Presentation on theme: "Effects of urbanization and forest fragmentation on water quality … Rachel Riemann USFS-FIA Karen Murray USGS-NAWQA."— Presentation transcript:
Effects of urbanization and forest fragmentation on water quality … Rachel Riemann USFS-FIA Karen Murray USGS-NAWQA
Opportunity for collaboration In this study we take advantage of current USGS- NAWQA water quality monitoring efforts and link it to USFS-FIA’s current investigations into monitoring forest fragmentation and urbanization -- in order to better understand the relations between the two.
Problem We already know that urbanization has been linked to water quality in other studies But, what aspects of urbanization and/or forest fragmentation are most highly correlated with the biological, chemical, and physical responses observed in streams?
a combination of interests… USGS-NAWQA –Improve understanding of the components impacting water quality in order to better provide management guidelines for preventing or minimizing degradation in the face of development pressure –Improve understanding of the forms and/or thresholds of that impact USFS-FIA –Identify the components of frag/urban that are most related to observed changes in water quality, –Develop methods to monitor these relevant parameters of frag/urbanization with sufficient accuracy over large areas
Two rapidly urbanizing areas: Appalachian ecoregion, especially the Pocono Mountains area –Fastest growing counties in Pennsylvania –Second home and primary home development –Transitioning from forested to suburban Piedmont ecoregion –Including Philadelphia – Trenton corridor –Rapidly transitioning from agriculture to suburban Appalachian Plateau Valley and Ridge Piedmont Coastal Plain
Objectives Identify which management-relevant landscape characteristics are most related to stream water quality and ecological health. Describe the forms of these relationships. Determine the influence of landscape data source (on interpretation/findings). If necessary develop corrections or recommendations for use of those broad-area datasets currently available.
Predictor data used NLCD’92 wasn’t sufficiently accurate to do the job, particularly in the less urbanized Poconos region %urban (source: NLCD’92)
Predictor data used From photointerpretation of land use and land cover from digital aerial photography ( ; some CIR, some B&W) –Land use polygons –land cover data recorded for each urban developed land use class (% tree, grass, house, road) From Census Bureau data (2000) –Population –House density –Roads and road density (2000 TIGER data)
Site selection Minimize point sources Minimize natural variation – - basin size all mi 2 - slope - all upland, riffle/pool sites Accessible for sampling during both low and high flows Selected representative sampling reach m long 33 sites
Similar %forest and same amount of urban development, but different % forest in buffer, and different %C/I Illus maps East Branch Red Clay East Branch Brandywine
Similar %forest and amount of development, but a different distribution of land uses (COR, AI, and forest patch size covariance) Illus maps Dingmans Hay
Macroinvertebrates Algae Field data collected Habitat & geomorphology Nutrients, ions Pesticides in water Discharge (instantaneous) Temperature
Primary responses related to urbanization (rdden) What are the primary biological, physical, and chemical responses that are related to urbanization? Loss of sensitive macroinvertebrates Decrease in habitat quality Increase in chloride, sulfate, other major ions Increase in nutrient concentrations
2 What are the primary biological, physical, and chemical responses that are related to urbanization? Increase in Pesticide Toxicity Index Increased variety and amounts of pesticides detected (especially insecticides) Increased potential toxicity of streamwater to fish and invertebrates
Ecosystem responses… Results--What are the changes we see? Loss of sensitive macroinvertebrates Decrease in habitat quality Increase in chloride, sulfate, other major ions Increase in nutrient concentrations Increase in Pesticide Toxicity Index
To what specific landscape characteristics (or combination of characteristics) are these responses related? basin-wide land use buffer zone land use fragmentation indices for basin
Buffer-zone variables “Buffer – zone” landscape variable Sensitive invertebrates Chloride conc. Pesticide toxicity Habitat quality Forested %+--+ Multi-family residential %-++- Commercial-industrial %-++- Impervious %-++- Urban %-++- Buffer-zone variables * Buffer zone = 100m on either side of the stream
Landscape indices Landscape index Sensitive invertebrates Chloride conc. Pesticide toxicity Habitat quality Mean patch size - forest+--+ Avg patch perimeter-forest+--+ Aggregation index - forest+-- Centroid connectivity - forest+-- Edge - urban-+ Avg patch perimeter - urban+ Distribution/frag measures
Can we combine some of these landscape factors to develop models of stream ecosystem response?
Multiple linear regression – Invertebrate community structure* Variable added to model Model R-Square (p<0.01) % Forest in basin (+) 0.77 % Commercial in basin (-) 0.82 % Urban in buffer (-) 0.86 *ordination site scores MLR - invertebrates
Multiple linear regression – Total nitrogen (spring sample) Landscape variable added to model Model R- Square % Forest in basin (-) 0.68 Relative contagion (-) 0.76 % commercial/industrial in basin (+) 0.81 MLR-total nitrogen (spring)
What is the form of the response? What is the form of the response And how does data source affect observed patterns? NLCD’92 –Currently available over entire US NLCD2000 –Currently only exists in pilot areas. Expected to have US-wide coverage in the next 5 years or so…
“Correcting” the NLCD92 dataset NLCD’92 …with roads overlaid on top NLCD’92 – ‘corrected’ using local road density example in to the Poconos area…
Differences-plots Basin stats Buffer stats
Where it helps and where it doesn’t… Helps: –%urban land in basin
Not much help: –%urban land in buffer Where it helps and where it doesn’t…
Differences between them Description, comparison maps and comparison plots Photointerpreted land use (1999) NLCD’92 (note missing development) NLCD2000 (note land cover focus)
Looking at NLCD2000… Being a land cover product, NLCD2000 urban developed land uses are more related to impervious surface than the entire developed area. And, areas that are sparsely developed, have small house footprints, and/or have trees overshadowing roads or buildings may still contain only a few ‘developed’ pixels in the NLCD2000 dataset within a background of grass (or forest)
Concluding thoughts… –%Forest in the basin (and its close opposite--%developed) –The type of developed land in the basin (e.g. C/I) –Distribution of land uses within the basin can be a factor Amount of forest or urban in the buffer COR, AI-Forest, diversity of forest patch sizes –Land cover % impervious And, although the data wasn’t fully analyzed, there was some evidence suggesting that the land cover of developed land uses may be a factor as well (e.g. forest vs. grass covered residential). Landscape variables most related to stream ecosystem response
Concluding thoughts Data source –You sometimes need the detailed land use/land cover information to find out what’s really going on –And you need an understanding of its relationship to the broadly available datasets for extrapolation over large areas Be very careful using threshold values derived using one land use data source and applying them to another
Concluding thoughts The cooperative effort provided a unique opportunity –To link forest and water studies to expand ecosystem knowledge –To investigate the linkage between a process-level study establishing relationships between factors and broad scale methods for scaling the results up to an entire region.