The URL can be used to link to this page

Home My WebLink AboutDWQ-2024-004590June 30, 2023 Maddie Keefer, Afshin Shabani, Cole Blasko, Jon Butcher, Kevin Kratt (Tetra Tech) Utah Lake Watershed Model Update Note: all information and data presented are considered draft, in-process material 2 Agenda Note: all information and data presented are considered draft, in-process material Hydrology Calibration Recap Sediment Calibration Results Nutrient Calibration Status Update Modeling Process Gather Data Build Model Calibrate Model Assess Current Conditions Run Scenarios 3Note: all information and data presented are considered draft, in-process material 4 Watershed Modeling QAPP •Quality objectives for measured and modeled data •Model framework to support the project goals and objectives •Data collection and acquisition to support model build and calibration •Specification of quality assurance/quality control (QA/QC) activities to assess model performance •Model usability assessment Note: all information and data presented are considered draft, in-process material Hydrology Calibration Recap 5Note: all information and data presented are considered draft, in-process material Note: all information and data presented are considered draft, in-process material Hydrologic Calibration (complete) •Calibrated to multiple endpoints to ensure robust model: ▪Remotely-sensed snow depth and water storage ▪Actual evapotranspiration ▪Daily, monthly, and cumulative gaged flow •Guided by multiple visuals and statistical metrics related to: ▪Total flow ▪Seasonal/monthly flows ▪High/low flow distribution ▪Nash Sutcliffe efficiency (NSE) coefficients •Daily streamflow records obtained from USGS monitoring sites •Calibration sought to obtain the best overall fit at multiple locations, with priority on larger tributaries to Utah Lake (Spanish Fork, Provo River) Note: all information and data presented are considered draft, in-process material Hydrologic Calibration Sites 7 Site Total Flow Relative Error Monthly NSE Provo River 10% 0.756 Spanish Fork 7% 0.814 American Fork 32% 0.737 Sixth Water Creek -2% 0.597 Salt Creek at Nephi -4% 0.994 Diamond Fork -4% 0.699 Currant Creek 42% 0.707 Hobble Creek 31% -0.054 Dry Creek -14% 0.582 Summit Creek -56% -0.017 Provo River (USGS Gage ID 10163000) Sediment Calibration Results 8Note: all information and data presented are considered draft, in-process material Note: all information and data presented are considered draft, in-process material •Observed sediment concentrations have various sources of uncertainty •Calibrate to multiple endpoints to ensure robust model: ▪Daily, monthly, and cumulative monitored sediment concentration ▪Sediment Source Assessment ▪Reach/Sediment Balance •Objective 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) ▪Using sediment loading rates that are consistent with available values and providing a reasonable match with instream sediment data 9 Multi-Objective Sediment Calibration Note: all information and data presented are considered draft, in-process material Example of Debris Flows after Spanish Fork Region Wildfires Note: all information and data presented are considered draft, in-process material Sediment Calibration •Guided by multiple visuals and statistical metrics related to: ▪Sediment loading rates from the landscape ▪Seasonal/monthly TSS concentrations ▪High/low flow distribution ▪Average and median concentration and load error •Observed TSS records obtained from Utah stream monitoring sites •Calibration seeks to obtain the best overall fit at multiple locations, with priority on larger tributaries to Utah Lake (Spanish Fork, Provo River) 11 Note: all information and data presented are considered draft, in-process material 12 Sediment/WQ Calibration Sites Site TSS Samples Provo River at Murdock and Olmstead 79 Provo River at Wildwood 44 Hobble Creek 132 Spanish Fork at Utah Lake 104 Spanish Fork at Moark 71 Diamond Fork 89 Starvation Creek/Upper Soldier Creek 11 Thistle Creek 21 Currant Creek 46 13 HSPF Simulated Average Annual Unit Area Sediment Loading Category 2016 NLCD Land Use (Years 2008-2016) Fire Modified Land Use (Years 2017-2021) Average Annual Unit Area Sediment Loading Rate (lb/ac/yr) Percent of Watershed Area Average Annual Unit Area Sediment Loading Rate (lb/ac/yr) Percent of Watershed Area Agriculture 0.33 10.03% 0.53 10.00% Barren 5.59 0.30% 11.63 0.28% Buildings/roofs 128.76 0.50% 115.31 0.50% Burned -- 0.00% 25.88 12.11% Forest 0.28 50.68% 0.003 42.46% Grass/Shrub 0.37 29.99% 0.81 26.28% Roads/parking lots 137.71 1.24% 132.00 1.24% Urban/lawn (not sewered)0.10 6.26% 0.23 6.14% Urban/lawn (sewered)0.04 0.02% 0.12 0.02% Water 0.13 0.99% 0.36 0.97% Note: all information and data presented are considered draft, in-process