Allen D. Roberts and Stephen D. Prince

1
Effects of Urban and Non-Urban Land Cover on
Nitrogen and Phosphorus Runoff to Chesapeake Bay
Allen D. Roberts and Stephen D. Prince
Department of Geography University of Maryland
College Park, MD 20742
Introduction
Land cover and land use (LC/LU) and its
human-induced changes have large effects on water
quality of streams, rivers, lakes, and estuaries.
One such region where anthropogenic-related
LC/LU changes are said to have affected regional
water quality is the Chesapeake Bay (Figure 1).
Transformations in the 166,534 km2 watershed
dating back to the mid-1600s have contributed to
conditions, such as eutrophication and hypoxia,
that consequently affect the overall health of
this aquatic ecosystem. Fortunately,
watershed-wide, remotely sensed LC/LU data are
recently available for landscape analysis,
including the Regional Earth Science Application
Centers (RESAC) LC/LU, percent impervious
surface area ( ISA), and percent tree cover (
TC) maps for year 20001-4. Furthermore, water
quality models such as the Hydrologic Simulation
Program-FORTRAN (HSPF)5 and the SPAtially
Referenced Regressions On Watershed Attributes
(SPARROW)6 have been developed that can
incorporate Chesapeake Bay watershed-wide LC/LU
data7-10 in an effort to predict and quantify
nutrient loadings entering the Chesapeake Bay.
However, HSPF and other deterministic models
developed to monitor Chesapeake Bay water quality
are quite complex and based upon extensive data
and calibration requirements that can limit their
widespread application. The United States
Geological Surveys (USGS) SPARROW model is a
hybrid-statistical/deterministic non-linear
regression model suitable for simulating LC/LU
effects. The model functions by allowing spatial
referencing of nutrient sources and watershed
attributes (such as Chesapeake Bay urban and
non-urban LC/LU) to surface water flow paths
defined according to a digital drainage network
and constrained by mass-balance laws. Although
SPARROW has been parameterized for the watershed,
current LC/LU in Chesapeake Bay SPARROW models
find no correlation between LC/LU composition and
the land-to-water delivery of non-point nitrogen
(N) and phosphorus (P). In addition, existing
SPARROW, HSPF, and other Chesapeake Bay water
quality models do not consider landscape
configuration. This is an important new
component of this study. Finally, none of the
models consider the possible role of spatial
configuration, such as the enhanced importance of
LC/LU in riparian zones influencing nutrient
transport. Previous research has found that
inclusion of the configuration of LC/LU, such as
the patchiness of each type, edge to area ratio
and other metrics used in landscape ecology, and
restricting LC/LU to stream buffers can improve
nutrient runoff predictions over those that use
total catchment-wide LC/LU areal extent11. Seven
compositional and configurational LC/LU-based
landscape metrics that specify 1) contagion, 2)
area-weighted mean radius of gyration, 3) patch
number, 4) percentage of landscape, 5)
area-weighted mean patch size, 6) area-weighted
mean edge contrast, and 7) area-weighted mean
Euclidean nearest neighbor distance have been
shown to be significant indicators of downstream
water impairment in other areas12. The overall
purpose of this study was to determine LC/LU
effects on total nitrogen (TN) and total
phosphorus (TP) runoff to the Chesapeake Bay.
Figure. 3 Map of catchments zoomed around the
District of Columbia (D.C.)., southern Maryland
(MD), northern Virginia (VA), and the Potomac
River.
Figure 2 Map of the 2339 catchments and
segmented-river network reaches used here.
Results

• 31 m riparian stream buffers accounted for
  the greatest variation in mean annual TN and TP
  yield (Table 2).
• 5 of the 167 TN and 3 of the 168 TP metrics
  were shown to be either significant (p value
  0.05) descriptors of non-point sources or
  land-to-water delivery variables to streams
  draining the Chesapeake Bay (Table 3).

