1. LANDSCAPE CHARACTERISTICS AFFECTING STREAMS IN URBANIZING SUBBASINS OF THE DELAWARE RIVER BASIN
Jeffrey M. Fischer1, Karen Riva-Murray1, Rachel Riemann2, and Peter S. Murdoch1
1 US Geological Survey
2 USDA Forest Service
And the development of approaches for monitoring the resulting parameters over large areas
Solid background is 18, 73, (27,127,217)
…and the impact of data source on results
…and the development of approaches for monitoring frag/urbanization over large areas
…a cooperative effort aimed at producing

2. What are the major effects of urbanization?
Major questions:
What are the major effects of urbanization?
What aspects of urbanization cause ecological responses? Which of these are ‘managable’?
How do these aspects compare among regions? What are the national commonalities?

3. Approach Control for natural variation among sites
Establish an apriori urbanization index based on available landscape data
Select sites so they are well-distributed along this gradient
Collect biological, physical, chemical data

4. Common problem: Landscape data issues
(Timeliness, accuracy, detail, availability)
NLCD 1992
Photo-
interpreted
Urb
For
Agr
NLCD
2000

5. Implications of NLCD inaccuracies
Mayflies:
Ephemeroptera
Stoneflies:
Plecoptera
Caddisflies:
Trichoptera
EPT richness: number of different:
But NAWQA didn’t have good land use data to work with – a problem since land use is known/highly suspected of having a major influence on water quality. Other variables (e.g. population and road density) are sometimes used as surrogates for urbanization/land use, but these don’t give us any opportunity for trying to tease out what aspects of urbanization or forest fragmentation are mostly highly correlated (the biggest culprits) with respected to the ecosystem repsponses observed in streams. “What aspects” can be important when we’re trying to to better provide management guidelines for preventing or minimizing degradation in the face of development pressure, or when we’re trying to more accurately model water quality and monitor for areas with these characteristics.
Developed a region wide method to correct for inaccuracies in satellite NLCD data using road density

6. Site selection for urban intensity gradient
Station
Road density (road miles/ sq. mi. basin)
Piedmont sites
Poconos sites 5 1 2 f l a t b u s r o d v n h y c w k p i e g m
43 sites
10-60 sq. mi. basins
Riffle/pool channels
Point sources avoided
Fitting these criteria, we were still able to get a range of urbanization (here using road density as a rough measure), although it is evident here (and became more so), that we needed to utilize basins from both areas to get the full range.
The pink are the Piemont basins and the green the Poconos.
You can tell there is substantial overlap, but you really need addition of the Poconos area to fill out the lower end, and vice versa…
Riemann and Riva-Murray, in process

7. Data collection elements
timing
Periphyton RTH
Late summer
Invertebrate RTH
Habitat & geomorphology
Nutrients, ions, field parameters
Late spring & late summer
Pesticides in water
Discharge (instantaneous)
Temperature
mid July-mid September
Contaminants in sediment & fish *
Once during
Mercury in sediment, fish, water*
Once during
*subset of sites

8. What stream ecosystem responses have we seen?
Chloride (mg/l)
Chloride
Pesticides
Nutrients
Sensitive
biota
Habitat
Again, just using roaddensity as a surrogate for urbanization…
Of all the things they looked at, which were more related to urbanization than any other source of variation… (natural or otherwise…)
EPT richness = # mayfly, stonefly, caddisfly species
(EPT richness: number of mayfly (Ephemeroptera), stonefly (Plecoptera), + caddisfly (Trichoptera) species
or other taxa)

9. Results - Detrended Correspondence Analysis
Light grey to black shading indicates percentage of invertebrate species that would be considered tolerant of pollution, habitat degradation, and other disturbance.
Note the increasing tolerance of the invertebrate community with increasing urbanization along Axis 1.
Again, just using roaddensity as a surrogate for urbanization…
Of all the things they looked at, which were more related to urbanization than any other source of variation… (natural or otherwise…)
EPT richness = # mayfly, stonefly, caddisfly species
(EPT richness: number of mayfly (Ephemeroptera), stonefly (Plecoptera), + caddisfly (Trichoptera) species
or other taxa)

