1. Biologically based urban response models for the South Atlantic gulf and Tennessee River basins T.F. Cuffney, E.M. Giddings, and M.B. Gregory North Carolina Water Science Center

2. Objectives: 1.Derive invertebrate response models for Atlanta, GA, Birmingham, AL, and Raleigh, NC. 2.Test and verify urban response models (ATL, BIR, RAL). a.Comparability among models b.Applicability of urban models to sites throughout the Southeast – can models developed in one urban area be applied to other urban areas? – can urban models be used to predict regional patterns in water-quality?

3. Denver Portland Dallas Birmingham Atlanta Raleigh Milwaukee Boston Salt Lake City South Atlantic gulf and Tennessee river basins Seattle Sacramento National Water-Quality Assessment (NAWQA) Program 11 Urban Stream Studies

4. Biological sampling sites Birmingham Atlanta Raleigh 090 180 Kilometers 360

5. Raleigh Urban Study Area Piedmont

6. Urban models: Site Characteristics AbbrevCore urban area Dominant ecoregion Habitat sampledBasin area (km 2 ) ATLAtlanta, GAPiedmontWoody snags 43-146 RALRaleigh, NCPiedmontRiffles5-82 BIRBirmingham, ALCentral Appalachian Ridges and Valleys Riffles5-66

7. Design of urban studies 1.Urban intensity gradient defined by a multimetric index based on land- cover, population, infrastructure and socioeconomic factors associated with population density. 2.Uniform environmental setting to keep natural environmental factors as constant as possible.

8. % Impervious and Urban Intensity index (UII) -- Raleigh

9. Urban models Multivariate (ordination) models: –Measure ecological distances among sites along an environmental gradient (urban) derived from differences in invertebrate assemblages. Multiple regression models: –Relate invertebrate assemblage metrics (EPT) to landscape variables: land cover, land use, census variables, infrastructure.

10. Multivariate (ordination) models

11. Ordinations scores & UII Atlanta

12. RTH Inverts: urban model 10% impervious surface

13. Ordination site scores vs. UII Urban areaabR2R2 Atlanta1.67-0.0170.81 Birmingham1.84-0.0200.74 Raleigh1.75-0.0150.74 Y = a + bX Y = ordination site score X = urban intensity UII

14. Ordination model: prediction ATL Site score Species abundance from ATL. Optima for species K: BIR model Predicting ATL using BIR model Eigenvalue from BIR model

15. Observed Axis 1 site score Woody snags Riffles Piedmont Ridge & Valley Piedmont

16. Observed Axis 1 site score

18. Inter-model comparisons: ordination models 1.Ordination models from Atlanta, Birmingham, and Raleigh are interchangeable. 2.Ordination models developed in one urban area can be applied to other closely allied urban areas.

19. Regression models Comparison among ATL, BIR, RAL

20. Multiple Regression Models EPTr: richness of pollution intolerant forms (mayflies, stoneflies, caddisflies) X i : –Land-use: (e.g., forest, developed, ag/urban grass) % basin area % stream buffer –Population density –Road density Model: EPTr = f(X i )

21. EPTr urban model variables % basin areaCensusInfrastructure CityDevelForestGrassPop denRoad den RAL0.110.01-1.44 ATL0.270.33-3.19 BIR-1.13 % stream buffer CityDevelForestGrassR2R2 P RAL0.130.470.81<0.001 ATL0.450.62<0.001 BIR0.59<0.001

22. Correspondence between predicted and observed EPT richness based on multiple regression models ATL modeled using:BIR modeled using: ATLBIRRALATLBIRRAL b0.670.30 b0.640.610.94 r2r2 0.660.510.33r2r2 0.190.590.44 p<0.001 p0.011<0.001 Expected = a + b * Observed

23. Predicted EPT richness based on multiple regression models RAL modeled using: ATLBIRRAL b0.49 0.47 r2r2 0.550.620.61 p<0.001 Expected = a + b * Observed

24. Inter-model comparisons: multiple regression models 1.Variability associated with estimating explanatory variables reduces the overall “fit” of the regression models. 2.Multiple regression models are not as interchangeable among urban study areas as are the ordination models.

25. Extrapolating urban models to South Atlantic gulf and Tennessee (SAGT) River basins Can urban models be used to predict regional patterns in water-quality?

26. Site Characteristics Small basins (< 160 km 2 ) Dominant ecoregionSitesUII range Piedmont83-80 Interior Plateau353-32 Mid Atlantic Coastal Plain1412-33 Southeastern Plains484-81 Central Appalachian Ridges and Valleys111 Southwestern Appalachians29-16

27. Ordination models Observed vs. expected site scores

28. Interior Plateau: Ordination MODEL: UII: 3-32 Sites: 35

29. Southeastern Plains: Ordination UII: 4-81 Sites: 48

30. Multiple regression models Observed vs. expected EPT richness

31. EPT: multiple regression model UII: 3-32 Sites: 35

32. EPT: multiple regression model UII: 4-81 Sites: 48

33. Tolerances to urbanization differ from tolerances to general pollution General pollution tolerance values: USEPA and NCDENR Urban pollution tolerance values: NAWQA (UII) Species tolerances

34. Conclusions 1.Invertebrate responses to urbanization provide very useful predictive models. 2.Ordination models were generally better predictors of responses than were multiple regression models using metrics. 3.Urban models based on ordination were transferable among urban areas -- models developed for one urban area are applicable to other urban areas. 4.Urban models are specific to urbanization and do not provide a general means of modeling regional responses to water quality – network design affects application.

35. Next steps Derive more generalized land-use disturbance models (forest, agriculture, urban). Couple biological response models with SPARROW modeling in South Atlantic gulf and Tennessee River basins.