Model Run Model Yield R2 Model RMSE
B P TN (1997) 0.9073 0.2834
RESAC 31 m TN 0.9366 0.2407
RESAC 62 m TN 0.9332 0.2454
RESAC 125 m TN 0.9332 0.2454
RESAC 250 m TN 0.9332 0.2454
RESAC 500 m TN 0.9332 0.2454
RESAC 1000 m TN 0.9332 0.2454
B P TP (1997) 0.7413 0.3257
RESAC 31 m TP 0.7503 0.3126
RESAC 62 m TP 0.7457 0.3246
RESAC 125 m TP 0.7353 0.3329
RESAC 250 m TP 0.7262 0.3368
RESAC 500 m TP 0.7220 0.3393
RESAC 1000 m TP 0.7209 0.3400
Significant (p value 0.05) 31 m Model Landscape Metric Variables P Value
Area-Weighted Mean Urban ( 10 ISA) Patch Size (N source) 7.2 x 10-7
Percentage of Cropland (N land-to-water delivery) 1.1 x 10-4
Percentage of Extractive Land (N land-to-water delivery) 7.4 x 10-4
Area-Weighted Mean Edge Contrast of Deciduous Forest (N land-to-water delivery) 7.2 x 10-3
Percentage of Evergreen Forest Within the Riparian Stream Buffer (N land-to-water delivery) 4.7 x 10-2
Area-Weighted Mean Non-Agricultural/Non-Urban Patch Size (P source) 5.8 x 10-13
Percentage of Barren Land Within the Riparian Stream Buffer (P land-to-water delivery) 1.2 x 10-6
Area-Weighted Mean Urban ( 10 ISA) Patch Size (P source) 1.0 x 10-4
Methodology
Percent impervious surface area (ISA) data were
used to identify urban areas ( 10 ISA). In a
second LC/LU classification, pixels were assigned
to one of twelve non-urban classes. A third
classification distinguished forested (gt 50 tree
cover) from non-forested. Overall, a total of
sixteen classes were used (Table 1). For this
study, the structure of the Version 3.0
Chesapeake Bay models (Brakebill and Preston,
2004), (B P) was used so that the overall
topology of the stream network was unchanged and
B P non-LC/LU variables, such as point sources,
could be used. 2339 separate catchments
averaging 71 km2 were modeled based upon stream
reaches in a segmented-river network known as the
enhanced river reach file (E3RF1) (Figures
2-3)10. Total nitrogen (TN) and total phosphorus
(TP) in runoff models were calibrated using
estimated loadings collected from 87 (TN) and 104
(TP) stream sites. Landscape metrics were
calculated for each of the catchments at the
catchment and one of the six riparian buffer
width (31, 62, 125, 250, 500, and 1000 m) scales.
In all, 167 (TN) and 168 (TP) LC/LU class metric
combinations were tested in each model
calibration run.
Table. 2 Comparison of yield r2 and root mean
square error between the B P10 and six TN and
six TP models.
Table. 3 Comparison of the significant (p value
0.05) landscape metrics in the 31 meter (m)
total nitrogen (TN) and total phosphorus (TP)
models.

• 24.89 and 0.92 kilograms of N and P,
  respectively, were estimated to be generated
  annually from each hectare of area-weighted mean
  urban ( 10 ISA) patch size annually throughout
  the watershed.
• 0.11 kilograms of P was estimated to be
  generated annually from each hectare of
  area-weighted mean non-agricultural/non-urban
  patch size annually throughout the watershed.
• Catchment-wide extractive land percentage,
  cropland percentage and area-weighted mean edge
  contrast of deciduous forests in conjunction with
  riparian stream buffer-wide evergreen forest land
  percentage increased delivery of non-point N by
  27.0, 2.1, 1.4, and 1.3, respectively.
• Riparian stream buffer-wide barren land
  percentage increased delivery of non-point P by
  28.1.
• Largest delivered TN yields were seen in
  the lower Susquehanna Basin near Harrisburg,
  Lancaster, and York (Pennsylvania) the middle
  Potomac Basin in Maryland the eastern shore of
  Maryland and Delaware (Figure 4).
• Highest delivered TP yields were seen in the
  lower Susquehanna Basin near Lancaster
  (Pennsylvania) (Figure 5).

Cover Class Class Reference Number
I. Urban/Non-Urban Classification
Urban ( 10 impervious surface) 1
Non-Urban (lt 10 impervious surface) 2
II. RESAC LU/LC Classification
Urban/Residential/Recreational Grasses 3
Extractive 4
Barren 5
Deciduous Forest 6
Evergreen Forest 7
Mixed (Deciduous-Evergreen) Forest 8
Pasture/Hay 9
Croplands 10
Natural Grass 11
Deciduous Wooded Wetland 12
Evergreen Wooded Wetland 13
Emergent (Sedge-Herb) Wetland 14
III. Forested/Non-Forested Classification
Non-Forested ( 50 TC) 15
Forested (gt 50 TC) 16
Figure. 1 The Chesapeake Bay watershed with
locations of major urban centers.
Table. 1 The three Chesapeake Bay
watershed-wide map products divided into sixteen
land cover classes.
References

