Loading...
HomeMy WebLinkAboutDWQ-2024-004577 1 Utah Lake Water Quality Study (ULWQS) Science Panel June 30, 8:00 AM to 3:00 PM State of Utah Multi-Agency Office Building, Salt Lake City Meeting Summary ATTENDANCE: Science Panel Members: Zach Aanderud, Mike Brett, Greg Carling, Mitch Hogsett, Theron Miller, Hans Paerl, Thad Scott, and Tim Wool Steering Committee Members and Alternates: Chris Cline, Eric Ellis, and Jay Olsen Members of the Public: Paul Abate, Jeff DenBleyker, Dan Potts Utah Division of Water Quality (DWQ) staff: Scott Daly, Jeff Ostermiller, and Nicholas von Stackelberg Technical Consultants: Jon Butcher, Rene Camacho, Mark Fernandez, Maddie Keefer, and Kevin Kratt Facilitation Team: Heather Bergman and Samuel Wallace ACTION ITEMS Who Action Item Due Date Date Completed Tetra Tech Produce a time-step graph using the Utah Lake nutrient model outputs for organic and inorganic matter fluxes. Nov. 30 Assess whether different taxa influence the translator between beaches/marinas and open water sites. Nov. 30 Work with DWQ Standards and Assessment staff to develop a recommendation for the prediction interval to discuss at a further meeting. Nov. 30 Create separate S-R analysis graphs for Provo Bay and main basin results. Nov. 30 Separate Provo Bay from the main basin on the overlay of chlorophyll-TP and chlorophyll-TN data from Utah Lake and the National Lakes Assessment. Nov. 30 Model Subgroup Meet to better understand the functionality of the Utah Lake nutrient model and discuss the next steps for applying it to the S-R analysis. Nov. 30 DWQ Produce a comprehensive list of water quality standards for the Science Panel. Nov. 30 DECISIONS AND APPROVALS No formal decisions or approvals were made at this meeting. 2 SCIENCE PANEL DIRECTION The Science Panel provided the following direction for the stressor-response (S-R) analysis: • Combine Goshen Bay and the main basin as one region for the S-R analysis. • Maintain the translation that has already been developed between beaches/marinas and open water sites and explore further whether different taxa influence the relationship between beaches/marinas and open water sites. • Have Tetra Tech work with DWQ Standards and Assessment staff to develop a recommendation for the prediction interval to discuss at a further meeting. • Have Tetra Tech develop different S-R graphs for main basin and Provo Bay data. • Do not use the relationship between Secchi depth and chlorophyll, TN, and TP to set targets for numeric nutrient criteria (NNC). • Use the National Lakes Assessment (NLA) dataset for chlorophyll-total nitrogen (TN) and chlorophyll-total phosphorus (TP) as a line of evidence in the Technical Support Document (TSD). WATERSHED MODEL UPDATE Maddie Keefer, Tetra Tech, presented an update on the Utah Lake watershed model. The presentation, the subsequent Science Panel discussion, and public comments are summarized below. ULWQS Watershed Model Update Presentation Below is a summary of the ULWQS watershed model update presentation. ULWQS Watershed Model Overview • The process for developing the in-lake water quality model takes the following steps: o Gather data o Build the model o Calibrate the model o Assess current conditions o Run scenarios • Tetra Tech follows the watershed model Quality Assurance Project Plan (QAPP). The QAPP o Outlines the quality objectives for measured and modeled data o Provides the model framework to support project goals and objectives o Lays out the steps for data collection and acquisition to support the model build and calibration o Identifies the quality assurance/quality control activities to assess model performance o Specifies the methods for a model usability assessment • Tetra Tech has completed the hydrology and sediment calibrations. They are now working on the nutrient calibration. • For the hydrology calibration, the Tetra Tech modeling team calibrated the model to multiple endpoints, including remotely sensed snow depth and water storage; actual evapotranspiration; and daily, monthly, and cumulative gage flows. They then generated multiple visuals and calculated statistical metrics related to total flow, seasonal/monthly flows, high/low flow distribution, and Nash-Sutcliffe efficiency (NSE) coefficients. In the calibration, they prioritized getting a better overall fit with the larger tributaries to Utah Lake (Spanish Fork and Provo River). 3 ULWQS Watershed Model Sediment Calibration Overview • For the sediment calibration, the Tetra Tech modeling team recognized that observed sediment concentrations have various sources of uncertainty. Due to these uncertainties, they have calibrated the model to multiple endpoints, including daily, monthly, and cumulative monitored sediment concentration; sediment source assessment; and reach/sediment balance. The objective of the model is to represent the overall sediment behavior of the watershed, with knowledge of the morphological characteristics of the stream (i.e., aggrading or degrading behavior). To achieve this objective, Tetra Tech will use sediment loading rates consistent with available values to provide a reasonable match with instream sediment data. • Like the hydrologic calibration, the sediment calibration is guided by multiple visuals and statistical metrics related to sediment loading rates from the landscape, seasonal/monthly total suspended sediment (TSS) concentrations, high/low flow distribution, and average and median concentration and load error. The observed TSS records are obtained from Utah stream monitoring sites. The calibration seeks to obtain the best overall fit at multiple locations, prioritizing the larger tributaries to Utah Lake (Spanish Fork and Provo River). • Tetra Tech calculated the sediment calibration using TSS data from seven sites: Provo River at Murdock and Olmstead, Hobble Creek, Spanish Fork at Utah Lake, Spanish Fork at Moark, Diamond Fork, Thistle Creek, and Currant Creek. Starvation Creek/Upper Solider Creek, the eighth potential calibration site, did not have enough TSS data for calibration purposes. • The watershed model uses an average annual sediment loading rate for each land use category in the watershed. The average annual sediment loading rate is derived from the 2016 National Land Cover Database (NLCD) and is calculated based on data from 2008-2016. Due to the impact of significant wildfires in 2017, the Tetra Tech modeling team used a fire-modified average annual sediment loading rate to capture the impact of wildfires on sediment loading. • The sediment calibration compared observed data to simulated results for the calibration period (2015-2021). The calibration results indicated a good fit between simulated and observed data based on model acceptance criteria provided in the literature for the Hydrologic Simulation Program – FORTRAN (HSPF) model, the modeling program selected for the watershed model. • To visualize the comparison, the Tetra Tech modeling team plotted the simulated results versus observed data. ULWQS Watershed Model Nutrient Calibration Overview • The Tetra Tech modeling team deliberately sequenced the modeling calibration, beginning with the hydrology calibration, then the sediment calibration, and ending with the nutrient calibration. • The steps in the nutrient calibration include: 1. Estimating all model parameters, including land use-specific accumulation