentration/intensity as distance extends from the source, much
like the radial pattern of throwing a stone onto a lake surface. Therefore, the SP subcommittee
conclusions based on Wilson and Serre are faulted by not accounting for radial jet theory. That
is, ammonia was dispersing laterally at least as much as it was dispersing and settling
longitudinally. Consequently, the decline in ammonia concentrations was not entirely due to
particle settling, but rather to horizontal diffusion and dispersion following radial jet theory.
Moreover, according to the SP slide deck, this attenuation to about 5 mg/cubic m was applied
to regional dust data on Utah Lake rather than the local dust that it was supposed to represent.
We all need to accept that regional dust can include additional contributions from local sources,
hence characteristics of transport include a reconfigured blending of distant and local sources,
depending upon size fraction, particle structure, as well as wind patterns. As such, the
attenuation pattern for regional dust used by the SP subcommittee was likely misapplied.
Recently, to adopt the attenuation pattern of Wilson and Serre (2007) may be inappropriate as
it only refers to local simple ammonia values and not the complex of dusts surrounding Utah
Lake. It could be construed by outside observers that this was just a matter of convenience on
part of the SP subcommittee (i.e., only ammonia attenuation downwind of a CAFO) to show
that attenuation is as rapid as possible in order to oppose any data that demonstrates that fine
particles (likely up to 10 um; characteristic of Great Basin playa dust) can be transported at
much greater distances that 2000 m.
Recently, Scott Daily (DWQ) mentioned that the source of P and N at Mosida could be the
nearby dairy. However, the dairy is actually located 5500 meters away, although some drying
ponds associated with the dairy are about 1200 m distance. NADP guidelines state that AD
samplers need to be >500 m distance needed to be considered valid. Nevertheless, Barrus et al.
(2021), removed two of 39 P and three of 39 N samples at Mosida as outliers to address this
problem. After removal, the P results were lower than the averages of the other four sites and
3

the average N was near the average of the other sites. Thus, if at all, there was very little
influence from the drying ponds on the Mosida site data.
In addition, another panel member stated that there are “thousands and thousands of birds on
Bird Island”. This was apparently accepted as fact by the rest of the Subcommittee without
evidence and is highly conjectural. Birds would virtually have to sit on top of each other to fit
“thousands and thousands” on this tiny ¼ acre island that frequently gets inundated by waves.
Frequent wave inundation of the island also suggests that the accumulation and mobilization of
dust particles is insignificant to non-existent to be a local source. Therefore, while the panel
believes that there is some type of point source emanating from Bird Island, there is no
scientific support for such an assumption, or for removing this data. Another example, the
average of ammonia from the Bird Island site, the only volatile compound in the study, remains
near the values for the other sites. Elevated values for N and P at Bird Island are most likely only
a reflection of the elevated N and P at Mosida which is directly upwind of the afternoon
prevailing wind coming out of the SW. Moreover, I recently presented a logical hypothesis
implicating converging daily prevailing breezes, down canyon early morning breezes from
Spanish Fork and Provo Canyons, passing across the urban areas, traveling passed the airport
(see windrose) and on across the lake. These winds are immediately followed by the prevailing
SW midday and afternoon breezes heading directly across the lake (see windrose), and toward
the suburbs as indicated by Goodman et al. (2019). These breezes converge and subside in the
central portion of the lake (the coolest part of the valley) as demonstrated by the nearly daily
inversions that develop in the very bottom of the valley which is directly over Utah Lake. Hence,
this inversion is comprised of fine particulates accumulated from both directions. Moreover,
elemental analysis by Telfer et al. (in preparation) identifies the Cherry Creek area (about 10
miles SW of Mosida) as being most similar to all of the shoreline sites and Bird Island. Notably,
this site is directly in line with the Sevier Lake playa- which was the dominant source of urban
dust in Provo, identified by Goodman et al. (2019). This point needs to be firmly established
into our collective understanding of AD.
With respect to the distribution of the size range of particles, the complementary nature of the
different types or fractions collected from different samplers of AD has been described in a
well-documented recent white paper by Williams (attached). Moreover, this report has been
reviewed by Dr. David Gay, Director of the NADP. He has expressed his support for the
complimentary nature of the fractions described by Williams and is expected to provide a
written comment letter by February 20.
Comments on the SP Subcommittee “Constraining” document:
Dr. Theron Miller comments are highlighted in blue.
Constraining Document: “Areas with extreme urban pollution and high rates of biomass burning
are anticipated to have the highest concentrations of P deposition due to the high P
concentrations of combustion products (Mahowald et al. 2008, Brahney et al. 2015)”.
4

Dr. Miller Comments: This statement is inaccurate and misleading. In fact, Mahowald et al.
(2008) states that combustion composes only about 5% and for only some locations around the
world. Rather, 82% of deposition is mineral aerosols – which is much closer to the dust sources
around Utah Lake (Goodman et al. 2019).
Constraining Document: “A deposition rate of 1000 mg P/m2/yr is 9.5 x the standard deviation
above the global max. A deposition rate of 500 mg P/m2/yr is 7x the standard deviation above the global mean. Given TP deposition measured worldwide, values above 175 tons of P to Utah Lake need an explanatory mechanism, which is currently unknown.”
Dr. Miller Comments: We have provided an explanation, please read our material. We have an
abundance of data through multiple years, including the occasional wind and dust storms mobilize dusts from the hundreds of square miles of playa dust that surround Utah Lake. This is not typical of any deserts or otherwise in the worldwide data base. In addition, the accumulation of dust over the lake in the near-daily inversions has been discussed before and again, above. In
addition, the recent white paper by Williams explains how most samplers are designed to sample
one of the three fractions of AD. Read the Williams White Paper for details.

Figure 1. Distribution of measurements of P deposition worldwide. Data include a range of
landscape types, from remote to urban, agricultural, tropical, and near other shallow polluted lakes (e.g., Lake Taihu). From TetraTech.

Constraining Document:
• Biological Material is on average 1% P, or 10,000 ppm.
• Average Wasatch Front long range ‘Regional’ dust is 0.09% P (900 ppm) based on 12 samples (Reynolds et al. 2014, Brahney and Skiles, unpub),
5

• Local playas are reported to be similar at 1000 ppm or 0.1% P, (n=?) (Williams et al. 2022).
• Distal playas are on average 2152 ppm P
• Local urban dust (Provo, SLC, Ogden) are on average 2959 ppm P (n=12), Provo average dust P is 3916 ppm (n=4)
• The average of all regional and local dust samples is 2058 ppm (n=64)
Dr. Miller Comments: Goodman and Carling noted the enrichment of P in urban dust but had no
explanation. Throughout their paper Goodman et al. (2019) noted that the urban dust was 90% similar to the Sevier Lake playa dust to the SW, yet no explanation was offered. Here’s the likely explanation: Urban environments are characterized as hardened surfaces with very little opportunity for infiltration or washing off, except for the occasional rain. Rather, dust
accumulates on road surfaces, parking lots, gutters etc. Traffic, wind, and other disturbances
frequently and randomly remobilizes this material – allowing transport for various distances, depending on wind direction, velocity, and particle size. This accumulation/enrichment may proceed for weeks, between rainstorms.
Why did the “Constraining” authors decide to just average regional and local dust with no
defined reason and place more weight on regional dust from great distances (i.e.. averaging regional dust (based on 12 data points from hundreds of miles away) than with local dust. Also, why include data from Logan, Ogden and SLC sites with the Provo site. The Provo site is 33% greater than from other Wasatch front sites that are 50 to 215 miles away? This only dilutes the
Provo site, which is actually much more representative of the influence on nearby Utah Lake.
Constraining Document: “No one knows the true deposition rate to Utah Lake. We again can create unrealistic upper and lower bounds. The regional dust deposition rate is 6 g/m2/yr as measured (Brahney et al. 2019 and citations within)”

Dr. Miller Comments: As described, this value is only ascribed to far-range regional AD, which is only a portion of total AD. Thus, it is not comparable to the total of upwind regional, local, bulk, and precipitation washout samples. The Brahney samples were simply not collected near Utah Lake. Also see Williams white paper.

