1. Developing relations among human activities, stressors, and stream ecosystem responses for integrated regional, multi- stressor models R. Jan Stevenson 1, M. J. Wiley 2 D. Hyndman 1, B. Pijanowski 3, P. Seelbach 2 1 Michigan State Univ., East Lansing, MI 2 Univ. Michigan, Ann Arbor, MI 3 Purdue University, West Lafayette, IN Project Period: 5/1/2003-4/30/2006; NCX 4/30/2007 Project Cost: $748,527 Stevenson et al.

2. Natural Ecosystems Are Complex but can be Organized for Management Septic Systems Silviculture Livestock Grazing Irrigation Crop & Lawn Fertilizers Construction Organic/ Part PNC PO 4 NO x NH 3 HeatSediments Hydrologic Variability Nitrifying Bacteria Periphytic Microalgae Benthic Macroalgae Other Bacteria Benthic Invertebrates Fish Dissolved Oxygen Sewers & Treatment Herb Buffer Strips Tree Canopy Livestock Fences Ret. Basins & Wetlands Other BMPs Light Human Activities Stressors Endpoints Ecosystem Services Valued Ecological Attributes – Management Targets

3. Understanding how it all works: Complicating Issues Non-linearity and thresholds: –graded responses may be rare in complex systems. –thresholds complicate management choices. Complex causation: –multiple actions simultaneously shape biological responses. –issues of direct and indirect causation (effects): spurious correlations Scale and dynamics: –Potential stressors operate at different spatial and dynamic scales –Scales complicate the diagnosis of stressor-response relationships obscure causal dependencies through time lags, ghosts of past events, and misidentification of natural spatial/temporal variability. Stevenson et al.

4. Goals Relate patterns of human landscape activity to commonly co–varying stressors (nutrients, temperature, sediment load, DO, and hydrologic alterations) Relate those stressors to valued fisheries capital and ecological integrity of stream ecosystems Link empirical and mechanistic modeling approaches as a means to improving understanding and prediction Stevenson et al. G2M104070 Developing relations among human activities, stressors, and stream ecosystem responses for integrated regional, multi-stressor models

5. Approach 1.Building on other regional assessment & modeling by our team ( MI, IN, KY, OH, IL, WI) 2.Focus on basic interactions between landuse, hydrology, nutrients (CNP), and DO 3.Multi-scale Analysis: –Regional (Michigan) –(6) Focal Watersheds –Detailed Site monitoring 4.Modeling 1.empirical (statistical) 2.process-based (mechanistic) 3.hybrids ( a little of both!) using existing platforms and an integrated modeling system

6. Ecological significance Our project is focused on the streams and rivers of the Lower Michigan Peninsula. These are the veins and arteries of the Laurentian Great Lakes, the largest and most complex river-lake ecosystem in the world. What we learn here about multiple stressors is applicable in fluvial ecosystems anywhere. G2M104070 Developing relations among human activities, stressors, and stream ecosystem responses for integrated regional, multi-stressor models

7. Key findings 1.Urban land use is a stronger stressor than agricultural land use but agricultural impacts are more widespread. 2.Legacy impacts of landuse can be as important as current impacts. 3.Agricultural impacts appear to occur through a complex but tractable interaction of nutrient, hydrologic and metabolic stressors. 4.Impacts of specific stressors and their interaction varies with ecological setting in general; and specific hydraulic setting in particular. 5.Management expectations (ecological targets and assessment scoring criteria) need to be conditioned by ecological context of the site in question. G2M104070 Developing relations among human activities, stressors, and stream ecosystem responses for integrated regional, multi-stressor models

8. Lessons Learned Where exactly you look (sample locale), and at what scale you look (sample extent and frequency), affects what you can see (and model) We need more concise language to talk about multiple stressors and stresses [incorporate concepts of frequency, duration, co-variation and interaction, contingency] G2M104070 Developing relations among human activities, stressors, and stream ecosystem responses for integrated regional, multi-stressor models