and depletion/removal rates, washoff rates, and subsurface concentrations 2. Comparing simulated nonpoint loading rates with the expected range of nonpoint loadings from each land use and adjusting loading parameters when necessary to improve agreement and consistency 3. Calibrating instream water temperature 4. Comparing simulated and observed instream concentrations at each of the calibration stations 4 5. Analyzing the results of the comparisons in steps 3 and 4 and making adjustments until calibration targets are achieved • The main objective of the nutrient calibration is to obtain acceptable agreement of observed and simulated concentrations while maintaining the instream water quality parameters within physically realistic bounds and the nonpoint loading rates within the expected ranges from literature. • The first step in nutrient calibration is the water temperature calibration. Instream temperature is an important parameter for simulating biochemical transformations. The HPSF modules representing water temperature include PSTEMP (soil and groundwater temperature) and HTRCH (heat exchange and water temperature within flowing reaches). The modeling team plotted observed water temperature data versus the simulated results. The calibration indicates very good agreement between observed and simulated results for Provo River at Murdock and Olmstead, Hobble Creek, Spanish Fork at Utah Lake, and Diamond Fork. • The Tetra Tech modeling team plotted observed nitrate concentrations against simulated values. It is important to use visual tools to assess whether simulated values follow seasonal trends. • The nutrient simulation for the uplands represents ammonia, nitrate, inorganic phosphorus, and organic matter as four distinct constituents. Inorganic phosphorus and organic matter on pervious surfaces are simulated using a sediment potency approach. Ammonia and nitrate on pervious surfaces and all four constituents on impervious surfaces are represented as a buildup/washoff process. The Tetra Tech modeling team created graphs to show the average annual loading rate for nitrogen and phosphorus based on land use type (e.g., impervious roads, impervious buildings, grassland/shrub, etc.). • The HSPF model simulates point and nonpoint sources of key stressors and provides the output necessary to develop source assessments. The modeling team conducted a nitrogen and phosphorus source assessment for the Utah Lake HSPF model area using the current version of the model. Results may change pending final calibration results. The results show the annual nutrient loads for total nitrogen (TN) and total phosphorus (TP) by category (barren/burned, grassland/shrub/pasture/hay, forest, water/wetland (with direct atmospheric deposition), agriculture, developing roads, and point sources). The results for point sources do not include the discharges from the Timpanogos Special Services District (TSSD) or Orem Wastewater Treatment Plant (WWTP) because these WWTPs discharge directly into Utah Lake. ULWQS Watershed Model Next Steps • The sediment calibration is completed. The Tetra Tech modeling team has also begun the water quality calibration for dissolved oxygen and temperature. The draft water quality calibration and model documentation report was submitted to DWQ and presented to the Science Panel at this meeting. • Over the next few months, the Tetra Tech modeling team will address comments from DWQ to improve the calibration. They plan to circulate the watershed model report to the Science Panel for review later this year. Science Panel Clarifying Questions Science Panel members asked clarifying questions about the watershed model update. Their questions are indicated in italics below, with the corresponding responses in plain text. 5 As a land use type, orchards appear to have a high average annual phosphorus and nitrogen loading rate compared to the other land uses. Is that loading rate based on general assumptions, or is that rate particular to the Utah Lake catchment? The nitrogen and phosphorus loading rates for orchards are significantly higher than other modeled land uses. Orchard operations are nutrient-intensive, so it is unsurprising to see high nutrient-loading rates. The Tetra Tech modeling team does not have specific nutrient loading rate data from orchards in the Utah Lake watershed. Instead, they used a sediment potency rate from previous HSPF models developed in Lower California for the Lake Hodges Watershed and Santa Maria Watershed. Does "organic matter" mean total organic matter or dissolved particulate matter? The Tetra Tech split organic matter into labile and refractory organic carbon, organic nitrogen, and organic phosphorus. "Organic matter" refers to total organic matter. Where did the atmospheric deposition values come from for the nutrient calibration? • The atmospheric deposition values for nitrogen and phosphorus came from the Science Panel's recent work to determine atmospheric deposition loading rates for phosphorus and nitrogen. The Science Panel recommended using the Community Multiscale Air Quality (CMAQ) model from Brahney (2019) to estimate nitrogen loading from atmospheric deposition to Utah Lake. The watershed modeling team used the same CMAQ model to estimate nitrogen loading from atmospheric deposition for the catchment. • The modeling team took the Science Panel's estimate of phosphorus loading from atmospheric deposition for Utah Lake specifically (32 tons/year) and scaled the value up to cover the whole catchment. Public Clarifying Questions Members of the public asked clarifying questions about the watershed model update. Their questions are indicated in italics below, with the corresponding responses in plain text. What is the coverage of orchards in the Utah Lake watershed? Are there plans to collect locally specific data to refine the loading rate for orchards? The coverage of orchards is very small; Orchards represent less than 0.5% of the greater watershed area. Local field monitoring and studies on orchards would be helpful to refine the loading rate. That being said, given that orchards represent less than 0.5% of the land use cover in the watershed, changes in the loading rate will most likely not make a significant difference. UTAH LAKE NUTRIENT MODEL SENSITIVITY AND UNCERTAINTY ANALYSES PRESENTATION Rene Camacho, Tetra Tech, presented an update on the Utah Lake nutrient model sensitivity and uncertainty analyses. The presentation, the subsequent Science Panel discussion, and public comments are summarized below. Utah Lake Nutrient Model Sensitivity and Uncertainty Analyses Presentation Below is a summary of the Utah Lake nutrient model sensitivity and uncertainty analyses presentation. • The sensitivity analysis evaluates how changes in model input affect model outputs. The analysis involves changing model inputs to see how the model responds. The sensitivity analysis identifies the most important physical or biochemical drivers of ecosystem response. It can also identify the most sensitive parameters to focus calibration efforts. • The uncertainty analysis is more complex. The uncertainty analysis estimates dispersion (variance) around calibrated model outputs. The goal is to minimize errors between 6 simulated and observed data. During calibration, the modeling team minimizes model fit errors. During the uncertainty