Abstract

Initial landscape assessment efforts for the USEPA Environmental Monitoring and
Assessment Program's Western Pilot focused on improving and testing landscape
indicators previously developedfor the mid-Atlantic region of the United States.
Landscape indicators representing three general categories were developed for six study areas in the
western United States. These categories represented indicators derived from (1) single spatial data-
bases, (2) intersections of two or more spatial databases, and (3) models. Study areas were chosen
to represent a variety of ecosystems distinct from those in the mid-Atlantic region and encompassed
multiple watersheds. An intensive human use index was developed to assess the relative distribution
of dramatic anthropogenic land cover change caused by urbanization and agricultural development.
The presence of agriculture on steep slopes was used as an initial indicator of upland soil loss poten-
tial. Dasymetric mapping of population density was used to provide a more realistic representation
of human occupancy of the landscape than simple mapping of census data. Indicator results were
highly skewed with many values near zero. Several areas (Willamette Valley in northwestern
Oregon, viticultural areas in northern California, and the coastal plains of southern California) had
high scores for all indicators and were expected to have the most degraded aquatic resource condi-
tions. Needed improvements in the indicators presented are also discussed.

Overview of Landscape
Indicators

Landscape indicators are measures, indices of measures, or models describing the
condition of an ecosystem or one of its critical components (Hunsaker et al., 1990).
An indicator may reflect biological, chemical or physical attributes of ecological condition.
Indicators are used primarily to characterize status and to track or predict significant change.
Landscape indicators may also be used to identify significant ecosystem stress.

Landscape indicators are generated in a geographic information system (GIS) from spatial data
derived from satellite imagery and other data sources (e.g., elevation coverages generated from
dense field measures and models). Three general categories of landscape indicators can be used in
ecological assessments: (1) those generated from single spatial databases (e.g., percentage of water-
shed area that is forested generated from a digital land cover map); (2) those generated by intersect-
ing two or more spatial databases (e.g., the percentage of watershed area with agriculture on slopes
greater than 10% calculated by comparing a map of land cover to a map of slopes); and (3) those
generated from relatively simple models (e.g., potential nitrogen loss from a watershed calculated by
applying a nitrogen export model to a land cover map).

For the EMAP western pilot, we have created an initial list of landscape indicators, based primarily
on previous work in the mid-Atlantic region of the United States (Jones et al., 1997), and will evalu-
ate the ability of these indicators to explain variation in aquatic resource condition through a set of
watershed studies. Our initial landscape assessment efforts have focused on testing landscape indica-
tors developed for the mid-Atlantic region in six multiple watershed areas in the western U.S. These
study areas represent a variety of ecosystems distinct from those occurring in the Mid-Atlantic
region. We will also develop additional landscape indicators potentially related to aquatic resource
conditions. These indicators will emphasize the effects of mining, logging, grazing and population
growth. Conceptual models of watershed processes and functions will be used to establish hypothe-
ses of those landscape indicators that might be most important in explaining aquatic resource condi-
tions.

This poster presents results of representative landscape indictors from each of the three categories
described above. Each landscape indicator was calculated for the six study areas in the western
United States. Assessment units were 10-digit hydrologic unit code (HUC) basins for the
Northwestern Oregon, Northern California, Southern California and Lower Yellowstone study areas;
8-digit HUC drainage areas were used for the Southern Rockies and Colorado Plateau study areas.
The assessment units in Northern California do not meet the minimum size criteria for 10-digit HUC
basins. The large number of small assessment units in this area skews the frequency distribution of
indicator results and prevents the results to be meaningfully reported on a quantile basis. Indicator
results are therefore reported using logical breaks in the data.

Data Sources

A variety ofgeospatial data was used to derive the indicators for the six study
areas.

Comprehensive databases do not currently exist for many of the spatial data types used.
Although this could potentially complicate the comparison of indicator results across the study
areas, where possible, data were reclassified to facilitate comparability. As national data sets
become available in the next few years, all affected indicators will be recalculated to improve
comparability.

For the Northern California, Southern California, Lower Yellowstone, Southern Rockies and
Colorado Plateau study areas, land cover data were obtained from State Gap Analysis Programs
(GAP). The Northwestern Oregon study area utilized the Multi-Resolution Land Characteristics
Interagency Consortium's (MRLC) National Land Cover Database. Because States did not
always use the same criteria to classify higher level land cover categories, data were reclassified
into broad land cover categories for indicator development. GAP data will be replaced by
MRLC data as they become available.