Note that the recent work of Telfer et al. (in preparation) identified a much closer source of dust (Sites 4 and 8; Figure 3 below) than the sources identified by Goodman et al. (2019). Moreover, these samples were highly similar to the dust samples filtered from the wet/dry samplers, including the samples from Bird Island. The Bird Island samples were more related to Cherry
Creek dust (10 km from the Mosida sampler and 23 km from the Bird Island sampler; Telfer et
al. in preparation) and its deposition by contact. There is much more discussion on the fraction of AD that is composed of regional dust from Dr. Williams’ analysis (attached). Constraining Document:
Deriving an associated dust deposition rate:
Bounding information:
• Biological Material is on average 1% P, or 10,000 ppm.
6

• Average Wasatch Front long range ‘Regional’ dust is 0.09% P (900 ppm) based on 12 samples (Reynolds et al. 2014, Brahney and Skiles, unpub),
• Local playas are reported to be similar at 1000 ppm or 0.1% P, (n=?) (Williams et al.
2022).
• Distal playas are on average 2152 ppm P
• Local urban dust (Provo, SLC, Ogden) are on average 2959 ppm P (n=12), Provo average dust P is 3916 ppm (n=4)
• The average of all regional and local dust samples is 2058 ppm (n=64)
*Urban ground level dust will not travel over the entirety of the lake Dr. Miller Comments: Apparently, the SP subcommittee authors have not carefully read
Goodman et al (2019). Nor have they witnessed the near-daily inversions that cover the entire
lake. 90% of urban dust in Provo is of similar composition to the Sevier Lake playa dust. Moreover, the fact that it is enriched before it enters a sampler on top of the Geology building at BYU proves that it is re-mobilized after it is enriched, so its physical nature likely has not changed. If it is the accumulated and resuspended playa dust from the Sevier Lake playa, as
reported by Goodman et al. (2019), it has already travelled long distances (>100 km). For
example, if they measured this “urban dust” at the top of the BYU geology building at the edge of the mountains (which experiences daily downslope breezes, heading toward Utah Lake), the” Constraining” authors have made an incorrect assumption about the distance urban dust can travel - similar to Brahney’s incorrect assumption that the urban dusts extend only 200 to 600 m
over the lake – which is actually less than 10% of the lake surface. It should be recognized that
dust from a local gravel road is not similar to dust that has already travelled more than 100 km and temporarily settled in urban Provo (Goodman et al. 2019). If it’s 90% the same dust, it is logical to hypothesize that it travels downslope and back over the lake, also as an aerosol, as the inversion develops or strengthens. We will be testing this hypothesis as soon as it is safe to travel
on the lake.

7

Figure 3. Samplers were placed around Utah Lake with one sampler on Bird Island.
The dust source locations were chosen at the southern half of Utah Lake and in a
southwest direction towards Sevier Dry Lake. Sampler abbreviations: PS=Pump
Station, BI=Bird Island, LS=Lake Shore, MO=Mosida. Source numbers: 1=Lake Shore,
2=Eagle Mountain, 3=5-Mile Pass, 4=Chimney Rock Pass, 5=Goshen WMA, 6=Elberta,
7==Mouth of Spanish Fork Canyon, 8=Cherry Creek, 9=Fumerole Butte, 10=Sunstone
Knolle, 11=Sevier Dry Lake, 12= Cricket Mountain, 13=White Hills near Sevier Lake,
14=Mid-Sevier Lake near road, 15=Highway 6 South of Delta, 16=Burraston Ponds,
and 17=Miners Canyon. (From Telfer, in preparation)
To revisit the “Constraining paper”
Constraining Document: “Average urban dust deposition rates in Utah are 40.5 g/m2/yr (range
25-57 g/m2/yr) (Goodman and Carling 2019, Scholz and Brahney 2019)[n = 37]. “These values are similar to urban measurements made elsewhere in the western US and world (Lawrence and Neff 2009, Brahney et al. 2015, Brahney 2019 and references within).
8

Given that urban areas receive the greatest dust deposition rates (from intense local generation.”
Dr. Miller Comments: No, actually it’s local enrichment/accumulation of dust that arrived from
distant playas. The enrichment of urban dust P is due to the accumulation of dust from the original sources (southwest playas; (Goodman et al. 2019). The continued accumulation of dust combined with the local turbulence, (e.g., cars passing by on hardened surfaces), that resuspends this very light material.

Constraining Document: “… that dust deposition rates will attenuate away from their ground level source, it can be expected that the average dust deposition rates to the entirety of Utah Lake should not be greater than 40.5 g/m2/yr.”
Dr. Miller Comments: This is highly unlikely. That urban sources are from playa dust 20 to >100
km away demonstrates that long-range transport has already occurred. The realization that these urban dusts originated from distant playas can now support the evidence that dust can travel back downslope and across the lake, including Bird Island, as evidenced by the frequent inversions – not just 600 m from town. Again, the panel has ignored the fact that inversions, obviously
comprised of local dusts, spread over the entire lake surface at the bottom of the valley, on a
near-daily frequency (transported by the local downslope breezes – as indicated by the Provo airport windrose), Moreover, we verified each weekly visit that there was no evidence of bird droppings or insects
in the bucket, on the screen or on the table surfaces. This is logical in that the water birds that
frequent Bird Island (predominantly California Gulls, Franklin Gulls, Caspian Terns, White Pelicans) have webbed feet that are poorly designed for grasping and perching on thin sampler bucket rims. The distance from Utah Lake’s shoreline to Bird Island precludes passerines (perching birds) to identify as a destination or to risk predation during the journey.

Also, there is no evidence supporting the assumption that Bird Island acts as its own point source. It’s just speculation to support the decision to throw out the data. We will test this hypothesis during the 2023 season using mobile air samplers. If one examines the N (and the P) data, Bird Island data is very similar (not significantly different) to the shoreline data (i.e., the
best chance for being a point source would be elevated values for volatile compounds such as
ammonia – but data doesn’t support this. And P has no gaseous phase. The similarity of urban dusts to distant playa dusts (Goodman et al. 2019) and the similarity of dusts deposited in the Bird Island sampler to Cherry Creek dusts (Telfer in Preparation) also does not support the conjecture that Bird Island functions as its own point source.

Constraining Document: “Data assimilated from global observations indicate that only a few sites in the world have deposition rates greater than 100 g/m2/yr, and these sites occur is desert areas such as the Gobi Desert in China as well as Libya and Niger in the central Saharan Desert. Areas that receive approximately 50 g/m2/yr include the Loess Plateau in China, regions of Israel,
and Phoenix, Arizona."
Dr. Miller Comments: Yes, but these locations are not affected by nutrient-rich playa dusts originating in the middle of an ancient lake, Lake Bonneville, that resided in the middle of even
9

more ancient and uplifted (and eroding) 300 to 500 million year old shallow sea sedimentary deposits.

Constraining Document: “No one knows the true deposition rate to Utah Lake. We again can create unrealistic upper and lower bounds. The regional dust deposition rate is 6 g/m2/yr as measured (Brahney et al. 2019 and citations within).”
Dr. Miller Comments: This is not comparable to the total of regional, local bulk and precipitation
washout samples. See Williams white paper, attached. Constraining Document: “…and an upper bound of 40.5 g/m2/yr as the urban rate. As above, neither is likely given that the average dust deposition rates (for the full lake area) will be
somewhere between the two boundaries.”
Dr. Miller Comments: Not necessarily, as I wrote above. Constraining Document: “Canyonlands, a well-known dust producing region in arid southern
Utah provides a median deposition rate, which is 29 g/m2/yr (Reheis and Urban 2011, Brahney et
al. 2020). Dr. Miller Comments: These samples were not collected near Utah Lake.
Please acknowledge that Utah Lake and GSL are in the middle of a giant playa - with huge
amounts of well-known P-rich dust emissions. For example, compare to the Owens Lake playa which is about 0.0000001 the size of playa as that surrounding Utah Lake and GSL( e.g.. see our SAP). Similar dust storms have been observed blowing over Utah Lake and GSL. Read Reheis (1997), another paper that was omitted in the Brahney white paper. Table 1, in that paper reports
dust deposition rates of 7.8 mg/m2/day to more than 2100 mg/m2/ day. Converted to an annual
rate this equates to 2.847 g/m2/yr to 792 g/m2/yr – a little higher (more than an order of magnitude) than your 40.5 g/m2/yr “unrealistic” upper bound from regional sites. To follow through, even at 1% P this will dwarf all SP Subcommittee estimates provided in Figure 1. And this is exactly comparable to playas surrounding UL as the dust source was the Owens Lake
playa formed from a century of diverting tributaries to LA.
Also, “Canyonlands is a well-known source of dust”? I don’t agree. (reference?). Where are lake-deposition-based, fine dust-producing playas in Canyonlands National Park. I have been there many times. Not too surprising, it’s full of canyons that trap dust and sediments, and is
completely different from the geography and geology surrounding Utah Lake. We have
presented considerable evidence contradicting or rather, supplementing, the Brahney dataset. For example, Brahney ignored an important paper (Jassby et al. 1994) which demonstrates that fine particles can travel 20 km, to the middle of Lake Tahoe with only a 12% reduction in deposition and which was not a statistically significant reduction. Notably, adjusting for lake size, the 4-yr
study by Jassby et al. (1994), results in 150% of the SRP as that purported to occur on Utah Lake
by Brahney (2019). This is surprising in that Lake Tahoe is in a forested alpine basin at 6500 ft elevation with the only sources of dust being the Central Valley of California (80 miles away), and which must travel over the 9500-ft Sierra Nevada mountains or the Mohave Desert of
10