analysis, the modeling team estimates calibration variance. The uncertainty analysis results include several statistical values that quantify error and bias; these statistical values indicate how accurate the model is and how confident one should be in its outputs. • Determining model uncertainty is very complex, as some sources of certainty are unquantifiable. The uncertainty analysis focuses on epistemic errors, which are errors related to the model structure, input data, and parameters. These sources of errors are fixable and reducible by creating a better mechanistic representation of a particular process. Some sources of uncertainty are not reducible; these uncertainties are ontological. An example of an ontological error is the ecosystem response to stochastic natural processes/episodes. For example, fires and floods can completely change how the ecosystem responds to different processes, which are not accounted for in the models. • Assumptions are necessary to allow for the quantification of uncertainty. Uncertainty bounds depend on the statistical distribution of model errors. In most cases, model errors are assumed to be independent, homoscedastic, and normally distributed. • Two methods used to calculate model uncertainty include stochastic unconstrained and stochastic constrained. The stochastic methods are state-of-the-art but computationally intensive. These methods require running thousands of model simulations to sample the range of model inputs and parameters. Because they are computationally intensive, these methods are often limited to lumped and/or empirical and conceptually simple models. • Another method for calculating model uncertainty is the deterministic method. The deterministic method includes conducting a first-order variance analysis (FOVA). A FOVA is computationally inexpensive and applicable to mechanistic models. The FOVA uses a Taylor series expansion to generate uncertainty bounds and a sensitivity coefficient. The uncertainty bounds are a calculation of the standard deviation for a given variable. The sensitivity coefficient measures how much one variable is expected to change under different input parameters; the sensitivity coefficient identifies which parameters are the drivers of model outputs. • Ultimately, the FOVA produces similar results to the stochastic-constrained methods. Since the FOVA produces similar results and is less computationally intensive, it is the preferred method for the uncertainty analysis. • Ultimately, the Utah Lake nutrient model will be used as one line of evidence to estimate key assessment endpoints (e.g., chlorophyll-a, TN, TP, pH, dissolved oxygen) under various conditions. The sensitivity analysis will help identify the most important physical or biochemical drivers of ecosystem response. The uncertainty analysis will help estimate dispersion (variance) around calibrated model outputs. The sensitivity and uncertainty analysis will help the Science Panel have a more informed discussion on how to use the model in decision-making. Science Panel Clarifying Questions Science Panel members asked clarifying questions about the Utah Lake nutrient model sensitivity and uncertainty analyses. Their questions are indicated in italics below, with the corresponding responses in plain text. Will the sensitivity and uncertainty analyses only be conducted for the Utah Lake nutrient model? Yes. 7 MECHANISTIC MODEL STRESSOR-RESPONSE ANALYSIS APPROACH PRESENTATION Kevin Kratt, Tetra Tech, presented the approach for using the mechanistic model in the stressor-response (S-R) analysis. The presentation, the subsequent Science Panel discussion, and public comments are summarized below. Mechanistic Model S-R Analysis Approach Presentation Below is a summary of the mechanistic model S-R analysis approach presentation. Role of Mechanistic Model in Numeric Nutrient Criteria (NNC) Setting Overview • Three types of evidence will be used to generate NNC: reference-based, S-R analysis, and scientific literature. o The reference-based lines of evidence include the results from the paleolimnological studies. It will also involve using the Utah Lake nutrient model to predict and extrapolate reference conditions (i.e., the conditions of the lake pre-European settlement). o The S-R analysis lines of evidence will rely on outputs from the Utah Lake nutrient model and other statistical models. o The scientific literature lines of evidence will incorporate the scientific studies of comparable/related lake ecosystems. • There are pros and cons to using mechanistic modeling to set NNC. o Advantages include:  Mechanistic models are based on deterministic relationships (i.e., causation, not correlation). The model involves inputting initial data parameters and then using equations to relate those inputs to water quality outputs.  Mechanistic models produce outputs that cover the model's full spatial and temporal extent. Unlike grab samples or installed sensors, which only capture conditions in one place at one moment, the model can provide outputs at every grid cell at every hour.  Mechanistic models provide information for a significant number of parameters and fate/transport issues. These parameters include hydrodynamic parameters, total suspended solids, all nutrient species, chlorophyll-a, dissolved oxygen, pH, algal growth, settling, sediment resuspension, and sediment diagenesis. o Disadvantages include:  There is uncertainty in the model; no models are perfect. The sensitivity and uncertainty analyses will help characterize the degree of uncertainty associated with the model.  The mechanistic model is a simplistic simulation of some key processes (e.g., food web dynamics).  The outputs from the model do not directly compare to the sampling data typically used to make assessments. Because water quality impairment decisions are made on sampling data, the protocols on whether a water body is impaired are oriented around grab samples at certain locations.  A mechanistic model cannot predict several assessment endpoints, including cyanobacterial abundance and microcystin concentrations. However, it will useful for assessing several other S-R response relationships. Mechanistic Model Scenario Overview • The mechanistic model can be used to evaluate nutrient load alternative scenarios. For example, in other processes, the Tetra Tech modeling team has assessed TN, TP, and 8 chlorophyll-a based on three scenarios: existing conditions, no anthropogenic loads, and a 50% reduction in existing nutrient loads. The Science Panel will need to consider how many scenarios to run, how to set up those scenarios (i.e., how to characterize pre-European settlement scenarios), and how to process the outputs. • The Science Panel has previously discussed running the model under existing conditions, reference conditions, and reduced loading conditions. Regarding the reduced loading scenarios, the Science Panel will need to consider how many of those scenarios to run and to what degree the loads should be reduced. • Additionally, the Science Panel will need to determine how to decrease nutrient loads. For example, the Science Panel could choose to decrease particulate and dissolved forms of nutrients equally. • Once the Science Panel has determined scenarios, they will need to consider how to process outputs spatially and temporally. Spatial considerations