U.S. Geological Survey (USGS) 30-m Digital Elevation Models (DEMs) were used for all study
areas except the Southern Rocky Mountain and Colorado Plateau where 90-m DEMs were used.
Population data (1990) were obtained from the U.S. Census Bureau.

Indicator Development for Landscape-Level Aquatic Ecological
Vu nerability Assessment in Western United States

K. BRUCE JONES1, DANIEL T. HEGGEM1, TIMOTHY G. WADE1, ANNE C. NEALE1, DONALD W. EBERT1, MALIHA S. NASH1, MEGAN H. MEHAFFEY1, KARL A. HERMANN2, ANTHONY R. SELLE2, SCOTT AUGUSTINE3,

IRIS A. GOODMAN1, JOEL A. PEDERSEN4, DAVID BOLGRIEN5, J. MAX VIGER6, DEAN CHIANG7, CINDY J. LIN4, YEHONG ZHONG8, JOAN BAKER9, AND RICK D. VAN REMORTEL10

'US Environmental Protection Agency, Las Vegas, NV 89193-3478, 2US Environmental Protection Agency, Region VIII Office, Denver, CO, 3US Environmental Protection Agency, Region X Office, Seattle, WA,

4US Environmental Protection Agency, Region IX Office, San Francisco, CA, 5US Environmental Protection Agency, Mid-Continent Ecology Division, Duluth, MN, 6ISSI CO, Inc., Denver, CO, 'Anteon Corporation, Monterey, CA,
sLoclcheed Martin, Port Orchard, WA, 9US Environmental Protection Agency, Western Ecology Division, Corvallis, OR '"Lockheed Martin, Las Vegas, NV

Intensive Human Use Index

Northern California
Study Area

Percent Area	

H0-1 L 1-5

~ 5

I 111

~ a

¦Is

Perhaps the simplest indicator of potential envimnmental impact is the extent that humans have converted
natural vegetation to urban or agriadture land cover. Such dramatic land cover conversion results in pro-
found changes in terrestrial and aquatic ecosystem structure and function.

Methodology

The proportion of watershed area with land cover corresponding to intensive human use was employed as an initial indicator
of potential anthropogenic impact on aquatic ecosystems. Land cover data were reclassified to nine land cover types (i.e.,
urban, cropland, pasture, rangeland, forest, wetland, barren, water and other). Only urban, cropland and pasture land cover
were expected to unambiguously correspond to intensive human use. The proportion of area having each of these land cover
types was summed and reported for each assessment unit. Grazing and forestry were explicitly excluded from this indicator
because these land uses do not result in permanent conversion to intensive human use. Both of these land uses involve short-
tenn disturbance of vegetative cover followed by recovery or replacement by a modified ecosystem. Furthermore, the land
cover classes in the GAP and MRLC databases did not distinguish between grazed and ungrazed rangeland (a category
including scrub-, shrub- and grassland), or logged and unlogged forests.

Results

The intensive human use index shows the relative distribution of dramatic anthropogenic land cover change in the study
areas. The table below summarizes the indicator results for the six study areas. Results are reported as the percentage of
assessment units or area having the designated percent intensive land use. In the Northwestern Oregon study area, the inten-

Agriculture on Steep Slopes

Northern California
1 Study Area

Southern California
Study Area

I 11 - 5 I 15-10 I 110-25

In the absence of effective soil conservation, agricultural practices can result in increased rates ofsoil
erosion, leading to increased sediment deposition in streams, lakes and estuaries. Potential soil loss fivm
agricultural areas is related to surface cover, slope steepness, slope length, rainfall emsivity, soil erodi-
bility and management practices (Hillel, 1998).

Methodology

As an initial indicator of upland soil erosion potential, the percentage of watershed area under intensive agricultural manage-
ment on steep slopes was calculated. This indicator represents a modified version of an indicator developed in the mid-
Atlantic region of the United States by Jones et al. (1997). Intensive agriculture was defined as land cover classified as row
crop, vineyard, orchard, pasture, small grains or fallow in the land cover vegetation databases. Grazing and timber harvesting
were not included in calculating the indicator. Percent slope was used as a measure of slope steepness, and slopes greater than
10% were considered to have increased erosion potential regardless of soil type. Slope was calculated from USGS 30-m
DEMs. The percentage of watershed area having intensive agriculture on steep slopes was calculated by intersecting areas of
intensive agriculture with areas having slopes greater than 10%. The output of this operation was then intersected with the
assessment unit polygons.