material Sediment Calibration Results Summary Note: all information and data presented are considered draft, in-process material 14 Station ID(s) Description HSPF Reach Average Concentration (mg/L) Relative Error on Concentration Observed Simulated Average Median UTAHDWQ_WQX-4996780, UTAHDWQ_WQX-4996810, UTAHDWQ_WQX-4996778 Provo River at Murdock Div, Provo River at Olmstead Div, Provo River at Murdock Div Dam 19 6.0 6.9 14.9% 4.4% UTAHDWQ_WQX-4996850 Provo River at Wildwood 20 8.2 8.2 -0.5% 8.3% UTAHDWQ_WQX-4996100, WFWQC_UT-4996100 Hobble Creek at I-15 BDG, Hobble Creek at I-15 BDG 40 9.1 10.6 15.8% 19.7% UTAHDWQ_WQX-4995580, UTAHDWQ_WQX-4995578 Spanish Fork River at Lakeshore, Spanish Fork River at Utah Lake Inlet 47 107.9 124.7 15.6% 7.2% UTAHDWQ_WQX-4995600, UTAHDWQ_WQX-5919980 Spanish Fork River at Moark Div, Spanish Fork River near Moark Junction 48 155.6 133.3 -14.3% 1.2% UTAHDWQ_WQX-4995690, UTAHDWQ_WQX-4995670 Diamond Fork above Monks Hollow, Diamond Fork at Campground 51 33.9 38.5 13.6% 9.5% UTAHDWQ_WQX-4995310 Currant Creek at US-6 Crossing 77 17.2 12.8 -25.6% 18.2% Note: all information and data presented are considered draft, in-process material 15 Provo River at Murdock and Olmstead (Reach 19) Note: all information and data presented are considered draft, in-process material 16 Spanish Fork at Utah Lake (Reach 47) Water Quality Calibration Update Note: all information and data presented are considered draft, in-process material Note: all information and data presented are considered draft, in-process material •Sediment calibration follows the hydrologic calibration and precedes nutrient calibration •Water temperature, DO, and nutrient calibration follows sediment 18 Approach for Watershed Model Calibration Hydrology Calibration Sediment Calibration Nutrient Calibration 19 Steps in Nutrient Calibration Note: all information and data presented are considered draft, in-process material 1.Estimate all model parameters, including land use specific accumulation and depletion/removal rates, washoff rates, and subsurface concentrations 2.Compare simulated nonpoint loading rates with expected range of nonpoint loadings from each land use and adjust loading parameters when necessary to improve agreement and consistency 3.Calibrate instream water temperature 4.Compare simulated and observed instream concentrations at each of the calibration stations 5.Analyze the results of comparisons in steps 3 and 4 to determine appropriate instream and/or nonpoint parameter adjustments, and repeat those steps as needed until calibration targets are achieved; ▪Watershed loadings are adjusted when the instream simulated and observed concentrations are not in full agreement ▪Instream parameters are adjusted within the range determined reasonable Note: all information and data presented are considered draft, in-process material •Instream temperature is an important parameter for simulating biochemical transformations •HSPF modules used to represent water temperature include PSTEMP (soil and ground water temperature) and HTRCH (heat exchange and water temperature within flowing reaches) 20 Water Temperature Calibration Station Monitoring Period Model Reach Average Temperature (°F)Average Error (°F) Average Relative Percent ErrorObserved Simulated Provo River at Murdock and Olmstead 2/28/2008- 12/7/2021 19 51.7 56.1 4.4 8.5% Provo River at Wildwood 3/25/2008- 12/7/2021 20 49.8 49.6 -0.1 -0.2% Hobble Creek 3/7/2009- 12/8/2021 40 51.9 50.5 -1.3 -2.5% Spanish Fork at Utah Lake 7/17/2008-10/28/2021 47 51.7 49.1 -2.7 -5.1% Diamond Fork 1/11/2005- 8/28/2019 51 47.7 42.1 -5.4 -11.4% Simulated vs. Observed Water Temperature – Diamond Fork Note: all information and data presented are considered draft, in-process material Main objective 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 21 Nutrient Calibration Concentration Metrics NO2 + NO3 TP Average Observed Concentration (mg/L) 0.353 0.117 Average Simulated Concentration (mg/L) 0.274 0.088 Relative Median Error -12% 1% Example Nutrient Calibration Results – Utah Lake Note: all information and data presented are considered draft, in-process material •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, while ammonia and nitrate on pervious surfaces and all four constituents on impervious surfaces are represented as a buildup/wash off process 22 Upland Nutrient Loading Example Utah Lake Upland Nutrient Loading (Post-Fire Scenario) Note: all information and data presented are considered draft, in-process material •HSPF model simulates point and nonpoint sources of key stressors and provides the output necessary to develop source assessments •We 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 •Recall that some sources are input directly into lake model and are