include whether to separate the outputs of Provo Bay from the rest of the lake and whether to use the average of the vertical layers or the surface layers only. Temporal considerations include whether to use all six years of model simulation and whether to use an entire year or only focus on the growing season. Science Panel Clarifying Questions Science Panel members asked clarifying questions about the mechanistic model S-R analysis approach. Their questions are indicated in italics below, with the corresponding responses in plain text. How can the model be used to assess the role calcite might have in removing phosphorus from Utah Lake? The simulation of calcite formation in the model is not represented mechanistically. The model takes a simplified approach, using a partitioning coefficient to represent the phosphorus-calcite binding. The partitioning coefficient approach estimates the net effect of the calcite present on the deposition of particulate inorganic phosphorus. The partitioning coefficients are based on the Phosphorus-Binding (P-Binding) Study by Dr. Josh LeMonte at Brigham Young University. Does the partitioning coefficient include the iron fraction? • The partitioning coefficient is global. The coefficient does not relate to pH or temperature; it only represents the net deposition of inorganic particulate phosphorus. • The model cannot mechanistically represent iron-phosphorus formation. How does the sediment diagenesis module treat phosphorus-iron? The sediment diagenesis module has a representation of iron, but it is not a mass balance term in the model. How does the model incorporate sediment and water column interaction? The model dynamically represents the sediment and water column interaction. The Utah Lake nutrient model contains the mechanistic sediment diagenesis module, which simulates the breakdown of organic matter in the sediment. The model then simulates the inorganic and organic matter fluxes to and from the water column. Have the simulated fluxes been compared to the result of completed studies? The calibration efforts involved comparing the simulated fluxes to observed fluxes from the Littoral Sediment Study and previous studies conducted by Dr. Mitch Hogsett. 9 How does the Utah Lake nutrient model represent the bioavailability of particulate and dissolved phosphorus fractions? Less input from WWTPs would potentially result in most phosphorus input into Utah Lake being dominated by particulate forms. The model does capture nutrient speciation. The model uses the partitioning coefficient approach to calculate particulate and dissolved fractions of orthophosphate. For organic forms, the modeling team has state variables for dissolved and particulate phosphorus. The model also accounts for chlorophyll and algae being more reactive to the dissolved fraction. Science Panel Discussion • It would be helpful to see the Utah Lake nutrient model outputs for organic and inorganic matter fluxes as a time-step to observe if there are seasonal trends. Tetra Tech can produce that time step. • The study conducted by Dr. Steve Nelson from BYU to determine the anthropogenic effects on eutrophication of Utah Lake indicated that there is a large amount of orthophosphate in the porewater. Under these conditions, reducing phosphorus in the water column will trigger phosphorus release from the porewater into the water column. The Utah Lake nutrient model could simulate this dynamic by accounting for the bulk density of sediment and calculating the porewater volume. However, this simulation is not currently part of the sediment diagenesis model. A significant departure from the model would result in the need to re-calibrate the model, which would require additional resources and capacity. In its current form, the model can simulate the flux of orthophosphate to the water column under multiple future conditions, including how the lake will respond if external phosphorus sources are reduced. The model is trying to keep the lake in equilibrium. • One limitation of the model is that the sediment diagenesis assumes the sediments are static, so it cannot model the impact of large wind events on nutrient release. The current sediment diagenesis representation in the model is the most sophisticated version available. There are efforts to simulate resuspension events as part of the sediment diagenesis module, but those efforts are in progress. • A foundational question for the Science Panel to evaluate using the mechanistic model is how the sediments will interact with the water column with a reduction of anthropogenic loads. Understanding the base model very well will be important to think through future scenarios. • Iron-bound phosphorus has a large role in Utah Lake nutrient cycling. The introduction of oxygen can release phosphorus from its iron-bound form, at which point it becomes available for cyanobacteria. This dynamic should be considered within the model. • High phosphorus porewater concentrations are a general feature of lakes. For most lakes, the sediments are still acting as a net sink. The question is whether removing phosphorus from the water column will result in the release of phosphorus from the sediment. Another consideration is that Utah Lake is shallow, so wind events impact the sediments. Eventually, Utah Lake will reach a new equilibrium if the phosphorus concentration in the water column is reduced; the question is how long it will take to reach that equilibrium. The mass balance model can help identify how long it will take for Utah Lake to reach a new steady state. • Phytoplankton also removes the phosphorus released from the sediment into the water column. The algal blooms are taking up phosphorus pulses, so it is important to consider these biological influences. • It would be helpful to have a graphic explaining what is represented and what is not in the model. This graphic would help the Science Panel understand how to interpret the results from the model. 10 • The intention of using the model scenarios to evaluate future conditions is to establish nutrient targets. By running scenarios, the Science Panel can better understand how relative decreases in nutrient loads impact the lake. The scenarios can help build a response curve for criteria setting by varying proportional load reductions. Once targets are in place, the Science Panel and Steering Committee can run scenarios for implementation planning. • The Science Panel will form a model subgroup to better understand the tool's functionality and how to apply it to the S-R analysis. Thad Scott, Theron Miller, Tim Wool, Mike Brett, and Greg Carling volunteered to join the model subgroup. The modeling subgroup will meet once the Utah Lake nutrient model's calibration, sensitivity, and uncertainty analyses are complete. Public Comment The phosphorus flux from the sediment on windy days is an important contribution to Utah Lake. Once the phosphorus is released, phytoplankton quickly consumes the phosphorus. Zooplankton, in turn, consume the phytoplankton, and fish, in turn, consume the zooplankton. This dynamic will not be observed on a monthly timescale because it is associated with sporadic events. The Limnocorral Study, particularly the experiment that involved increasing phosphorus concentrations in one of the limnocorrals, provides evidence that this dynamic is occurring in