Southern California or Southern Nevada, (about 200 miles away). This 4-yr study, with hundreds of samples, simply contradicts Brahney’s original attenuation estimates of 200, 400,
600, and now 2000 m for regional dust from the eastern shoreline. The panel needs to be using
actual comparable data rather than picking unrepresentative data or making conjecture or rates just to put arbitrary “constraints” on attenuation. It is inappropriate to ignore relevant data (such as Jassby et al. (1994) and Reheis, 1997), especially when playa dust located >100 km SW of Utah Lake is the dominant dust in urban Provo, on the opposite side of the lake, using non-peer-
reviewed estimates of P transport across the lake while attempting to ignore peer reviewed data
from Barrus et al. (2021). I suggest that as unbiased scientists, we need to reconfigure this whole paradigm. Constraining Document: Goshen Bay sedimentation rate 1.7 mm/yr, *Mitch Power freeze cores
north of Provo Bay show similar rate.
Dr. Miller Comments: This has not been actually quantified or peer-reviewed. Freeze cores don’t have that level of resolution as sediment traps. We are all aware that sediment mixing and resuspension in shallow lakes (for which Utah Lake is famous) from frequent winds and carp
bioturbation precludes accurate measurement of sedimentation rates. Because sediment traps
were not used, this is questionable data. Constraining Document continues:
Provo Bay sedimentation rate: 2.6 mm/yr
Bird Island sedimentation rate: ~1 mm/yr Dry density of Utah Lake surface sediments at Goshen Bay: 0.7 g/cm3 Dry density of Utah Lake surface sediments at Bird Island 0.55 g,
Dry density of Utah Lake surface sediments at Provo Bay: 0.5 g/cm3
The above information is consistent with modern sedimentation rates measured in other waterbodies throughout North America (Brothers et al. 2008)(Appendix A).
Dr. Miller Comments: Possibly, but sedimentation on GSL is 3.5 – 4.5 mm per year – in the
middle of GSL (USGS Se study; using sediment traps). Given the several years of the SP’s existence, the SP Subcommittee now needs to acknowledge that Utah Lake and GSL are not similar to most other lakes in the world, particularly in regard to AD, morphology and geography. Utah Lake and GSL are in the middle of a unique dust-filled giant playa. For
example, compare these to Owens Lake which is about 1/100000 the size of playa as that
surrounding Utah Lake and GSL, see our SAP. Constraining Document continues: One would anticipate areas closer to the shore and thus closer to sediment sources and areas with higher production, would have greater sedimentation
rates.
Dr. Miller Comments: This assumption ignores the principle of sediment focusing which distinctly occurs in shallow GSL.
11

Constraining Document Continues: In addition to catchment and production sources of material
for sedimentation, authigenic calcite makes up a sizeable portion of the sediment accumulation.
Using estimated AD deposition rates from 5 to 350 Tons we can determine plausible AD deposition rates of P given that deposition rates should (at least) not exceed measured sediment accumulation rates in all parts of the lake

Dr. Miller Comments: Sediment accumulation rates used here are not accurate. They were not measured. AD Tons of P to Utah Lake 0.55 g/cm3 0.64 g/cm3 mm/y mm/y
5 0.03 0.02
25 0.13 0.11
75 0.39 0.33
175 0.91 0.78
350 1.82 1.56

The constraining document continues: “If we again assume a dust deposition rate of 29 g/m2/yr, we arrive at a sediment accumulation rate of 0.17-0.28 mm/yr, or 10 to 16% of the measured sediment accumulation rate”

Dr. Miller Comments: What happened to the 1.7 mm reported by Mitch Powers?, whose data, I mentioned, is 1.5 orders of magnitude less than MEASURED (using sediment traps) GSL sediment accumulation rate. And 1.5 orders of magnitude less than MEASURED dust deposition from the Owens, Lake Playa. If we use the GSL MEASURED rate, and even close to that
measured from the Owens Playa dust emissions there is plenty of deposition to cover the 350
tons of P to Utah Lake.

Concerning Speciation
From Dr. Wood Miller’s data (approximately 2000 samples):
The Dr. Wood Miller data set includes bulk samples but uses quite a different method from other
bulk samplers (i.e., Bunt cake tins or plastic totes with the bottoms generally covered with
marbles) which may or may not have been analyzed for both Total P and ortho-P. Throughout all
the sample sites and the entire sampling period, the proportion of ortho-P ranged from 41 to 62%
of the total P (see inserted “pictures” of data tables). A large amount of the total P is biologically
available.
Jassby’s et al. 1994 work on Lake Tahoe resulted in SRP to TP ratios of 44 to 47% (Table 4
below). Jassby et al. 1994 and Wood Miller’s data agree quite well. Yet, both are quite different
12

from the summary by Tetra Tech which uses a value of 33 to 37%. Where does this data come
from?

Dust content of ca and mg
Mitch Hogsett (Science Panel co-chair) stated that the majority of P in playa dust would be
bound to calcite.
Dr. Miller Comments: This is likely not true. Dust data from Goodman et al. 2019 found Mg to
be 2X to 3X more concentrated than Ca in the playa dust and Carling et al. and Randall et al. has
reported that 40% of sediment P is associated with Fe. In short, redox reactions between the high
concentrations of Fe/Mg and P are occurring. This was clearly illustrated by Hogsett et al. (2019)
where they report that 1500 tons of P is potentially mobilized from the sediment per year. This
was the conclusion from Hogsett’s experiment using DO chambers, in contact with the sediment
surface, to near 0 mg/L DO – an experiment illustrating the effect of redox reactions on P
speciation. This value dwarfs atmospheric deposition of P, POTW discharge stormwater runoff
and tributary contribution combined. Large releases of sediment P during hypoxic events are not
unusual. The release of phosphorus from bottom sediments can contribute up to 99% of the total
P pool in shallow lakes (Bostrom et al. 1988, Jensen and Andersen 1992; Hullebusch et al.
2003). Unfortunately, Hogsett’s paper, including the contribution of P and N, have not been
discussed by the SP and therefore, I am concerned that this large contribution of nutrients to the
water column – likely causing the blooms, as Michael Brett pointed out, will be
underrepresented.
There is another important measurement within the Jassby et al. (1994) paper. Dry deposition
data indicated that attenuation of SRP from AD included a reduction by about 12% from a
shoreline sampler to a sampling bucket located 20 kilometers (12 miles) from shore. This value
is not significantly different from the shoreline data. For comparison, the sampler at Bird Island
was only about 6 kilometers from the shoreline. In turn, there was no attenuation of P in the AD
at Bird Island as compared to the shoreline sites, a value quite comparable to the 20-km range in
the Lake Tahoe Study. Also, particle sizes were notably smaller than from other sampling sites –
indicating that they can travel long distances (See Williams white paper).
13

Another important comparison: dry deposition of P on Lake Tahoe was about 3.5X greater than
wet deposition (Jassby et al. 1994). Dry deposition on Utah Lake is about 3.5X greater than wet
deposition (Olsen et al. 2018).
However, Jassby et al. (1994) noted that AD collected in 2-m “snow tubes”, used to estimate wet
deposition rates during winter, were quite different. P measured in the snow tubes at the 20 km
site fell off by 90% compared to the shoreline site. There is no explanation given – neither by
the authors nor Tetra Tech. After “Google Scholaring” for about 45 minutes, I could not find a
peer-reviewed paper that addresses the difference in wind-driven snowfall (in a horizontal
direction) across flat surfaces such as ice cover vs snowfall in rugged or forested terrain such as
that surrounding Lake Tahoe. This is related to the idea of snow fences throughout the high
plains of Wyoming – allowing snow to settle rather than continuing to drift horizontally. The
point being that snow is much more likely to settle in the wind-protected rough, forested zones
around Lake Tahoe (falling vertically into the snow tube), than on the ice surface on a 40 by 70-
kilometer lake – where comparatively, any breeze will blow the snow horizontally across the
opening of the snow tube rather than allowing it to fall in it. I have witnessed this effect
personally during my research on the ice-covered lakes in the boreal forest of Northern Alberta.
Lake surfaces would often be snow-free while several feet of snow would accumulate within the
adjacent aspen forest. This was particularly true for the larger lakes.
In conclusion, I have provided additional evidence for supporting the Williams and W Miller
data, reports and publications. At the same time, I have provided substantial published data and
other scientific measurements that dispute many of the assumptions used to cull or otherwise
ignore Williams and W Miller’s data and particularly when the panel has chosen to impose an
unrealistic attenuation pattern on the lake. The empirical evidence just doesn’t go there. Finally, I
urge all panel members to read the attached white paper by Williams that presents a logical and
scientifically supported arrangement of ALL data presented that describes how these different
sampling strategies actually supplement or complement each other in an additive manner – rather
than contradict each other.

14

Literature Cited
Anonymous, 2022. Constraining atmospheric deposition of phosphorus based on dust
deposition, sediment accumulation, mass balance, and bootstrap models.