10. What landscape variables are related to stream ecological responses?
from photo-interpretation
EPT
index
Habitat quality
Chloride conc.
Pesticide toxicity
Basin road, house, population density - +
Basin & buffer % urban, %imperv.
Basin % commercial/industrial
Buffer % commercial/industrial
Urban edge
Basin & buffer % forest
Forest aggregation index
Forest centroid connectivity
Note: not all variables come up as substantial for all measures of ecosystem response
Spearman rho >0.60, p<0.05

11. Algal Communities Trees increase shading which reduce algal growth
Nutrients shift algal community composition to more eutrophic types

12. Results – Examples of landscape variables that contribute to 2-variable model of disturbance gradient (based on Multiple Linear Regression using ranked data)
Percent impervious alone
(no additional variable)
Model R2 is 0.49
Variable added to model:
(i.e. If we increase ……….)
Model R2 becomes:
Invertebrate Disturbance P
Percent basin that is tree cover
0.78
decreases
<0.0001
Percent buffer that is grass cover
increases
Percent buffer that is tree cover
0.77
Percent basin that is grass cover
0.76
Percent basin that is agricultural
0.68
Aggregation index of forest patches
Percent of basin that is herbaceous wetland
0.64
0.0008
Area-normalized shared edge
0.63
0.002
Multiple linear regression results – 2 variable model, with forced inclusion of % impervious. That is, given a certain amt of impervious surface (i.e. given that urbanization is increasing), what characteristics tend to make things better (push to left along disturbance gradient), or worse (move to right along disturbance gradient)
Again, just using roaddensity as a surrogate for urbanization…
Of all the things they looked at, which were more related to urbanization than any other source of variation… (natural or otherwise…)
EPT richness = # mayfly, stonefly, caddisfly species
(EPT richness: number of mayfly (Ephemeroptera), stonefly (Plecoptera), + caddisfly (Trichoptera) species
or other taxa)

13. 2 variable model improves prediction

14. Results – Principal Components Analysis - chemistry
Analyte
Ordination axes
Spring base flow
Summer base flow I
(0.29) II
(0.22)
III
(0.21)
(0.24)
Silica
0.83
0.18
-0.08
0.91
0.23
0.03
Chloride
0.21
0.87
0.12
0.19
0.35
0.74
Nitrate plus Nitrite
0.85
0.30
-0.03
0.81
0.36
Boron
0.11
0.75
0.09
Dissolved organic carbon
-0.11
0.07
0.94
-0.19
Ammonium + total organic nitrogen
0.10
0.72
0.17
Dissolved phosphorous
0.38
0.24
0.34
0.66
0.15
0.53
Pesticide Toxicity Index (invertebrates)
0.63
0.39
0.56
0.41
0.70
0.42
Total pesticide concentration
0.77
0.51
0.80
0.14
Alkalinity
0.40
0.06
0.76
0.50
Results of principal components analysis of chemical data from spring and from late summer. The table shows loadings, or correlations, of original input variables along each of the first 3 axes from two separate PCA’s – one for each season. The bolded loadings show variables that were most important in contributing to each of these new ‘synthetic’ factors. The axes that are significantly correlated with the disturbance gradient (DCA axis) are in yellow.
Again, just using roaddensity as a surrogate for urbanization…
Of all the things they looked at, which were more related to urbanization than any other source of variation… (natural or otherwise…)
EPT richness = # mayfly, stonefly, caddisfly species
(EPT richness: number of mayfly (Ephemeroptera), stonefly (Plecoptera), + caddisfly (Trichoptera) species
or other taxa)

15. Summary Stream responses to urbanization include:
Invertebrate commmunity change; reduced # of sensitive taxa
Increased nutrients, chloride, pesticide concentrations
Decreased habitat quality
Potentially important types of landscape variables include:
Amount of forest or disturbed land in the watershed
Type of disturbed land in the watershed
Amount of forested buffer zone.
Patchiness of the landscape
So, not only is the overall amount of forest land important, but the type of landscape disturbance, the extent to which the stream corridor is disturbed or buffered, and the overall patchiness of the landscape.