9. Interactions & Users MDEQ nutrient criteria development MDNR groundwater protection criteria EPA nutrient criteria workgroups MDNR Ecoregional management teams GLFT Lake Michigan Tributary Assessments Local watershed groups (MWA, HRWC, MiCORP) G2M104070 Developing relations among human activities, stressors, and stream ecosystem responses for integrated regional, multi-stressor models

10. Graduate students supported: total of 10 across all 3 institutions M.S. theses developed/completed: 4 Extensive linkage with other EPA-Star, NSF, Great Lakes Fisheries Trust, and Great Lakes Fisheries Commission projects G2M104070 Developing relations among human activities, stressors, and stream ecosystem responses for integrated regional, multi-stressor models

11. 2006a Progress Report 1.Late start first year, 2004 first extensive field year, NCX to 2007 2.Analyses of regional, aggregated data sets underway! {first looks} 3.Analysis of 2004 and 2005 focal basin surveys continues {some highlights} 4.intensive hydrologic and WQ monitoring continues in Cedar and Crane Creeks 5.Integrated process modeling running for Cedar, underway for Brooks, Bigelow, & Crane {description and early results}

12. Large, Regional-Scale Statistical Modeling Urban and agricultural land use as key multiple stressors Urban and agricultural land use as key multiple stressors –Relative impacts? –Direct and indirect effects? {watersheds and riparian buffers} –Causal relationships? Intervening variables? Data assembled from MDEQ, Michigan Rivers Inventory, previous EPA-STAR, NSF, Muskegon River Assessment; registered on attributed NHD database (EPA-STAR/USGS AQGAP product) Data assembled from MDEQ, Michigan Rivers Inventory, previous EPA-STAR, NSF, Muskegon River Assessment; registered on attributed NHD database (EPA-STAR/USGS AQGAP product) Used regional Normalization approach to standardize datasets and metrics (fish and invertebrate) Used regional Normalization approach to standardize datasets and metrics (fish and invertebrate)

13. Regional modeling of Multiple-source assessment datasets: Patterns of human activities and fluvial ecosystem response Data coverage Fish & Invertebrate Multi-Metric

14. 5%, 50% 1%, 8% r= -.36 r= -.20 r= -.29 Regional “~dose-response” relationships to Land use Stressors Indicator: normalized EPT score [(obs-exp)/sd] %Urban in riparian buffer %Ag in riparian buffer Urb and Ag: geom. mean Noisey Linear(izable) Urb > Ag thresholds

15. Standardized Total Effects - Estimates xWT_urbxWT_agxRT_agxRT_urb xWT_ag-0.1520.0000.0000.000 xRT_ag-0.1180.7760.0000.000 xRT_urb0.9230.0000.0000.000 nEPT- 0.354-0.189-0.244-0.023 Issues of direct and indirect effects: Urbanization of Ag areas Multiple ways to represent land use/cover Structural Equation Modeling to sort out direct, indirect and total effects VEA:EPT score watershed Riparian buffer Results: Overall Urban stronger than Ag Riparian Ag > than Basin AG Basin Urban > Riparian Urban Best fitting, structurally plausible model

16. Training data Predicted ObservedN% Correct1-2345 99578.3927801386314 337223.656225884118 458719.0829610411275 524341.975822930102 Test (%20 withheld from training) Predicted ObservedN% Correct1-2345 24877.01619131206 310516.196817155 414513.10379261921 55836.2071612921 Attainment class thresholds Basin Urban 22.5% Basin Ag <=48.5% CART model fish & invert based Attainment Class CART of normalized overall fish and invert multi-metric

17. Statistical Modeling of Focal Basin dataset Agricultural impacts on Stream Ecosystems (6 )100-300 mi 2 systems representing a targeted gradient of agricultural land cover Cedar Creek –hIgh value fishery with Ag impacts, threatened by development Bigelow –Pristine high value fishery resource Mill Creek Brooks Creek –threatened by development currently with signif agricultural Crane Creek Sycamore Creek –intensive agricultural impacts What is the nature of biological responses to agricultural land use? 1.The case for chronic metabolic stresses –Agricultural land use and nutrients –Agricultural land use and dissolved oxygen dynamics 2.Highly variable response tied to variation in hydrologic/hydraulic/DO regime