Utah Lake. TECHNICAL SUPPORT DOCUMENT (TSD) AND REMAINING ULWQS TASKS PRESENTATION Kevin Kratt, Tetra Tech, and Mark Fernandez, Tetra Tech, presented an overview of the TSD and the remaining tasks for the ULWQS. The presentation is summarized below. • The TSD report will bring all the different lines of evidence to link management goals to response conditions to nutrients. The TSD will present the statistical analyses that will form the technical basis for the criteria recommendations. • The Science Panel will translate the TSD into protective nutrient ranges, identify uncertainty, and report the results to the Steering Committee. • The Science Panel will also describe Utah Lake's complex limnology by responding to the Steering Committee charge questions. The Science Panel has already completed the first draft of responses to the charge questions. They will update the responses as new studies are completed. • The Steering Committee will begin implementation planning to determine how NNC will be achieved with the management strategies available. The implementation planning phase will consider source partitioning, cost and feasibility, and timelines. • The lines of evidence (e.g., ULWQS studies, empirical S-R modeling, watershed mechanistic model, lake mechanistic model, literature) will be used differently across portions of the project. S-R ANALYSIS – SPATIAL AGGREGATION DISCUSSION Science Panel members discussed whether to consider Goshen Bay as a unique third region for the S-R analysis or combine it with the main basin. The Science Panel discussion and public comments are summarized below. Science Panel Discussion • The S-R analysis separates Provo Bay from the lake's main basin as two unique regions. At the May Science Panel meetings, Science Panel members discussed whether to separate Goshen Bay from the main lake as a third unique region. Science Panel members requested additional analyses to determine whether Goshen Bay has unique hydrological characteristics compared to the lake's main basin. An analysis of total dissolved solids 11 between the main basin and Goshen Bay suggests that Goshen Bay has unique properties compared to the main basin. However, the sample size is limited for Goshen Bay, which may make it difficult to conduct S-R analyses. • The total dissolved solids in Provo Bay are lower than the total dissolved solids in the main basin because of the freshwater input into Provo Bay. The water residence time is shorter in Provo Bay than in Utah Lake. Similarly, there is almost no freshwater inflow to Goshen Bay, which is why its total dissolved solid measurements are relatively high. • The number of data points for Provo Bay and Goshen Bay appear to be similar, which may be a justification for considering Goshen Bay as a third unique region. • One challenge with separating Goshen Bay as a unique region is delineating where Goshen Bay begins, and the main basin ends. The observational data does not provide sufficient data to draw a boundary between Goshen Bay and the main basin. The mechanistic model may provide better information to draw a boundary. • The only reason to identify Goshen Bay as a third unique region is if it has unique loading conditions. Science Panel Direction The Science Panel supported combining Goshen Bay and the main basin as one region for the S-R analysis. Public Comment One difference between Goshen Bay and the main basin is how wind affects the water bodies. Goshen Bay is so shallow that there are no clear circulation patterns. Despite this difference, the direction from the Science Panel to combine Goshen Bay and the main basin makes sense. S-R ANALYSIS – MICROCYSTIN TRANSLATOR APPROACH DISCUSSION Science Panel members discussed whether to pursue a translator approach for microcystin concentrations between beaches/marinas and open water. The Science Panel discussion and public comments are summarized below. Science Panel Discussion • Following the direction from the Science Panel at the May meetings, Tetra Tech developed several graphs to compare beach, marina, and open water data. They also developed a scatter plot that compares beach/marina data with open water data collected on the same date to see if a linear regression translator would be viable. The scatter plots show variable results, with some showing patterns while others are more random. • Something for the Science Panel to consider is whether Microcystis forms better near beaches than in the open water or if Microcystis is being blown from open water to the beaches. • More phytoplankton and toxin data are available for beaches/marinas than nutrient data. The purpose of the translator is to assess the relationship between open water conditions and phytoplankton and toxin data at beaches/marinas. The statistical translator would help define the open water conditions that would support designated uses in nearshore locations. Essentially, instead of having beaches as a separate assessment unit, one could manage the rest of the lake to support designated uses at the beaches. • The eastern side of Utah Lake has higher sediment nutrient content and is more likely to recycle local nutrients. The middle and west side of Utah Lake does not have the same issues with beach conditions as the eastern side of the lake. One potential explanation is that the wind is blowing the algae to the northeastern side of the lake. 12 • There are different sources of nutrients at different places around Utah Lake. The different nutrient recycling dynamics across Utah Lake make it difficult to use a translator approach. • The algal cells are undergoing lysis near the beaches, so there will be higher toxicity levels near the beaches. The algae are likely blown or transported to the beach. Microcystin levels are higher at the beach than at the open water sites, but there is no difference in cyanobacterial cell counts between open water and nearshore locations. • The translator approach is an assessment of the relationship between data points; it is less an explanation of the mechanics of why that relationship is occurring. • Future assessments of lake conditions to determine whether uses are being attained will still occur at the site scale. Even if the Science Panel adopts the translator approach, samples collected close to the beach will still be assessed independently. • The ULWQS Bioassay Study conducted by Dr. Zach Aanderud at BYU used cubitainers to assess the impact of phosphorus and nitrogen on toxin concentrations. The study indicated the relationship between toxins and nutrients varied by season. Cyanotoxins were highest in the spring, possibly associated with high runoff into the system. Toxins were not directly related to the size of the blooms themselves. • The toxin concentrations measured in the Bioassay Study varied by location on the lake. The toxin concentrations were lower on the western side of the lake compared to the concentrations in Provo Bay. The toxin concentrations also depended on carbon, nitrogen, and phosphorus requirements. Temperature was the biggest indicator of cyanotoxin production. • Species with the genetics to produce toxins do not always produce them. The literature is clear under what nutrient conditions those species will produce toxins. It might be good to develop a screening level based on toxin production and the risk of exposure. For