not in the HSPF model, such as: ▪Timpanogos Special Service District▪Orem Wastewater Treatment Plant (WWTP) 23 Source Summary of Annual Average Nitrogen and Phosphorus Loads Category 2006-2015 2016-2021 TN (lb/yr) TP (lb/yr) TN (lb/yr) TP (lb/yr) Barren, Burned 7,134 551 335,690 23,709 Grassland/Shrub, Pasture/Hay 37,766 7,199 50,412 9,042 Forest 34,233 10,425 34,865 10,551 Water/Wetland (with direct Atm. Dep) 13,030 1,074 21,437 1,090 Agriculture 28,736 2,332 43,570 3,528 Developed inc. Roads 117,378 13,772 153,681 17,108 Point Sources 78,774 10,970 78,774 10,970 24 Source Summary of Annual Average Nitrogen and Phosphorus Loads 2016 NLCD Land Use (Simulation Years 2006-2016) Fire Modified Land Use (Simulation Years 2017-2021) 25 Timeline Update Note: all information and data presented are considered draft, in-process material Sediment Calibration Completed May 2023 Draft Water Quality Calibration + Model Documentation Submitted to DWQ Present to SP June 2023 Address DWQ Comments/Improve Calibration June/July 2023 Submit Model Documentation to SP for Review July/August 2023 Questions and Discussion Note: all information and data presented are considered draft, in-process material Introduction to the Lake Model Sensitivity and Uncertainty Analysis Approach Science Panel Meeting | June 30, 2023 Note: all information and data presented are considered draft, in-process material Purpose and Agenda 2 •Purpose: •Provide background to Science Panel regarding sensitivity analysis of lake model Note: all information and data presented are considered draft, in-process material •Agenda •Definition of sensitivity analysis •Definition of model uncertainty, assumptions and methods •Description of First Order Variance Analysis (FOVA) •Uncertainty in TMDL implementation Simulation : 10% reduction of parameter/input data Sensitivity Analysis 3 Objective(s) •Evaluate impacts of parameter or model input changes on model outputs •Help identify most important physical or biochemical drivers of ecosystem response (mechanistic models) •Useful during model calibration to focus effort on sensitive parameters Results •High sensitivity: Model outputs change in same or higher proportion to model parameter/input changes •Low sensitivity: Model outputs show little to no response to model parameter/input changes Var Time Calibrated simulation Simulation : 10% increase of parameter/input data Ra n g e o f v a r i a t i o n Note: all information and data presented are considered draft, in-process material Uncertainty Analysis 4 Definition of Uncertainty •Estimate of dispersion (variance) around calibrated model outputs Objective(s) •Calibration:Minimize model fit errors •Uncertainty analysis:Estimate calibration variance Results •Calibration:Parameter values and RMSE, BIAS, MAE etc. •Uncertainty analysis:Calibration variance Tim e Variabl e Observation s Model output εi+ 1ε i 95% uncertainty bounds Note: all information and data presented are considered draft, in-process material Model Uncertainty 5 Very complex problem as some sources of uncertainty are unquantifiable Epistemic Ontological •Model structure errors •Input data errors •Parameter errors •Ecosystem response to stochastic natural processes/episodes•Fires, floods etc. Reducibl e Irreducibl e Note: all information and data presented are considered draft, in-process material Model Uncertainty - Example 6 From: https://www.opb.org/article/2022/09/27/predicting-hurricane-ian-s-track- has-been-difficult-an-expert-tells-us-why/ From: https://cbs12.com/news/local/tropical-storm-ian-strengthening-impacts- to-south-florida-lessening Note: all information and data presented are considered draft, in-process material Model Uncertainty – Methods and Assumptions 7 Assumptions are necessary to allow quantification of uncertainty •Uncertainty bounds depend on the statistical distribution of model errors •In most cases model errors are assumed independent, homoscedastic and normally distributed •Result:Uncertainty bounds unbiased or symmetrical around the calibrated (most probable) value. Problem reduces to find an error model that fits a Normal distribution! From: https://www.scribbr.com/statistics/normal- distribution/#:~:text=In%20a%20normal%20distribution%2C%20data%20are%20symmetrically %20distributed%20with%20no,same%20in%20a%20normal%20distribution.Note: all information and data presented are considered draft, in-process material Model Uncertainty - Methods 8 Stochastic Unconstrained •Monte Carlo Analysis Stochastic Constrained •Monte Carlo Analysis •Bayesian Monte Carlo Analysis •Bayesian Markov Chain Monte Carlo Analysis •Bayesian Model Averaging Deterministic •Fist Order Variance Analysis (FOVA) • Robust state of the art methods • Computationally intensive even using important