Boström, B., Andersen, J. M., Fleischer, S., & Jansson, M. (1988). Exchange of phosphorus across
the sediment-water interface. Hydrobiologia, 170, 229–244.
Brahney, J. (2019). Estimating total and bioavailable nutrient loading to Utah Lake from the atmosphere.
Goodman, M. M., Carling, G. T., Fernandez, D. P., Rey, K. A., Hale, C. A., Bickmore, B. R., ...
& Munroe, J. S. (2019). Trace element chemistry of atmospheric deposition along the Wasatch Front (Utah, USA) reflects regional playa dust and local urban aerosols. Chemical Geology, 530, 119317.
Hogsett, M, L. Hanyan and R. Goel. 2019. The Role of Internal Nutrient Cycling in a Freshwater
Shallow Alkaline Lake. ENVIRONMENTAL ENGINEERING SCIENCE Volume 00, Number 00,
2019

Hullebusch, E., Auvray, F., Deluchat, V., Chazal, P., & Baudu, M. (2003). Phosphorus
fractionation and short-term mobility in the surface sediment of a polymictic shallow
lake treated with a low dose of alum (Courtille Lake, France). Water, Air and Soil
Pollution, 146, 75–91.
Jassby, A.D., J.E. Reuter, R. P. Axler, C. R. Goldman, and S. H. Hackley. 1994. Atmosphere
deposition of nitrogen and phosphorus in the annual nutrient load of Lake Tahoe
(California-Nevada) Water Resources Research. VOL. 30, NO. 7, 2207-2216

Jensen, H. S., Kristensen, P., Jeppesen, E., & Skytthe, A. (1992). Iron-phosphorus ratio in surface
sediment as an indicator of phosphate release from aerobic sediments in shallow
lakes. Hydrobiologia, 235(236), 731–743.
Randall MC, Carling GT, Dastrup, DB, Miller T, Nelson ST, Rey KA, et al. (2019) Sediment
potentially controls in-lake phosphorus cycling and harmful cyanobacteria in shallow,
eutrophic Utah Lake. PLoS ONE 14(2): e0212238.
https://doi.org/10.1371/journal.pone.0212238

15

Reheis (1997; Dust deposition downwind of Owens (dry) Lake, 1991-1994: Preliminary
findings, J. Geophysical Research. Vol 102, No. D22, 25,999-26,008
Tyler, S.W., S. Kranz, M.B. Parlange, J. Albertson, G.G. Katuld , G.F. Cochranb , B.A. Lylesb , G.
Holdere 1997. Estimation of groundwater evaporation and salt flux from Owens Lake,
California, USA. Journal of Hydrology 200: 110–135. Version 1.9

Winter, J.G., P. J. Dillon, M. N. Futter, K. H. Nicholls, W. A. Scheider and L. D. Scott 2002. Total
Phosphorus Budgets and Nitrogen Loads: Lake Simcoe, Ontario (1990 to 1998). Journal
of Great Lakes Research 28, (3), 301-331.

Atmospheric Deposition of Nutrients to Utah Lake:
Process and Research Overview

Developed for:
Wasatch Front Water Quality Council Leland Myers, Project Manager

Developed By:
Gustavious Williams, Ph.D.

February 2023

UTAH LAKE ATMOSPHERIC DEPOSITION
Page ii
Executive Summary
In this paper, we classify atmospheric deposition (AD) into three different processes: settlement (dust),
contact, and washout. Settlement occurs when large particles, 10 – 100 µm leave the atmosphere due to gravity. They settle on the ground and are only resuspended by a strong wind or mechanical action. Contact
refers to smaller particles, less than 10 µm (PM10), and especially less than 2.5 µm (PM2.5) which do not settle (in general) and only leave the atmosphere when they contact a surface Washout refers to particles that are washed out of the atmosphere during a precipitation event. This includes dust (> 10 µm), fines (< 10 µm),
and gases. For all three processes, if the particles come in contact with the Utah Lake surface, they are captured and not resuspended, they stick when they contact the water surface.
Using this classification, we can describe nutrient AD using the following equation: 𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡=𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡+𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡+𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝 Eq 1
Where 𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 is the total nutrient AD on Utah Lake, 𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡 is the nutrient AD from settleable dust, 𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡 is the nutrient AD from fine particles less than 10 µm that are deposited by contact with the water
surface, and 𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝 is the nutrient AD from materials washed out of the atmosphere from a precipitation
event.
We have results from separate studies of nutrient AD on Utah Lake, each focusing on a different type of
deposition, settlement (𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡), contact (𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡), or washout (𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝). Brahney [1] performed a literature
review and estimated that 𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡 is in the range of 2 to 9 tons/yr. Miller (unpublished) collected 850 samples
around Utah Lake and measured concentrations in rainwater and estimated 𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝 of 88 to 142 tons/yr.
Barrus, et al. [2] evaluated 306 samples collected around Utah Lake and estimated that the total AD (𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡)
was between 133 to 262 tons/yr. Using these estimates, we can conclude that AD from fines suspended in
the atmosphere (𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡) ranges from 36 to 43 tons/yr, depending on which of the above numbers are used.
Based on these calculations, this means that the AD from dust is only 1.5% to 6.8% of the total nutrient AD. This is supported by recent work by Telfer (unpublished) who measured dust concentrations in samples
around the lake and found minimal dust, with annual dust deposition rates of 2.14 to 5.85 g/m2/yr. These rates are similar to, but significantly lower than those reported by Brahney [1]. However, these samples were
not designed to measure dust, some samples were discarded before being measured, and any soluble dust was
not measured, so these numbers are reasonable.
The Utah Lake AD studies initial appear to contradict each other because of the wide range of AD estimates. However, when considering that AD is driven by different processes, contact, dust (settlement), and precipitation, and that each study mostly measured only a subset of the total, it is clear that the studies are not
contradictory, but rather complement and strengthen each other.
Based on this analysis, we conclude that an annual AD TP loading of rate of 250 tons/yr to Utah Lake is accurate. However, in consideration of the range of findings and potential implications, we propose 150 tons TP/yr could be used as a consensus-based value for evaluation.

UTAH LAKE ATMOSPHERIC DEPOSITION
Page iii
Table of Contents
Executive Summary ii Background 4
Particulate Matter 4 Atmospheric Particulate Settling 4 PM2.5 and PM10 4
Atmospheric Deposition Mechanisms 5 Utah Lake Nutrient AD Studies and Measurements 6 AD Studies 6
Supporting Study 7 Discussion 8
Conclusion 9 References 10 Appendix 12

UTAH LAKE ATMOSPHERIC DEPOSITION
Page 4
Atmospheric Deposition of Nutrients to Utah Lake:
Process and Research Overview
Gustavious Paul Williams, Ph.D. Background
Particulate Matter
Particulate matter in the atmosphere (also called PM or particle pollution) is a complex airborne mixture of solid particles and liquid droplets. Atmospheric particle size is generally reported in micrometers (µm) or 10-6
m. The particulates in the atmosphere can range in size from a few nanometers to several micrometers. Though PM ranges widely in size, it has been divided into two categories based on diameter. PM2.5 are
particles with a diameter smaller than 2.5 µm and are also called fine particles PM10 are particles with a
diameter between 2.5 µm and 10 µm and are also called inhalable coarse particles. Particles larger than 10 µm (e.g., sand and large dust) are not regulated by EPA (https://health.utah.gov/utahair/pollutants/PM)
Particulates in the atmosphere are composed of a mixture of gases and particulates such as fumes, smokes
and other small solid and liquid particles. One common set of gases relevant to nutrient deposition are
nitrogen-oxygen species typically called NOx. NOx is mostly anthropogenic and a major contributor to atmospheric pollution and can be notices as a brown haze during summer months. NOx reacts with other
pollutants to form fine particulate matter in the atmosphere [3]. The process of NOx formation from organic
compounds involves the reaction of nitrogen-containing volatile organic compounds (VOCs) with atmospheric ozone in the presence of ultraviolet (UV) light. This reaction produces nitrogen oxide radicals
that further react with other atmospheric species to form nitrogen dioxide (NO2) and other species. NO2 is a key component of photochemical smog, which is a type of air pollution that is associated with urban and
industrial areas. These compound form particulates, with most of the nitrogen compounds present as the
ammonium salts in particulate form, except for ammonium nitrite which is a gas. Atmospheric Particulate Settling
The way particulates settle in the atmosphere depends on their size and weight, with larger particulates
settling faster than smaller particulates. For example, dust particles may settle within a few hours, while
smaller particles, like PM2.5, can stay in the atmosphere for days to weeks.
Hinds, et al. [4] state that particle size is the most important parameter for characterizing aerosol behavior. They show that particles with equivalent diameters of 0.1, 1.0, 10, and 100 µm settle in perfectly calm air at
8.8x10-7, 3.7x10-5, and 3.1x10-3, and 2.5x10-1 meters per second (m/s), respectively. In terms of time, this
means that the 0.1, 1.0, 10, and 100 µm particles require 315 hours, 7.5 hours, 5 minutes, and 4 seconds to settle 1 meter in perfectly calm air, respectively. This means that for particles smaller than about 10 µm
(PM10) a light breeze can keep the particle from settling. While PM2.5 particulates such as photochemical
smog (mostly nitrogen particles), smokes, and fine dust essentially do not settle from the atmosphere, but are kept aloft by Brownian motion and wind currents [4]. For these particles, gravity is not an effective removal
mechanisms, but they are removed from the atmosphere by contact with a surface or by washout from precipitation. Contact can either by a dry surface where static charges capture the particle, or wet surfaces. Static surfaces soon fill, and subsequent particles either are not captured or displace an existing particle which
is resuspended. Wet surfaces, such as Utah Lake, capture any of the fine particles that touch the surface. PM2.5 and PM10
Particulate matter less than 10 µm (PM10) and particulate matter less than 2.5 µm (PM2.5) in diameter are
monitored as indicators of air quality. Kuprov, et al. [5] studied atmospheric pollution in Utah and noted that
Utah Valley, the location of Utah Lake, is a non-attainment area for PM10 and PM2.5 which means that particulate levels are high for these particle sizes.