example, once there is a bloom in the lake, excess nitrogen in the water column could potentially trigger greater toxin production. Supersaturated oxygen levels can also produce high microcystin levels because the algal calls produce microcystin to detoxify themselves of hydrogen peroxide, which forms under very high oxygen levels. Oxygen and nitrogen levels are both connected to the production of microcystin. • There may be different species producing toxins between the beaches and the open water. For example, there is more microcystin near the beaches than in open water and more anatoxin-a in open water than near the beaches. • Most toxin measurements in Utah Lake are of microcystin, partly because it is difficult to monitor anatoxin-a and cylindrospermopsin. • Developing a translator that can operate under different conditions may be one way to capture various dynamics. • It is important to protect the beaches. Whether a translator is appropriate depends on how close a relationship there is between open water and the beaches. If the relationship is not close, it may be important to develop two different assessment strategies. Public Comment • TSSD is researching the relationship between blooms and toxins and may have data on what conditions trigger toxin production. • Geosmin may be another chemical of interest. Geosmin is responsible for the off-flavor of fish in Utah Lake. Considering how valuable of a resource the fish are to Utah Lake, it may be another chemical to track. 13 Science Panel Direction The Science Panel supported maintaining the translation already developed between beaches/marinas and open water sites and exploring further whether different taxa influence the relationship between beaches/marinas and open water sites. S-R ANALYSIS – PREDICTION INTERVAL DISCUSSION Science Panel members discussed the appropriate prediction interval to assess S-R relationships. The Science Panel discussion is summarized below. Science Panel Discussion • The Tetra Tech team plotted linear and logistical regressions for the different S-R relationships. All the linear regression graphs from the S-R relationship analyses conducted for this meeting are based solely on open-water data and do not include beach or marina data. The logistic regression graphs used open water, beach, and marina data. • Tetra Tech plotted a linear regression of chlorophyll-a measurements against the daily maximum pH to assess the relationship between the two variables. The extent to which the Science Panel can draw potential endpoints for chlorophyll based on this S-R relationship depends on the size of the prediction interval. Tetra Tech currently set the prediction interval at 50%. • It would be helpful for the Science Panel to consider and provide direction on the appropriate size of the prediction interval for the S-R relationships. • The prediction interval identifies the exceedance probability and should be related to the exceedance threshold standard. Ultimately, the size of the prediction interval is a risk management decision. A larger prediction interval would result in an NNC that assumes less risk of exceedance. • The question on the appropriate size of the prediction interval applies to several linear S-R relationships, including chlorophyll-daily maximum pH, cyano biovolume-microcystin, chlorophyll-cyano biovolume, TP-cyanobiovolume, and TN-cyanobiovolume. • Some S-R assessments do not have established thresholds, such as chlorophyll-cyano biovolume. The S-R analysis approach involves developing a threshold for cyano biovolume using the cyano biovolume-microcystin relationship and then using that threshold to identify a chlorophyll threshold. Due to this approach, using the same prediction interval to assess the cyano biovolume-microcystin relationship and the chlorophyll-cyano biovolume relationship may make sense. • The daily maximum pH threshold is a not-to-exceed limit of 9. There is also a minimum of 6.5, but pH conditions that low do not exist in Utah Lake. • In the Environmental Protection Agency (EPA) National Lakes Assessment (NLA), they used a credible interval and developed a slider to assess the credible interval from 50% to 95%. • Six samples in the open water data exceeded the microcystin threshold of eight micrograms/liter. All of those samples were surface grab samples in the photic zone, which raises the question of whether only using samples from the photic zone biases the results. Many more samples in the beach and marina data exceed the microcystin thresholds. A translator approach may be a better way to compare open water data and beach/marina data to determine NNC in open water that are protective of uses in the beaches and marinas. Science Panel Direction • The Science Panel supported having Tetra Tech create separate S-R analysis graphs for Provo Bay and main basin results. They also supported having Tetra Tech work with the 14 DWQ Standards and Assessment staff to develop a recommendation for the prediction interval to discuss at a further meeting. • It would be helpful to have a comprehensive list of the existing standards in Utah. DWQ can produce a list of existing water quality standards in Utah for the Science Panel. S-R ANALYSIS – SECCHI DEPTH DISCUSSION Science Panel members discussed whether to use the S-R relationships that related Secchi depth with chlorophyll, TN, and TP. The Science Panel discussion and public comments are summarized below. Science Panel Discussion • Secchi depth in Utah Lake is confounded by the fact that approximately 75% of turbidity in Utah Lake comes from non-algal materials (e.g., suspended sediment, dissolved organic matter). The remaining 25% is related to primary production. • Inorganic particulate matter interferes too much with Secchi depth to make Secchi depth a good predictor of algal blooms. • There are seasonal patterns associated with algal blooms that affect Secchi depth. Since blooms seasonally affect Secchi depth, including the S-R relationship between Secchi depth and chlorophyll, TN, and TP would be useful. • Cyanobacteria can modulate their position in the water column by altering buoyancy. The position of cyanobacteria in the water column at any given time can affect Secchi depth, which makes it less useful as an indicator in the S-R relationship. • Using the S-R relationship between Secchi depth and chlorophyll, TN, and TP to establish NNC is not useful. However, this analysis is useful when considering what light penetration is needed for macrophyte restoration at the bottom of the lake. If the goal is macrophyte restoration, there would need to be a different method to establish light targets besides the S-R relationship between Secchi depth and chlorophyll, TN, and TP. One would need to know what clarity macrophytes need to re-establish. Public Comments Dr. David Richards at Oreo Helix Ecological, Inc. and Dr. Gus Williams at BYU have collected weekly measurements of Secchi depth and other indicators in 2021 and 2022. DWQ also collects photosynthetically active radiation (PAR) measurements in monthly grab samples. That information may be useful when considering how light limitations affect efforts to re-establish macrophytes. Science Panel Direction The Science Panel supported not using the relationship between Secchi depth and chlorophyll, TN, and TP to set targets for NNC due to the uncertainty created by the prevalence of suspended sediments. S-R ANALYSIS – CYANOBACTERIA CELL COUNT/BIOVOLUME CRITERIA DISCUSSION Science Panel members discussed whether using cyanobacteria cell counts and biovolume metrics to establish NNC is appropriate. The Science Panel discussion and public comments are summarized below. 