sampling - Latin Hypercube • Require thousands of model simulations. • Typically limited to lumped and/or empirical models. • Hydrological lumped models • Computationally inexpensive • Applicable to mechanistic models Note: all information and data presented are considered draft, in-process material First Order Variance Analysis – Sensitivity Coefficients 12 DS C (v a r i a b l e ) Note: all information and data presented are considered draft, in-process material First Order Variance Analysis – Uncertainty Bounds 13 Va r i a b l e Co n c e n t r a t i o n Note: all information and data presented are considered draft, in-process material Uncertainty Analysis in the Context of Setting Numeric Nutrient Criteria •Model will be used to estimate key assessment endpoints under various conditions ▪Chlorophyll a, TN, TP, pH, Dissolved Oxygen •Sensitivity analysis will help identify most important physical or biochemical drivers of ecosystem response •Uncertainty analysis will help estimate dispersion (variance) around calibrated model outputs Questions and Discussion Note: all information and data presented are considered draft, in-process material Mechanistic Model Stressor Response Analysis Approach Science Panel Meeting | June 30, 2023 Note: all information and data presented are considered draft, in-process material Purpose and Agenda 2 •Purpose: •Remind Science Panel of Role of Models in Setting Numeric Nutrient Criteria •Request feedback on which scenarios to run, how to setup models, and how to process output Note: all information and data presented are considered draft, in-process material •Agenda •Use of Mechanistic Modeling in Setting Numeric Nutrient Criteria •Discussion of Specific Model Scenarios Lines of Evidence 1.Reference-based Results from paleolimnological studies Utah Lake Nutrient Model prediction/extrapolation of reference conditions 2.Stressor-response analysis Utah Lake Nutrient Model output Statistical models 3.Scientific literature Scientific studies of comparable/related lake ecosystems Support/supplement other lines of evidence Lines of Evidence 1.Reference-based Results from paleolimnological studies Utah Lake Nutrient Model prediction/extrapolation of reference conditions 2.Stressor-response analysis Utah Lake Nutrient Model output Statistical models 3.Scientific literature Scientific studies of comparable/related lake ecosystems Support/supplement other lines of evidence Pros and Cons of Mechanistic Modeling (Lake Model) •Pros Output based on deterministic relationships (i.e., causation not correlation) Output cover full spatial and temporal extent of model –Six years of hourly output within each grid cell –Multiple vertical layers for most cells Information available for significant number of parameters and fate/transport issues –Hydrodynamic parameters, TSS, all nutrient species, chlorophyll a, dissolved oxygen, pH –Algal growth, settling, re-suspension, sediment diagenesis •Cons Uncertainty (no models are perfect) Simplistic simulation of some key processes (e.g., food web dynamics) Output not directly comparable to sampling data typically used to make assessments Can’t output several potential assessment endpoints: –Cyanobacterial abundance, Microcystin concentration Stressor-response Relationship Pairs for Use in Deriving Endpoints Note: all information and data presented are considered draft, in-process material Use Assessment Endpoint Stressor Response Mechanistic Model Output Recreation Algal blooms Chlorophyll a Cyanobacterial abundance Yes Recreation, Aquatic Life pH Chlorophyll a pH Yes Aquatic Life DO Chlorophyll a DO Yes Aquatic Life Food resources Chlorophyll a Proportion cyanobacteria Yes Aquatic Life Light Chlorophyll a Kd, Secchi depth Yes Criteria Setting TN Chlorophyll a Yes Criteria Setting TP Chlorophyll a Yes Criteria Setting TN Cyanobacterial abundance Yes Criteria Setting TP Cyanobacterial abundance Yes Criteria Setting TN Kd, Secchi depth Yes Criteria Setting TP Kd, Secchi depth Yes Direct Model Output Estimated from Model Output Example Scenarios and Associated Output Note: all information and data presented are considered draft, in-process material Parameter Nutrient load alternative scenarios S1 (existing conditions) S2(no anthropogenic loads) S3(50% reduction in existing nutrient loads) TN (mg/L)1.44 0.98 1.20 TP (mg/L)0.070 0.053 0.062 Chlorophyll-a (µg/L)34.70 30.90 33.00 Questions for Science Panel? •How many scenarios? •How to setup models? •How to process output? How Many Scenarios? •Previously Discussed Existing conditions Prediction/extrapolation of reference conditions Reduced loading conditions •Potential Additional reduced loading conditions –How many? Others How to Setup Scenarios •Existing conditions = Calibrated model •Reference