UTAH LAKE ATMOSPHERIC DEPOSITION
Page 5
PM2.5 levels 98th percentile averaged across 18 air quality monitors throughout the Wasatch Front from 2004 to 2015 (https://health.utah.gov/utahair/pollutants/PM/) was between 40 and 50 µg/m3 while the 98th
percentile for PM10 averaged about 10 µg/m3 over the same time period. The Wasatch front met the 1997 PM2.5 24-hour primary standard of 65 µg/m3. However, when this standard was lowered to 35 µg/m3 in
2006, the Wasatch Front were unable to comply and were re-designated as nonattainment with levels that exceeded that value. For PM10, Salt Lake and Utah Counties have been designated as nonattainment for the 24-hour primary standard of 150 µg/m3. In general, both areas have been in compliance with the national
standards since 1996, but disagreements regarding the classification of exceedances due to windborne dust during high wind events have prevented re-designation to attainment or maintenance (https://health.utah.gov/utahair/pollutants/PM/).
The State of Utah (https://health.utah.gov/utahair/pollutants/PM/) estimates that over 30% of primary
PM2.5 particle emissions in Utah came from dust in 2011. Fires contributed over 15%, while fuel combustion and mobile sources emitted 15% and 12.5% of the total primary PM2.5 particles, respectively. However, most PM2.5 is made of secondary particles, those formed in the atmosphere from other pollutants such as
NOx as discussed above. Regarding PM10, the State says that in 2014, 67% of the primary PM10 particles in Utah came from dust, largely from unpaved roads with other sources including agriculture and industrial processes which contributed approximately 17% and 5%, respectively, of the total primary PM10 particles.
There are also secondary PM10 particulates, but not as prevalent as secondary PM2.5 particles.
During a precipitation event, most of the particulate matter in the atmosphere, including PM10 and PM2.5, is washed-out or deposited with the precipitation. For discussion purposes, if we assume that the particulate pollution above Utah Lake extends 1,500 meters above the valley floor, about half way up the mountains, and
we assume that the concentration is 10 µg/m3 (the approximate average from 2004 to 2015) there would be 15 mg/m2 of PM2.5 AD for each precipitation event if the event washed out all the particulates. For PM10, assuming 150 µg/m3, (the non-attainment value), AD would be 225 mg/m2 per precipitation event.
Atmospheric Deposition Mechanisms
For this discussion we classify atmospheric deposition (AD) into three different processes: settlement (dust),
contact, and washout. Settlement occurs when large particles, 10 – 100 µm leave the atmosphere due to gravity. They settle on the ground and are only resuspended by a strong wind or mechanical action. Contact refers to smaller particles, less than 10 µm, and especially less than 2.5 µm, which leave the atmosphere when
they contact a surface and “stick” because of electrostatic charge or moisture. These particles do not settle, and surfaces soon become “saturated” so that additional particles either are not captured or displace an existing particle. These smaller particles can settle slightly, but they are easily resuspended if they are not
attached to a surface. Washout refers to particles that are washed out of the atmosphere during a precipitation event. This includes dust (> 10 µm), fines (< 10 µm), and gases. For all three processes, if the particles come
in contact with the Utah Lake surface, they are captured and not resuspended, they stick when they contact the water surface.
Using this classification, we can describe nutrient AD using the following equation: 𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡=𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡+𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡+𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝 Eq 1
Where 𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 is the total nutrient AD on Utah Lake, 𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡 is the nutrient AD from settleable dust, 𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡 is the nutrient AD from fine particles less than 10 µm that are deposited by contact with the water
surface, and 𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝 is the nutrient AD from materials washed out of the atmosphere from a precipitation
event.

UTAH LAKE ATMOSPHERIC DEPOSITION
Page 6
Utah Lake Nutrient AD Studies and Measurements
AD Studies
We have results from at least four separate studies of nutrient AD on Utah Lake, each focusing on a different
type of deposition, settlement (𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡), total (𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡), or washout (𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝). Some of these studies
collected measurements designed to quantify AD on Utah Lake, others use data from literature studies to
estimate AD rates, including literature studies from Utah locations, but not near Utah Lake. While all these studies also report on nitrogen loads, for discussion here we will only consider phosphorous. Nitrogen loads are similar, with the exception that gas solution from the atmosphere to the lake or into precipitation also
occurs.
Dr. Janice Brahney performed an in-depth literature review of dust deposition in Utah and elsewhere in the
Great Basin [1]. This work summarizes current knowledge on total and soluble phosphorus loading from dusts and summarizes atmospheric deposition rates of nitrogen from wet, gaseous, and particulate sources as
reported in the literature. This includes urban depositions, both on the Wasatch Front and other locations, and deposition in the mountains and summarizes 7 different Utah dust measurements reported in two different studies (see Table 2 in [1]). Brahney [1] generates estimates of total (urban + regional) nutrient
loading to Utah Lake based on these literature values and found that 80% of estimates fell between 2 and 9 metric tons of Total Phosphorus (TP) deposition to Utah Lake per year, with estimates of the bioavailable
fraction at a minimum of 0.5 to a maximum of 7.9 metric tons, with probable deposition rates between 2 to
2.5 metric tons per year. Brahney [1] states that the study does not consider wet deposition. The Utah studies (Table 2 in [1]) all use the bulk marble collection method which generally captures dust, but not smaller
particles such as PM10 or PM2.5. While some small portion of these fine particles are captured, many are resuspended or not captured at all because they do not settle onto the collector. Brahney [1] reports TP dust concentrations as deposition rates, rather than as mass concentrations in the dust. It is not clear how dust
mass measurements from Table 2 were converted to TP deposition rates. It appears that phosphorous concentrations and deposition rates from other studies were used to estimate TP rates, rather than data from
the Utah sites, though this is unclear. This study mostly measures dust (𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡) though some fines are
captured.
Dr. Greg Carling (personal communication with Theron Miller) measured AD at four locations along the Wasatch front at university campuses in Provo, Salt Lake, Ogden, and Logan using the bulk marble method. He measured phosphorous concentrations in the dust and estimated that ~55 tons/year of TP is deposited
on Utah Lake. The closest measurement in this case was at the BYU campus in an urban area surrounded by lawns and pavement about 300 meters, or more, above the Utah Lake surface and away from dust storms on
the valley floor. Again, the bulk marble method mainly measures dust (gravity) AD, and not contact or
precipitation AD. This study mostly measures dust (𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡) AD, though some fines are captured.
Dr. Woodruff Miller has collected rainfall at 9 locations around Utah Lake for the past 6 years amounting to a total of 850 samples. These samples were collected close to the lake, away from any local dust source. He
collects the precipitation samples shortly after a rainfall event and sends the water to a certified laboratory for analysis. The laboratory measures and reports concentration of TP in mg/L. Dr. Miller uses the precipitation gage at the Utah Lake outfall to determine the amount of precipitation that falls over the lake surface. His
previous reports used an average TP concentration along with the precipitation among and the area of Utah Lake on that date, to compute the amount of TP deposited to the Lake. His estimates range from 88 - 142
tons/year of TP. A more recent analysis (not yet published) uses 7 rain gages around the lake along with geostatistics to compute rainfall intensity maps with the same methods used with data from the 9 sample sites to compute concentration maps across the lake for each precipitation event. These maps are then combined
with the lake area to estimate deposition for each event. Preliminary results from this study are similar to those reported by Miller. These studies focus on measuring AD from precipitation. While some dust or fines
were captured by the rain gages, these gages were not designed to capture or retain the dust or fines, so we
UTAH LAKE ATMOSPHERIC DEPOSITION
Page 7
expect these contributions are minimal. This study mostly measures precipitation (𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝) AD, though
some dust is captured.

The final study, reported by Barrus, et al. [2], uses buckets to capture all the settlement (𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡), contact
(𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡), and washout (𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝) AD [2,6]. This study uses an open bucket filled with deionized water to
capture contact and dust AD and a separate bucket to capture precipitation AD. However, for much of the
time, the mechanism that was used to switch the buckets did not work, so the total of all three AD processes was reported. This study collected 336 measurements in locations near Utah Lake. For these studies, both the
concentration and volume of water in the buckets were measured and converted to mass (mg) of TP per area per sample period (typically one week, though occasionally longer).