18. Meso-scale empirical modeling (6) stream systems sampled across Ag and Hydrologic gradients Organic Carbon (COD) Inorg Nitrogen (ppm) Phosphorus (ppm) PM oxygen (ppm) % metabolic conformers EPT Taxa % surface breathers % Riparian Buffer area in Ag % Watershed area in Ag % Watershed in Ag % Watershed Ag % Watershed in Ag Multiple Local (direct) Stressors response to Agriculture (indirect stressor) % Riparian Buffer area in Ag Biological response to indirect Landscape stressors

19. Early Morning D.O. levels

20. Site-Intensive data collection & Integrated Mechanistic Modeling Test hypothesis that cause-effect relations in regional statistical models are plausible Test hypothesis that cause-effect relations in regional statistical models are plausible Understand how multiple stressors interact to cause biological response Understand how multiple stressors interact to cause biological response –Cedar Creek ** –Mill Creek* –Brooks Creek* –Crane Creek * –Sycamore Creek –Bigelow*

21. Integrated Modeling of Cedar Creek - Spatially & temporally intensive water chemistry and biological sampling

22. Holten River Rd. Holten to River Rd. Ratios Catchment area ratio= 26% Typical storm peak ratio = 80% Average flow ratio= 3% Max Q = 250 cfs Mean Q =2cfs groundwater Runoff [ 67%] Max Q = 200 cfs Mean Q =46cfs Groundwater [95%] Runoff [ 5%]

23. Holten Gage River Rd. Gage Poor Below expectation Acceptable Excellent Biological Quality Cedar Creek Bason Multi-Stressor Project Observed/Expected diversity

24. Habitat stress oxygen temperature bed transport Surface abstraction Weather model* Groundwater Model Basin Routing transforms Channel Routing transforms Channel hydraulics width depth velocity shear Thermograph HEC-HMS um HEC-RAS um MODFLOW msu KendallPREP msu DOSMOSC um SRTMum Landcover model* * or historical data Model accumulates hrs [or relative freq] of oxygen and bed mobilization stress over long period runs (e.g. 1-2 years) LTM2 purdue Linking local-scale mechanistic models for Causal evaluation and modeling experiments MT3Dmsu QUAL2Kmsu Or Water Quality Data

25. Hydrologic Modeling: Simulate Transient Fluxes to SW Preprocessor & MODFLOW –Inputs: Land Use (historical & LTM2) Regional Geology NEXRAD Precipitation NOAA Snow Depth MODIS LAI DEM Solar radiation HEC-HMS –Surface Water and channel routing

26. NEXRAD for Expanded Muskegon 10 yrs + 10 synth

27. Monthly Vegetation Density Distribution in Expanded Muskegon and Cedar Creek Weekly Leaf Area Index Model Based on MODIS coverage

28. Results % of precipitation that becomes recharge Landuse effects Recharge Cedar Creek well recharge monitoring Regional analyses indicate reduced recharge in agricultural vs forest watersheds

29. Results –Observations MODFLOW All head observations: R 2 = 0.81 Pre-1988: R 2 = 0.79 1988-2004: R 2 =0.89

30. Results MODFLOW Upper Cedar Creek Lower Cedar Creek

31. Nitrate Transport Simulation (MT3D) Used GW model fluxes Nitrate sources –Atmosphere –Agricultural lands –CAFOs –Septic systems Nitrate fluxes exported to stream ecohydrology model NO3, mg/L

32. Simulating Water Chemistry and Biological Response in Cedar Creek Using nitrate fluxes to Cedar Creek calculated in transport model QUAL2K 8 9 10 11 12 13 14 05101520 Distance Downstream (km) Water Temperature (°C) Simulated Water Temperature Observed Water Temperature 0 500 1000 1500 2000 05101520 Distance Downstream (km) Nitrate + Nitrite (ugN/L) Observed Nitrate Simulated Nitrate