15 Science Panel Discussion • The EPA does not have established cyanobacteria cell count or biovolume thresholds above which there is a risk to public health. Given that the EPA does not have cell count or biovolume thresholds, there is a question on whether it is appropriate for the State of Utah to use cell count or biovolume thresholds to inform NNC development. • The states can create standards for whatever metrics they want so long as EPA approves the criteria. • The ULWQS Steering Committee had a similar discussion on whether using cell counts or biovolume to establish criteria was appropriate. The relationship between human health and toxin concentrations is more clear than the relationship between human health and cell counts. There were different perspectives on whether it was appropriate, but the Steering Committee ultimately decided to explore the relationship. The State of Utah sets recreational health advisory guidelines for harmful algal blooms (HABs) based on cell counts and toxin concentrations. The cell count threshold for the recreational advisory is 100,000 cells/milliliter. • Some studies showing a relationship between cell count and allergic reactions (e.g., Pilotto et al. (1997)) may be confounded by other variables. • Individual cell size is another confounding variable in the impact of cell count on human health. A cyanobacteria cell can grow up to three times larger than an "average" size. Using phycocyanin measurements, which DWQ collects every 15 minutes at the buoys, would better assess cyanobacteria presence. Phycocyanin is a better measurement than biovolume because some small cyanobacteria that can produce toxins do not get counted in biovolume measurements. For example, picoplankton do not get counted, and DNA tests in the past have indicated that picoplankton make up 30-60% of the DNA in the water column. The calculation from cell counts to biovolume is normally based on a standard multiplier, which can vary from lake to lake. It would be helpful to conduct an S-R assessment between chlorophyll and phycocyanin to see if there is a relationship. • The goal of the S-R analysis is to be responsive to the Steering Committee's management goals. They seek guidance from the Science Panel on how to approach some of these complex issues. Public Comment The goal for recreational use in Utah Lake is that someone can swim and ingest small amounts of water without experiencing negative health impacts. S-R ANALYSIS – CHLOROPHYLL-TN AND TP RELATIONSHIP DISCUSSION Science Panel members discussed the relationship between chlorophyll-TN and chlorophyll-TP. The Science Panel discussion and public comments are summarized below. Science Panel Discussion • Tetra Tech assessed the relationship between TN and TP with chlorophyll using a quantile regression due to the heterogeneous variance across the distribution of the relationships, which violated the assumptions of linear regression. Science Panel input is needed to define what exact quantiles should be used. • The TN-chlorophyll quantile regression shows that Utah Lake is nitrogen-limited and also displays the inefficiency of nitrogen fixation in Utah Lake. • Similar to the outcome of the discussion on prediction intervals for linear regressions, Tetra Tech will work with DWQ Standards and Assessment staff to develop a recommendation. 16 S-R ANALYSIS – NLA MODEL DISCUSSION Science Panel members discussed how to apply the NLA model to assess S-R relationships. The Science Panel discussion and public comments are summarized below. Science Panel Discussion • Tetra Tech overlaid paired microcystin-chlorophyll data from Utah Lake on the paired microcystin-chlorophyll data from the NLA. The number of paired microcystin-chlorophyll data for Utah Lake is limited, but the existing data overlaps well with the NLA dataset. The plan is to use the NLA model to identify lakes with similar characteristics to Utah Lake (ecoregion and lake maximum depth) to compare chlorophyll targets. • Tetra Tech also overlaid the paired data for chlorophyll-TN and chlorophyll-TP from Utah Lake with the NLA dataset for lakes across the continental United States. The overlay showed a strong relationship between the distribution of chlorophyll-TP data at Utah Lake and the distribution of chlorophyll-TP data at the national level, particularly at chlorophyll levels above ten micrograms/liter. The relationship between the distribution of chlorophyll-TN at Utah Lake and the distribution of chlorophyll-TN at the national level was weaker. Tetra Tech recommends using the national relationship between TP and chlorophyll as another line of evidence for developing NNC but not TN and chlorophyll. • It is more difficult to relate nitrogen to chlorophyll than phosphorous to chlorophyll because there are more confounding variables in the TN-chlorophyll relationship, such as denitrification. Phosphorus cannot leave a system like nitrogen can. Nitrogen fixation occurs in Utah Lake but not at a quantifiably important rate. • Light limitation could be a confounding variable in the TP-chlorophyll relationship. There has to be light for phosphorus to impact chlorophyll levels. • The results of the ULWQS Bioassay Study indicated that photosynthetic organisms were being limited by nitrogen and phosphorus, but the response of cyanobacteria was mostly related to phosphorus. The chlorophyll measurement is capturing cyanobacteria and phytoplankton. • The very high levels of chlorophyll and TP are potentially the result of taking scum as a sample. When one takes a sample of scum, they are collecting a lot of biomass, which has a lot of phosphorus. These samples would not be representative of average conditions. However, any chlorophyll sample in Utah Lake would have been an integrated composite sample, which negates this concern. • It would be helpful to see an overlay with the NLA dataset that separates Provo Bay and the main basin. Science Panel Direction The Science Panel supported using the NLA dataset for chlorophyll-TN and chlorophyll-TP as a line of evidence in the TSD. S-R ANALYSIS –LAKE VISITATION OVERVIEW Science Panel members discussed how to use lake visitation data in the S-R analysis. The Science Panel discussion and public comments are summarized below. Science Panel Discussion • The State of Utah does not track exact visitation numbers for Utah Lake State Park. They use a formula to estimate visitation numbers based on the number of annual passes sold, which allow access to any state park. Additionally, Utah Lake State Park is not the only entry point to Utah Lake, so it is difficult to estimate the total number of people visiting Utah Lake. 17 • The Utah Lake Authority is installing monitors at all marinas to count cars, which will provide more reliable data in the future. • Due to the lack of available data on visitation numbers, it is difficult to connect the number of visitors to water quality conditions. Public Comments The Montana Department of Water developed a water quality standard for recreation by surveying people to see