The first study [6], was provided to the Utah Lake Science panel and they had concerns about collection height, location of solar panels related to the buckets, potential splash from bucket lids, and insects in the
collections. In the subsequent study, [2], these issues were all addressed with side-by-side collections or other
approaches. While collection methods were changed, (buckets raised to 2 m, solar panels moved, and miner’s moss placed on bucket lids), statistically these were shown to be no different from the initial collections. The
addition of screens on the buckets to exclude insects did have an impact on some sites, but not others.
This subsequent study compared the data from all the collection sites and found that the data from any given
site, with the exception of Saratoga Springs, were not statistically different from any other site. This included the site at Bird Island. In other words, the stations were all measuring the same process. Since the
measurements at Bird Island, were not statistically different than shoreline measurements, this indicates that
there is little to no attenuation in AD rates. Barrus, et al. [2] estimated total TP loads of 262 and 133 tons/year for unscreened and screened samplers, respectively.
Supporting Study
An upcoming report from Telfer et al (2023 – not yet submitted), attempts to determine the source for dust in the samples around Utah Lake using samples from the on-going study reported by [2,6]. As part of this study, laboratory analysis requires at least 0.1 grams of solids. To acquire the dust samples for analysis, Telfer
filtered the samples that had been archived from the 5.5 month summer period from the 2022 sample year. These samples were not meant to measure dust, and not all the samples were retained, though the majority
were available. Telfer needed filter all the samples to obtain enough dust for analysis. Telfer found that over
the 5.5 months fall and spring periods the study, he collected the amounts of dust shown in Table 1.
Table 1 Dust collected over spring and fall 5.5-month periods.
Location Dust
Collection (grams)
Mosida (fall) 0.05
Mosida (spring) 0.07 Lake Shore (fall) 0.11 Lake Shore (spring) 0.11
Pump Station (fall) 0.03 Pump Station (spring) 0.16 Orem (fall) 0.05
Orem (spring) .0.07 Bird Island (total) 0.05
Bird Island (total - minus last sample) 0.02
This study was done during the high dust production period for the area.
UTAH LAKE ATMOSPHERIC DEPOSITION
Page 8
Bird Island results are interesting. There was so little dust in the samples, that they could not be split into spring and fall intervals, but the entire sample set was filtered to obtain 0.05 grams of dust. Also of interest,
the last sample contained 0.03 grams of that total, while the remaining 10 months contained only 0.02 grams. One large dust storm deposited more than half the total dust over the 11 month period on Bird Island. If the
last sample collected is ignored, the Bird Island sampler only collected one fifth the dust compared to Mosida and even less compared to the other samplers. This episodic nature of dust deposition is consistent with
Brahney [1] who notes that over 15 years of dust-on-snow measurements, some years had as few as 3 events
per year, while others have up to 12. While filtering the samples, Telfer noted that there were samples from Bird Island containing zero or near zero measurable dust for several of the weeks. Bird Islands proximity to West Mountain may protect it from some winds, creating the reduced dust deposition.
These Bird Island data indicates that there is attenuation in dust AD over Utah lake (large, settleable particles)
even though the nutrient AD data from [2] indicated no attenuation. To collect the dust, Telfer filtered the sample water from the buckets using a 0.45 µm filter. In the Telfer
study, each bucket has an area of 0.041 m2, so deposition rates for the minimum and maximum samples of 0.04 and 0.11 grams results over 5.5 month periods range from 0.98 g/m2 to 2.68 g/m2, respectively. These correspond to annual rates of 2.14 and 5.85 g/m2/yr, respectively.
This study and these samplers were not designed to capture dust. A few samples were missing, the samplers
were not designed to measure dust, and any soluble particles, such as NOx salts or organics, would have dissolved. Several of the samples collected for the study contained algae, which plugged the filters quickly and were therefore discarded. This also reduced the amount of dust collected. Brahney reported a Utah Urban
average of 24.7-56.7 g/m2/yr which is higher than these values, but a similar order of magnitude. We know that the majority of PM10 and PM2.5 are soluble nitrogen salts, so these values are reasonable. Discussion
These four separate studies measure different AD processes, settlement (𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡), contact (𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡),
washout (𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝) and (𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡). Brahney and Carling measure dust (𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡) with some fines, Miller
measures precipitation (𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝) with some dust, while Olsen and Barrus measure (𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡), or all three
processes.

Table 2 AD study summary
Study Process Amount (TP tons/yr) Notes Number of Samples
Brahney dust 2 – 9 Includes some minimal contact Varies Carling dust 55 Include some minimal contact 4 Miller precipitation 88 - 142 Includes some minimal dust and contact 850 Barrus contact, dust, precipitation 133 - 262 The lower value is screened 306
Using these studies, we can estimate contributions from the different processes. For example, if we assume
that 𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡 = 2 tons/yr, 𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝 = 88 tons, and 𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = 133 tons/yr, then 𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡 = 43 tons/yr.
Using these assumptions, then 𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡, 𝐴𝐴𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐𝑝𝑝𝑝𝑝, and 𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡 contribute 1.5%, 66%, and 32% of the AD,
respectively. If we assume that 𝐴𝐴𝐴𝐴𝑑𝑑𝑑𝑑𝑑𝑑𝑡𝑡 = 9 tons/yr, then 𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡 would be 36 tons/yr with percentages of 6.8%, 66%, and 27% for dust, precipitation, and contact processes, respectively.
This shows that these studies do not contradict each other, but rather support each other. The low percentage of AD from dust, shows that the attenuation demonstrated by the Bird Island dust samples is less than the
variance in the data, demonstrating why the total AD measurements at Bird Island are not statistically different from the measurements at the other sample sites.

UTAH LAKE ATMOSPHERIC DEPOSITION
Page 9
While 𝐴𝐴𝐴𝐴𝑐𝑐𝑡𝑡𝑐𝑐𝑡𝑡𝑡𝑡𝑐𝑐𝑡𝑡 estimates of 36 to 43 tons/yr are high, recall that Utah Valley suffers from high PM2.5 and PM10 levels, so contact deposition in this range is reasonable.
We would also like to stress that only the Miller, Olsen, and Barrus studies measured conditions near the lake,
both the Brahney and Carling studies measured deposition significant distances from the lake under different conditions. The Utah campuses (BYU, UofU, Weber, and USU) are all elevated and located in urban green areas with few large dry dust sources. In the summer you can see the haze in the valley from these campuses
and they are above a good part of it. In addition, Brahney lowers AD estimates based on estimated attenuation over the lake. For dust (large
gravity settling particles) this is correct and supported by the Bird Island data. But for smaller particles, attenuation is minimal and the data measured by both Miller and Barrus show no evidence of attenuation.
The attenuation of the dust AD is within the variance of the data. Conclusion
The Utah Lake AD studies initial appear to contradict each other because of the wide range of AD estimates.
However, when considering that AD is driven by different processes, contact, dust (settlement), and precipitation, and that each study mostly measured only a subset of the total, it is clear that the studies are not contradictory, but rather complement and strengthen each other.
Based on this analysis, we conclude that an annual AD TP loading of rate of 250 tons/yr to Utah Lake is
accurate. However, in consideration of the range of findings and potential implications, we propose 150 tons TP/yr could be used as a consensus-based value for evaluation.

UTAH LAKE ATMOSPHERIC DEPOSITION
Page 10
References
1. Brahney, J. Estimating total and bioavailable nutrient loading to Utah Lake from the atmosphere. 2019.
2. Barrus, S.M.; Williams, G.P.; Miller, A.W.; Borup, M.B.; Merritt, L.B.; Richards, D.C.; Miller, T.G. Nutrient Atmospheric Deposition on Utah Lake: A Comparison of Sampling and Analytical
Methods. Hydrology 2021, 8, 123. 3. Atkinson, R. Atmospheric chemistry of VOCs and NOx. Atmospheric Environment 2000, 34, 2063-2101, doi:https://doi.org/10.1016/S1352-2310(99)00460-4.
4. Hinds, W.C.; Zhu, Y. Aerosol technology: properties, behavior, and measurement of airborne particles; John Wiley & Sons: 2022. 5. Kuprov, R.; Eatough, D.J.; Cruickshank, T.; Olson, N.; Cropper, P.M.; Hansen, J.C. Composition
and secondary formation of fine particulate matter in the Salt Lake Valley: Winter 2009. Journal of the Air & Waste Management Association 2014, 64, 957-969, doi:10.1080/10962247.2014.903878.
6. Olsen, J.M.; Williams, G.P.; Miller, A.W.; Merritt, L. Measuring and calculating current atmospheric phosphorous and nitrogen loadings to utah lake using field samples and geostatistical analysis. Hydrology 2018, 5, 45.

UTAH LAKE ATMOSPHERIC DEPOSITION
Page 11

UTAH LAKE ATMOSPHERIC DEPOSITION
Page 12
Appendix
From Atkinson [3]

Table 3.1 Effect of Pressure on Terminal Settling Velocity of Standard Density
UTAH LAKE ATMOSPHERIC DEPOSITION
Page 13

2/18/2023
Comments by David Gay
Dr. G. Williams, “Atmospheric Deposition of Nutrients to Utah Lake: Process and Research
Overview” and Dr. Theron Miller email.