Conditions Removal of anthropogenic pollutant sources in the watershed •Reduced Loading Conditions Maintain current flow inputs to model Reduce nutrient concentrations Maintain sediment loads? How to Process Output •Spatial Issues One average value for all Provo Bay grid cells and another for all other grid cells? Average vertical layers or only use surface layer? •Temporal Issues All six years of model simulation? Entire year or growing season? Questions and Discussion Note: all information and data presented are considered draft, in-process material Project Orientation & Review of Technical Support Document Science Panel Meeting | June 30, 2023 Note: all information and data presented are considered draft, in-process material •Discuss details of more complex decisions o Period of analysis o Lake Regions o Phytoplankton o Zooplankton •Present initial draft of S-R models •Discuss Tt recommendations for S-R next steps Agenda Note: all information and data presented are considered draft, in-process material ULWQS Ongoing and Upcoming Activities •Technical Support Document Theme: link management goals to response conditions to nutrients Reference analysis, stressor-response analysis, literature evidence •SP NNC Recommendation Theme: translate TSD into protective nutrient ranges, with uncertainty & report to SC •SC Charge Questions Responses Theme: describe the complex limnology of Utah Lake Draft 1 completed previously 🡪look toward updating responses as new studies are completed •Implementation Planning Theme: Determine how NNC will be achieved with the “levers” available Takes into account source partitioning, cost & feasibility, timelines Note: all information and data presented are considered draft, in-process material Available Lines of Evidence Lines of evidence will be used differently across portions of the project Note: all information and data presented are considered draft, in-process material Technical Support Document: Reference Technical Support Document: S-R Analysis Technical Support Document: Literature SC Charge Questions Responses Implementation Planning ULWQS Studies Empirical S-R Modeling Watershed mechanistic model Lake mechanistic model Literature Pathway to Criteria (ULWQS Technical Framework) •SP recommendation SC recommendation Utah Lake Authority endorsement •Regulatory process multiple other groups involved with the process DWQ Water Quality Standards workgroup Legislature EPA Note: all information and data presented are considered draft, in-process material •Provide the technical basis for the development of numeric nutrient criteria (NNC) to protect designated uses •Recreation •Aquatic Life •Others (Agriculture, Downstream) •Conduct analyses to support multiple lines of evidence in the NNC framework Purpose of the Technical Support Document Note: all information and data presented are considered draft, in-process material Lines of Evidence 1.Reference-based Results from paleolimnological studies Utah Lake Nutrient Model prediction/extrapolation of reference conditions 2.Stressor-response analysis Utah Lake Nutrient Model output Statistical models 3.Scientific literature Scientific studies of comparable/related lake ecosystems Support/supplement other lines of evidence Note: all information and data presented are considered draft, in-process material Weight of Evidence •Ranges of nutrients deemed protective of uses across lines of evidence •How to distill these lines of evidence into a recommendation? Statistical distributions of endpoints Interpret endpoints in the context of their uncertainty weigh lines against each other by their relevance, strength, and reliability Note: all information and data presented are considered draft, in-process material Magnitude, Frequency, and Duration •Magnitude: “the maximum amount of the contaminant that may be present in a water body that supports the designated use” •Frequency: “the number of times the contaminant may be present above the magnitude over the specified period (duration)” Not to be exceeded x exceedances in a season x seasonal central tendency exceedances in y years •Duration: “the period over which the magnitude is calculated" Grab Seasonal average Weeks in exceedance •Extent: spatial aggregation Single station Lake regions (e.g., Provo Bay, Goshen Bay, etc.) Lakewide average For seasonal averages, average across sites then dates or across dates then sites? •Choose confidence/credible intervals and prediction intervals/quantiles to represent quantification of uncertainty and protectiveness General S-R Guidance Needed for Next Steps Note: all information and data presented are considered draft, in-process material Questions and Discussion Note: all information and data presented are considered draft, in-process material