1. Goetz, S. J., R. Wright, A. J. Smith, E.
   Zinecker, and E. Schaub. 2003. Ikonos imagery
   for resource management tree cover, impervious
   surfaces and riparian buffer analyses in the
   mid-Atlantic region. Remote Sensing of
   Environment 88 195-208.
2. Goetz S. J., C. A. Jantz, S. D. Prince, A. J.
   Smith, R. Wright, and D. Varlyguin. 2004.
   Integrated analysis of ecosystem interactions
   with land use change the Chesapeake Bay
   watershed. Ecosystems and Land Use Change, pp.
   263-275. Eds. R. S. DeFries, G. P. Asner and R.
   A. Houghton. Geophysical Monograph Series,
   American Geophysical Union, Washington, D.C.,
   USA.
3. Goetz, S. J., D. Varlyguin, A. J. Smith, R. K.
   Wright, C. Jantz, J. Tringe, S. D. Prince, M. E
   Mazzacato, and B. Melchoir. 2004. Application of
   multitemporal Landsat data to map and monitor
   land cover and land use change in the Chesapeake
   Bay watershed.
4. Jantz, P., S. Goetz, and C. Jantz. 2005.
   Urbanization and loss of resources land in the
   Chesapeake Bay watershed. Environmental
   Management 36(6) 805-825.
5. Bicknell, B. R., J. C. Imhoff, J. L. Kittle, Jr.,
   T. H. Jobes, and A. S. Donigian, Jr. 2001.
   Hydrological Simulation Program-Fortran (HSPF).
   Users Manual for Release 12. U.S. Environmental
   Protection Agency, National Environmental
   Research Laboratory, Ecosystems Research
   Division, Athens, GA, in cooperation with U.S.
   Geological Survey, Water Resources Division,
   Reston, VA.
6. Smith, R. A., G. E. Schwarz, and R. B. Alexander.
   1997. Regional interpretation of water-quality
   monitoring data. Water Resources Research
   33(12) 27812798.
7. Preston, S. D. and J. W. Brakebill. 1999.
   Application of spatially referenced regression
   modeling for the evaluation of total nitrogen
   loading in the Chesapeake Bay Watershed. U.S.
   Geological Survey Water-Resources Investigations
   Report 99-4054.
8. Linker, L. C., G. W. Shenk, R. L. Dennis, and J.
   S. Sweeney. 2000. Cross-media models of the
   Chesapeake Bay watershed and airshed. Water
   Quality and Ecosystems Modeling 1(1-4) 91-122.
9. Brakebill, J. W., S. D. Preston, and S. K.
   Martucci. 2001. Digital data used to relate
   nutrient inputs to water quality in the
   Chesapeake Bay, Version 2.0. U.S. Geological
   Survey Open-File Report OFR-01-251.
10. Brakebill, J. W. and S. D. Preston. 2004.
    Digital data used to relate nutrient inputs to
    water quality in the Chesapeake Bay, Version 3.0.
    U.S. Geological Survey Open-File Report
    OFR-2004-1433.
11. Carle, M. V., P. N. Halpin, and C. A. Stow.
    2005. Patterns of watershed urbanization and
    impacts on water quality. Journal of the
    American Water Resources Association 41(3)
    693-708.
12. Leitao, A. B., J. Miller, J. Ahern, and K.
    McGarigal. 2006. Measuring Landscapes A
    Planners Handbook. Island Press, Washington,
    District of Columbia, USA.

Figure. 5 Per catchment estimated Chesapeake
Bay delivered total phosphorus (TP) yield in
kilograms per hectare per year (kg/ha/yr).
Figure. 4 Per catchment estimated Chesapeake
Bay delivered total nitrogen (TN) yield in
kilograms per hectare per year (kg/ha/yr).

• TN annual loadings (kg/yr) estimated from
  the 31 m model were 1.449 x 108 as compared to
  1.480 x 108, 1.420 x 108, and 1.292 x 108
  calculated from the respective B P, mean
  1985-1994, and year 2000 HSPF runs.
• TP annual loadings (kg/yr) estimated from
  the 31 m model were 5.367 x 106 as compared to
  5.210 x 106, 9.991 x 106, and 8.673 x 106
  calculated from the respective B P, mean
  1985-1994, and year 2000 HSPF runs.
• TP findings are lower than the loadings
  estimated by the HSPF model, perhaps because
  SPARROW is not event-based and cannot simulate P
  transport by sediments in episodic storm events.

This work was partly supported by a grant from
the NASA News program to Dr. Prince and a
fellowship from the University of Maryland to
Allen Roberts.
Acknowledgement
Conclusions

• Effects of LC/LU in riparian stream buffers
  were observed on TN and TP runoff at a much finer
  scale (smaller streams and localized scales) than
  before. Thus, riparian stream buffer LC/LU
  should be used in future data-driven water
  quality simulations representative of the entire
  Chesapeake Bay TN and TP watershed runoff.
• All newly significant land-to-water delivery
  variables at catchment and riparian stream
  buffer-wide scales increased non-point N and P
  delivery to the Chesapeake Bay as a result of
  higher overland and shallow subsurface flow
  generated from decreased hydraulic conductivity
  associated with the surface properties of these
  particular LC/LU classes or surrounding LC/LU
  classes.
• Largest delivered annual TN and TP yields
  (kg/ha/yr) were augmented from non-point
  agricultural (manure and fertilizer application)
  and urban (impervious wash off) sources of N and
  P that were increasingly transported to streams
  draining the Chesapeake Bay when spatially
  located in catchments having greater influences
  of new land-to-water delivery variables.
• Catchments far from the Chesapeake Bay may
  deliver high annual TN and TP yields (kg/ha/yr)
  to the estuary as a result of non-point N and P
  losses decreasing with addition into larger
  streams.
• The use of riparian stream buffer LC/LU
  composition improved the precision of simulated
  annual TN and TP loading estimates reaching the
  Chesapeake Bay.
• These findings are relevant not only to the
  Chesapeake Bay, but in similar watersheds
  elsewhere.