Complementary nature of the different samples: I agree with Dr. Williams on this conclusion of
complementariness. The different measurements are complementary to the total Atmospheric
Deposition (ADtotal). The studies mentioned all measured a part of the ADtotal equation, and not
the total deposition (except Barrus). This is driven in part by the difficulty, cost, assumptions
made, and disagreements of how to make dry deposition measurements. Dry deposition
measurements can be made, but are fraught with error, inconsistencies, and assumptions. As
Dr. Williams suggests, you need to compare like to like measurements, realizing that they are
all part of the total deposition equation. Additionally as he suggests, a bulk measurement of wet
deposition (without a closing top), will generally accurately measure wet deposition, but the
concentrations are likely biased high by added dry deposition (ADdust, ADcontact).
I will go one step further and suggest, based on the above, that the wet deposition collections
made with closing samplers are likely the best measurements (true ADprecip) as compared to the
dry deposition measurements (ADdust, ADcontact) because they are easier measurements with fewer
assumptions. Therefore I maintain that true ADprecip measurements can provide the basic
deposition or minimum deposition to the lake, and that the dry deposition should be added to
these measurements with the associated inaccuracies, errors and assumptions.
Going further out on a limb and following from the idea that this paper focuses on TP, I would
argue that in Equation 1:
𝐴𝐷௧௢௧௔௟ =𝐴𝐷ௗ௨௦௧ +𝐴𝐷௖௢௡௧௔௖௧ +𝐴𝐷௣௥௘௖௜௣
The ADdust and ADprecip are the most important components for the total lake deposition.
Phosphorus does not have gaseous compounds, therefore gaseous particle formation of small
particles is less likely to occur, so PM2.5 and smaller sized particles with phosphorus compounds
should be less important. Therefore, the more important category of TP deposition would
primarily be ADdust, and ADprecip.

Another Calculation
Table 2 AD study summary
Study Process Amount
(TP tons/yr)
DG Notes
Brahney dust 2 – 9 Dry only, literature study only
Carling dust 55 Dry only, few measurements
Miller precipitation 88 - 142 Wet plus some dry contam.
Barrus contact, dust, precipitation 133 - 262 Wet plus dry measurement, but with
problems

This is Dr. Williams table, but with the deposition description added in the final column. I don’t
know (or remember) the conditions surrounding the Barrus measurement problems mentioned
by Dr. Williams. However, let’s just assume that they are close to correct. And, if you add the
Carling dry P measurements to the Miller wet P measurements, you get about the same range
as the Barrus total P measurements. I think Brahney estimates are pretty low versus everything
else I have seen. Carling measurements are only a few number and have the normal dry
deposition measurement problems. Miller’s wet deposition measurements have some problems
too (not closed after precip, not a dry deposition sampler, etc.), but there are a lot of
measurements. I have mentioned I don’t know what Dr. Williams means with the problems for
Barrus. But this summation of Carling + Miller estimates a range of about 140 to 200 for total P
deposition. That is about the same range as Barrus. So with three independent studies, you
arrive at about the same answer. This all suggests the 150 to 250 range noted by Dr. Williams is
reasonable.

Another Calculation
Here is another way to get to the reasonableness of all of this.
The US EPA estimates total (wet and dry) deposition of N through modeling. See this location
for the maps: https://nadp.slh.wisc.edu/committees/tdep/ with individual map links below at
this site. Unfortunately they have not tried to model TP yet. However, my idea is this.
I calculated (see excel sheets) what total N deposition was for Utah Lake based on these
modeling estimates for 2021. I get the following:

These are big numbers. My idea is to do the following.
 Compare the 384 tons of deposition as N (or 1747 tons as NO3) versus the Utah Board’s
preferred number for atmospheric N deposition. Does it compare well?
 If EPA> Utah Board, this is another qualified estimate of N deposition to Utah Lake that
says the Utah Board values are too low.
 If you have data that says the mass ratio of N to P in your samples is X, then you can
estimate TP deposition from the EPA TN deposition values.
 Then if this estimate of TP is much greater than the Utah Board estimate, then you have
a logical estimate to argue that the TP values they want to use are also probably low.
 For example, if wet and dry deposition measurements say that there is 1 atom/molecule
of P for every 10 atoms/molecules of N, then this would argue that the total deposition
of P to the lake is 384 tons N * 0.10 = 38.4 ton P.
 This is something of a stretch, but it is somewhat reasonable.

Based on my calculations and other observations, it seems that a 40-50% estimate of 50 TP: 100N
is about right. Brahney’s values are very, very low and are only for dry deposition. The same is
true for Carling. Miller is only for wet deposition plus some dry. Barrus’s are actually for total
deposition, and range from 0.43 to 1.70 P:N. If you just assume the low end of this (40-50%),
then you get an estimate of total deposition to the lake.
If you multiply this by the EPA estimate of Total N deposition (wet and dry), you come in at
about 150 to 195 tons P per year to the Lake (both wet and modeled dry deposition estimates).
This makes Dr. William’s estimate of 150 or 250 about right; it strikes me as at least reasonable,
based on others data.
Again, there might be better estimates of the ratio of P to N, which could be used.

Other Matters
Also, I have noted before, that insects into the lake may not officially be considered “wet
deposition” by anyone. We don’t with NADP. But for the TN and TP load, it would seem to me
that this could be a significant source of both to the lake, and should be considered in the TN/TP
load estimates and calculations.

Opinion on the “stickyness” of a water surface. Basically, I agree with your opinion, Theron.
Almost any particle coming in contact with a free water surface is essentially going to get stuck
to the water through chemical charge interaction, or if it is heavy enough it will sink into the
water. And, it seems to me that no particle is going to leave the fluid on its own. A lake, with
waves and wind is going to generate water aerosols certainly that could carry suspended
particles with them back into the atmosphere. But the idea that the lake is a somewhat passive
collectors of atmospheric particles is correct. Any particle with a significant deposition velocity
and quiet atmospheric conditions will pick up particles from the atmosphere, whether they are
anthropogenic or naturally occurring. Therefore, this should be occurring with Provo/Salt Lake
urban smog into Utah Lake. Larger particles of 10 microns and above will settle out fairly
quickly, and as dust plumes move over the lake, I would certainly expect the larger dust
particles to settle into the water and the dust cloud would become less and less concentrated as
it moves across the lake. I would expect significant attenuation of large particles by the time you
reach Bird Island and the east side of the lake.
How significant is this? That is a very good question. It would depend upon the atmospheric
concentrations, the size of the particles (likely very small, fine fraction), and the atmospheric
conditions. Higher concentrations, larger particles, and more quiet atmospheric conditions will
increase this deposition, and vice versa.
This phenomena is essentially dry deposition but to a liquid surface.

I do not understand why Dr. Williams concludes “we recommend a annual TP loading
of 150 tons/yr (rounded), though we feel that a rate of 250 tons/yr is more accurate.”. I
would think he would recommend 250 tons TP to the lake.

Additionally, I have included comments to the document, as you will see below. Some of these
you might find useful. Most of what Dr. Williams says in his report I agree with, as you will see
in my comments. However, it comes down to who’s measurements or calculations for dry
deposition that you believe. It is very, very difficult to measure, and is always controversial.

D. A. Gay

OreoHelix Ecological “Dedicated to Evaluating and Protecting the World’s Ecological Health, Integrity, and Well Being…. One
Snail at a Time”

Filename: Effects of screens on AD of TP in samplers

November 3, 2022

Screen Effects on TP in AD samplers
Technical Memo

To
Wasatch Front Water Quality Council
Salt Lake City, UT

By
David C. Richards, Ph.D.
OreoHelix Ecological, Vineyard UT 84059
Phone: [phone]
Email: [e-mail]

Justification
There is much concern by DWQ Utah Lake Science Panel (ULSP) on the amount of nutrients
accumulating on Utah Lake from atmospheric deposition (AD). Presently, the ULSP is
considering using only screened sampler data from Barrus et al. (2020) raw data after removal
of insect or debris contaminated samples to calculate AD loads. However, Barrus et al. (2021)
and Richards (2020) reported that screened samplers significantly reduced TP deposition.
Accurate estimates of AD of nutrients will not be possible if the effects of screens on AD are not
accounted for. This cursory analysis addresses this concern.
Methods
Raw data from Seth Barrus Excel file titled, “AD_Results_Barrus”, sheet name: “CombinedStats
per m2” were analyzed. Table II on that sheet provided 48 sampler data from Central Davis High
and Orem paired screened and unscreened (NADP) samplers (Table 1).

Table 1. II. Comparison between NADP and SDSD Sample Tables (No filter - NADP, Filter - SDSD): Total Phosphorus (mg/m2)
from Barrus 2020 Excel spreadsheet.