how likely they are to use a stream depending on how much filamentous green algae are at the bottom. Researchers at Utah State University (USU) conducted a public survey that also identifies how likely a person is to recreate in Utah Lake based on the color of the water. Those results will be available soon. MASS BALANCE MODEL DISCUSSION Science Panel members discussed the option to create a mass balance model. The Science Panel discussion and public comments are summarized below. Science Panel Discussion • An interactive user interface that allows a person to adjust conditions in the mass balance model could help support the work of the Science Panel and Steering Committee. The Steering Committee developed an implementation framework that maps out several implementation scenarios. An interactive mass balance model could help the Steering Committee assess these scenarios to evaluate cost-effectiveness and explore other solutions, such as macrophyte recovery. The mass balance model can also help inform some of the responses to the charge questions. • The mass balance model can help explore what happens to the lake under different flow or loading values. It can also assess how long the lake will take to reach a new stable state based on impacts from hydrologic/climatic variability, lake level, population growth, and internal loading. • The regional estimates for the population around Utah Lake are that the population will double over the next 30 years. The population around Utah Lake has more than doubled since the 1980s, and the lake's phosphorus and chlorophyll concentrations have not changed significantly in that period. The intention of the mass balance model is to understand what happens in the lake under future conditions (e.g., population doubles) and develop management strategies to respond accordingly. For example, the model may indicate that doubling the population may not impact lake conditions. The Steering Committee can then use that information to identify or not identify relevant management strategies. • DWQ, with data from the Utah Lake Authority, put together graphs to show how lake levels have changed over time. They also developed graphs to show how Utah Lake's volume and surface have changed over time for the main basin and Provo Bay. The lake volume estimates came from the Bureau of Reclamation as well as light detection and range (LiDAR) surveys of the lake. The current mass balance model is the average for the whole lake and excludes Provo Bay. The lake levels and surface area data can help determine how nutrient transport and flow through Provo Bay to help set up a separate mass balance model for Provo Bay. • The mass balance model can assess how long the lake will take to reach a new steady state under different future conditions. Phosphorus concentrations could then be related to chlorophyll to assess use support. 18 • Iron-phosphorus decoupling and nutrients released from porewater are important to the mass balance equation. WFWQC is currently conducting experiments to assess the importance of these dynamics in nutrient cycling in Utah Lake. • The Steering Committee suggested evaluating the following scenarios: o Wastewater scenarios  1 milligram/liter TP (assumed 10 milligram/liter of total inorganic nitrogen (TIN))  0.3 milligram/liter TP and 6 milligram/liter TIN  Reverse osmosis o Stormwater (percent reductions) o Nonpoint sources (percent reductions) o In-lake nutrient management (percent reductions) • Suggestions from Science Panel members on what conditions to simulate using the mass balance model include: o Assess conditions backward in time during the 1980s o Explore how snowmelt variability (i.e., lower streamflow conditions) would impact lake level o Evaluate how an increase in impervious surfaces would impact the quickness of stormwater delivery o Adjust nonpoint source input into tributaries using studies and data (e.g., Olsen's research on reference conditions, citizen scientists sampling) that establish background nutrient concentrations Public Clarifying Questions Members of the public asked clarifying questions about the mass balance model. Their questions are indicated in italics below, with the corresponding responses in plain text. Does the mass balance consider top-down biological effects like carp? Dr. Mike Brett assessed how much phosphorus is removed from the lake due to carp harvest. The assessment estimated that roughly five tons of phosphorus/year is removed from Utah Lake due to carp harvest. The phosphorus removed is in a bioavailable form, unlike calcite-bound phosphorus. Another way carp can affect nutrients in Utah Lake is by stirring up sediment and releasing nutrients captured in porewater; this dynamic is not captured in the mass balance model. Public Comments • The WWTPs will have the best forecasting data on the impacts of population growth. It would be helpful to use their forecasting data to inform scenarios. The Steering Committee indicated they would like to explore scenarios based on low, medium, and high population growth projects. • The new Provo River Delta will capture phosphorus and nitrogen and reduce the loads to the lake. • There are geothermal springs that serve as an input into Utah Lake. They likely do not provide significant amounts of nitrogen or phosphorus into the lake, but they do provide marl. • When developing the mass balance model user interface, it is important to think of the audience. When using these tools, the public will usually assess the maximum conditions (e.g., maximum population growth, maximum nutrient input). It is helpful to bound the mass balance model by realistic scenarios. 19 PUBLIC COMMENT Members of the public commented on the Science Panel discussion at the meeting. Their comments are summarized below. • The goal of the June Sucker Recovery Implementation Program is to restore June sucker populations in Utah Lake. If they are successful and the June sucker populations begin to grow, they will be competing directly with carp. These dynamics will change the food web in Utah Lake. NEXT STEPS • The Tetra Tech team will continue to make progress on developing and finalizing the Utah Lake nutrient model and watershed model. The next step for the models is to complete the calibrations and conduct the uncertainty analyses. The Model Subgroup will plan to meet once the reports are ready. • Tetra Tech will continue to develop the TSD with the input provided by the Science Panel over the next several meetings. • Dr. Janice Brahney, USU, is working on the final report for the ULWQS Paleolimnology study. The P-Binding Study is also nearly complete. The Paleolimnology Study Subgroup and P-Binding Study Subgroups will convene to review and finalize those reports. Lastly, the researchers at USU are developing the final report of the recreation public survey. • DWQ is creating a technical support contract to support the ULWQS Steering Committee's future implementation planning process. • Tim Wool, Mike Brett, Hans Paerl, Mitch Hogsett, and Theron Miller volunteered to join a Mass Balance Subgroup to help provide input on developing the mass balance model user interface. • The next Science Panel meeting will occur once the Model Subgroup has reviewed and provided feedback on the models. Any future Science Panel meetings towards the end of the year will likely be virtual.