33. Coupling models to generate realistic processes Recharge Model MODFLOW MT3D QUAL2Kw Site Biological response (annual) Recharge Groundwater fluxes Nitrate fluxes Stream concentrations Recharge Model MODFLOW HEC-HMS HEC-RAS MRI-DOHSAM Recharge Groundwater fluxes Watershed hydrology Channel hydraulics Cum metabolic stress (hr) (day)

34. Exceedence frequencies for Dissolved oxygen and bed mobilization Specified stress thresholds: O2 : 4 ppm Incipient Bed mobilization : ratio of ave. shear to D84 critical shear /5 Stress summary: as % of period Scour_stress = 56.8 O 2 stress = 2.5 Combined = 59.1 Simultaneous = <.1 CMSI MRI_DOHSAM cumulative DO & Hydraulic Stress Assessment Model 8 day simulation for Crane Creek Outlet channel using observed flow temp, depth and velocity data from an up-looking doppler sensor. Loading parameters BOD = 8 ppm, NH4=.2 ppm d84 4 ppm

35. %MC cum O2 stress 1 :.533.0.153.00.00.031 cum bed mobil 2 :.00.003.01.02.06.00 % Ag in Basin 57% 42% 37% 18% 18% 15% % Ag in RT 41% 33% 29% 21% 21% 14% Integrated Modeling of Cedar Creek Stress Assessment: year 2003 NexRAD with 1998 Landcover %MC EPT %MC = % of taxa that are Metabolic Conformers EPT = count (# species) of EPT Taxa Field data from our Biological Assessment 0 5 10 15 123456 Modeling Multiple stressors: hydraulics, temp, NH 4, TP, BOD Sensitive taxa EPT Metabolic conformers Number of genera @Brickyard @Crystal @M-120 @ Ryerson @Sweeter @River Rd

36. 2 4 6 8 10 12 14 -0.100.10.20.30.40.50.6 Cedar_metrics EPT Taxa Metabolic Conformers y = 12.094 - 16.126x R= 0.96021 y = 9.9731 - 13.722x R= 0.9414 Observed Number of genera Modeled cumulative oxygen stress

37. COD TP NH 4 Temp Hydraulics Relative effect {as % reduction} in total stress score -53% -0% -4% -73% -81% Cedar Creek e.g. Model “experiment” 1 Cedar@Brickyard site What are the individual effects of each stressor On cumulative stress? Sum >100% Hydraulics>temp>WQ

38. -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 050100150200250300350 [simulating simple response to a single stressor] @brickyard @crystal lake rd @m-120 @ryerson rd @sweeteer rd Below river rd Cumulative Metabolic Stress Index TP ppb e.g. Cedar Creek Modeling “experiment” 2 eliminating BOD and NH4 effects How do the sites respond to a TP gradient? @brickyard @m-120 Below river rd All others How spatially variable is Cedar Creeks response to TP loading? C&N set low BOD=1 NH 4 =.02 ppm

39. -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 050100150200250300350 Cumulative Metabolic Stress Index TP ppb @brickyard @crystal lake rd @m-120 @ryerson rd @sweeteer rd Below river rd Given current BOD and NH 4 stressors How do the sites respond to a TP gradient? e.g. Cedar Creek Modeling “experiment” 3 [simulating response to a single stressor in a Multi-Stressor setting] @brickyard @m-120 Below river rd All others Current concs How spatially variable is Cedar Creeks response to TP loading? Current elevated C and N concs

40. -120 -100 -80 -60 -40 -20 0 20 40 050100150200250300350 Cumulative Metabolic Stress Index TP ppb [simulating response to a single stressor in a multi-stressor setting] @brickyard @crystal lake rd @m-120 @ryerson rd @sweeteer rd Below river rd e.g. Cedar Creek Modeling “experiment” 3 Response to TP relative to current conditions @brickyard @m-120 Below river rd All others

41. Final Steps Model refinements –Regional & focal watersheds Complete model integration for focal watersheds Validate using bio-assessment data Re-visit regional empirical models based on mechanistic model insights; improve with stratification?