Date Central Davis High Central Davis NADP Orem Orem NADP
6/25/20 1.9736 4.4820 2.0095 22.9642
7/2/20 2.6770 3.0465 2.6660 7.0593

OreoHelix Ecological “Dedicated to Evaluating and Protecting the World’s Ecological Health, Integrity, and Well Being…. One
Snail at a Time”

7/10/20 0.8675 3.3352 0.6988 9.4431
7/17/20 N/A N/A 2.4678 5.4925
7/23/20 2.7951 N/A 4.9859 4.5119
7/30/20 N/A N/A 0.9500 23.7307
8/10/20 2.4922 6.2892 1.0434 66.5338
8/21/20 5.0232 7.9270 2.3631 232.5352
8/28/20 42.2723 41.6804 3.5383 49.7937
9/4/20 N/A N/A 2.4220 5.2501
9/11/20 N/A 5.6627 59.9064 83.9371
9/18/20 19.0458 4.0492 1.2663 4.2267
9/25/20 2.1758 3.3842 2.1909 4.0811
10/2/20 2.9420 4.3022 1.6954 8.2823
10/9/20 1.0739 3.6930 2.1695 3.4091
10/15/20 4.0970 4.0859 2.5183 3.6299
10/23/20 1.7816 11.1853 1.2217 3.7653
10/29/20 1.5314 4.7025 34.4981 1.6064
11/12/20 N/A N/A 23.8514 42.5714
11/19/20 N/A N/A 3.1592 5.3603
11/25/20 7.1850 15.9369 1.5063 10.5526
12/3/20 6.9812 1.1643 2.3049 8.6842
12/10/20 0.9690 0.6500 1.5589 2.5119
12/16/20 1.2232 1.8235 1.2048 4.3610

The following table (Table 2) is reordered with Bug/Debris added from Barrus 2020 sheet name:
“Overall”.

Table 2. Reordered Table 1 with bug/debris samples added from sheet “Overall” Barrus spreadsheet.

Date Screened Unscreened Location Bugs/Debris
8/21/20 5.0232 7.9270 Central Davis High 3
7
13
50
debris
y
y
y debris
10/23/20 1.7816 11.1853 Central Davis High
10/23/20 1.2217 3.7653 Orem
8/21/20 2.3631 232.5352 Orem
11/12/20 23.8514 42.5714 Orem
10/15/20 4.0970 4.0859 Central Davis High
10/15/20 2.5183 3.6299 Orem
10/29/20 1.5314 4.7025 Central Davis High
6/25/20 1.9736 4.4820 Central Davis High
7/2/20 2.6770 3.0465 Central Davis High
7/10/20 0.8675 3.3352 Central Davis High

OreoHelix Ecological “Dedicated to Evaluating and Protecting the World’s Ecological Health, Integrity, and Well Being…. One
Snail at a Time”

7/17/20 N/A N/A Central Davis High
7/23/20 2.7951 N/A Central Davis High
7/30/20 N/A N/A Central Davis High
8/10/20 2.4922 6.2892 Central Davis High
8/28/20 15.9369 41.6804 Central Davis High
9/4/20 N/A N/A Central Davis High
9/11/20 N/A 5.6627 Central Davis High
9/18/20 19.0458 4.0492 Central Davis High
9/25/20 2.1758 3.3842 Central Davis High
10/2/20 2.9420 4.3022 Central Davis High
10/9/20 1.0739 3.6930 Central Davis High
11/12/20 N/A N/A Central Davis High
11/19/20 N/A N/A Central Davis High
11/25/20 7.1850 15.9369 Central Davis High
12/3/20 6.9812 1.1643 Central Davis High
12/10/20 4.5119 0.6500 Central Davis High
12/16/20 23.7307 1.8235 Central Davis High
6/25/20 2.0095 22.9642 Orem
7/2/20 2.6660 7.0593 Orem
7/10/20 0.6988 9.4431 Orem
7/17/20 2.4678 5.4925 Orem
7/23/20 4.9859 4.5119 Orem
7/30/20 0.9500 23.7307 Orem
8/10/20 1.0434 66.5338 Orem
8/28/20 3.5383 49.7937 Orem
9/4/20 2.4220 5.2501 Orem
9/11/20 59.9064 83.9371 Orem
9/18/20 1.2663 4.2267 Orem
9/25/20 2.1909 4.0811 Orem
10/2/20 1.6954 8.2823 Orem
10/9/20 2.1695 3.4091 Orem
10/29/20 34.4981 1.6064 Orem
11/19/20 3.1592 5.3603 Orem
11/25/20 1.5063 10.5526 Orem
12/3/20 2.3049 8.6842 Orem

OreoHelix Ecological “Dedicated to Evaluating and Protecting the World’s Ecological Health, Integrity, and Well Being…. One
Snail at a Time”

12/10/20 1.5589 2.5119 Orem
12/16/20 1.2048 4.3610 Orem

There were eight bug/debris ‘contaminated’ samples that I removed from further analysis.
I then calculated Difference = unscreened – screened and descriptive statistics.
Results
The difference between paired screened and unscreened data was calculated (Table 3).
Table 3. Difference between screened and unscreened TP concentrations mg/m2.

date screened unscreened site Difference
6/25/20 1.9736 4.482 Central Davis High 2.51
7/2/20 2.677 3.0465 Central Davis High 0.37
7/10/20 0.8675 3.3352 Central Davis High 2.47
7/17/20 0.05 0.05 Central Davis High 0.00
7/23/20 2.7951 0.05 Central Davis High -2.75
7/30/20 0.05 0.05 Central Davis High 0.00
8/10/20 2.4922 6.2892 Central Davis High 3.80
8/28/20 4.9859 41.6804 Central Davis High 36.69
9/4/20 0.05 0.05 Central Davis High 0.00
9/11/20 0.05 5.6627 Central Davis High 5.61
9/18/20 19.0458 4.0492 Central Davis High -15.00
9/25/20 2.1758 3.3842 Central Davis High 1.21
10/2/20 2.942 4.3022 Central Davis High 1.36
10/9/20 1.0739 3.693 Central Davis High 2.62
11/12/20 0.05 0.05 Central Davis High 0.00
11/19/20 0.05 0.05 Central Davis High 0.00
11/25/20 7.185 15.9369 Central Davis High 8.75
12/3/20 6.9812 1.1643 Central Davis High -5.82
12/10/20 3.5383 0.65 Central Davis High -2.89
12/16/20 0.05 1.8235 Central Davis High 1.77
6/25/20 2.0095 22.9642 Orem 20.95
7/2/20 2.666 7.0593 Orem 4.39
7/10/20 0.6988 9.4431 Orem 8.74
7/17/20 2.4678 5.4925 Orem 3.02
7/23/20 4.9859 4.5119 Orem -0.47
7/30/20 0.95 23.7307 Orem 22.78

OreoHelix Ecological “Dedicated to Evaluating and Protecting the World’s Ecological Health, Integrity, and Well Being…. One
Snail at a Time”

8/10/20 1.0434 66.5338 Orem 65.49
8/28/20 3.5383 49.7937 Orem 46.26
9/4/20 2.422 5.2501 Orem 2.83
9/11/20 59.9064 83.9371 Orem 24.03
9/18/20 1.2663 4.2267 Orem 2.96
9/25/20 2.1909 4.0811 Orem 1.89
10/2/20 1.6954 8.2823 Orem 6.59
10/9/20 2.1695 3.4091 Orem 1.24
10/29/20 34.4981 1.6064 Orem -32.89
11/19/20 3.1592 5.3603 Orem 2.20
11/25/20 1.5063 10.5526 Orem 9.05
12/3/20 2.3049 8.6842 Orem 6.38
12/10/20 1.5589 2.5119 Orem 0.95
12/16/20 1.2048 4.361 Orem 3.16

The mean difference in TP (mg/m2) between screened and unscreened side by side paired
samples was 6.02 mg/m2 and the proportion difference (mean unscreened/mean screened)
was 2.26 mg/m2 from samples with bugs/debris removed (Table 4). This shows that screens
had a very large effect on reducing the amount of AD that went into a sampler. Reasons are
speculative, for example screens accumulated AD, wind blew AD off screens, etc.

Table 4. Descriptive statistics of screened and unscreened TP mg/m2/week.

stats screened unscre~d

mean 4.78 10.8
sd 10.7 18.3
p50 2.17 4.33
p25 .997 2.17
p75 3.05 8.48
max 59.9 83.9
min .05 .05
range 59.9 83.9

stats Difference
mean 6.01
sd 15.8
p50 2.49
p25 0
p75 6.48

OreoHelix Ecological “Dedicated to Evaluating and Protecting the World’s Ecological Health, Integrity, and Well Being…. One
Snail at a Time”

max 65.5
min -32.9
range 98.4

Conclusion
AD samplers with screens had a very significant negative effect on TP deposition measurements
and can significantly bias AD nutrient load estimation on Utah Lake.
Recommendation
Do not use screened data only, because screens reduced TP by about 56%, which is consistent
with Barrus et al. 2021 publication and my initial analyses, Richards 2020. I recommend using
both screened and unscreened data after adjusting screened data to account for screen effect
and after removing contaminated samples to estimate nutrient loads more accurately to Utah
Lake from AD.
Literature Cited
Barrus, S. M. et al. 2021. Nutrient Atmospheric Deposition Sampling and Analysis
Improvements: Utah Lake Impacts. Hydrology.

Richards, D.C. 2020. Nutrient Atmospheric Deposition on Utah Lake and Great Salt Lake
Locations 2020, including Effects of Sampler Type: Statistical Analyses and Results.
Report to Wasatch Front Water Quality Council, Salt Lake City. OreoHelix Ecological,
Vineyard, UT.