Colorado Plateau
Study Area

sive human use index was highest in the two regions dominated by
agricultural and urban land uses, the Willamette Valley and Deschutes
River basin. For Northern California, the highest values of the inten-
sive use index occurred in viticultural areas of Sonoma and Mendocino
counties, the Humbolt Bay area, and agricultural areas in the northeast-
ern part of the study area. In Southern California, the high proportion
of intensive human use was due to extensive urban and agricultural
development on the coastal plains. Little intensive human use occurred
in portions of the Transverse and Coastal ranges. The highest indicator
values for the Lower Yellowstone study area occurred in agricultural
areas in the west and north (52% and 55% percent of assessment unit
area). The Colorado Plateau had generally low values of the intensive use index (less than 8%). In the Southern Rockies, the
highest values for the index were in the eastern and southeastern portions of the study area. These regions correspond to agri-
cultural areas and the city of Evergreen.

Discussion

High index values correspond to areas that have experienced dramatic land cover conversion as a result of anthropogenic
activity. The aquatic resources in drainage basins having high index values are hypothesized to be degraded by pollutant
inputs and habitat alteration. This hypothesis will be tested in the indicator quantification phase of the project. Future indica-
tor development efforts will focus on assessing the extent and severity of impacts from forestry and grazing. This may
require the acquisition of imagery from more recently deployed sensors and the use of new image processing techniques.

Results

The percentage of assessment unit area with agriculture on slopes
greater than 10% provides an indication of upland soil erosion
potential. The results from this indicator are summarized in the
table below. Few assessment units had agriculture on steep slopes
over a large portion (> 25%) of their areas. Those that did were
located primarily in viticultural areas of Northern and Southern
California. The Willamette Valley also contained a number of
assessment units with fairly high proportions (10-14%) of their
area cultivated on slopes greater than 10%.

Agriculture on
Steep Slopes

(%)

Percent of
Assessment Units

Percent of
Area

<1

90%

87%

1 -4.9

4%

8%

5-9.9

2%

2%

10 - 24.9

2%

2%

25 - 49.9

1%

0.3%

50 - 75.3

1%

0.3%

Discussion

This indicator provides an indication of some of the lands most vulnerable to soil loss. However, many lands vulnerable to
soil erosion are not captured by this indicator. For example, depending on soil texture and structure, erosion may occur on
slopes less than 10%. Similarly, rainfall intensity, slope length and land use practices (e.g., irrigation methods, mulching,
cover crops) also affect soil erosion. The results of this indicator should therefore be viewed as minimum percentages of
potentially erodible areas. The focus on erosion from agricultural lands also precludes consideration of soil loss from grazed
and forested landscapes. The current indicator only considered sheet and rill erosion potential; channel erosion and mass
wasting (e.g., slumping) are not addressed. Future efforts will focus on improving estimates of upland soil erosion potential
by application of the Revised Universal Soil Loss Equation on a 10-digit HUC scale.

Intensive Human Use

Percent of

Percent of

(%)

Assessment Units

Area

<1

72%

35%

1 -4.9

6%

20%

5-9.9

4%

0%

10 - 24.9

6%

14%

25 - 49.9

5%

10%

50 - 74.9

4%

4%

75 - 100

3%

5%

Southern Rockies	Colorado Plateau

Study Area

Dasymetric Population Density

Northern California
Study Area

Southern California
Study Area

Capita/Square Kilometer

1H 0-1 mil -10 I 110-100 I 1100- 1.000

J 1,000 - 10,000

Population density can serve as a measure of human use intensity. Current spatial data-
bases of human population distribution (e.g., 1990 U.S. Census Bureau data) do not nec-
essarily relate to human use intensity. Alternative approaches suggested in the literature
often integrate remotely sensed imagery with data from the U.S. Census Bureau (Yuan et
al, 1997).

Because census data locate human population primarily in residential areas, actual spatial patterns of human use
are not accurately depicted when these data are used alone. For example, heavily used commercial and industrial
areas may have much greater daytime population densities than suburban residential areas.

Methodology

A dasymetric mapping approach was used to apportion human population to specific land cover types within
assessment units (Schumacher et al., 1999). This indicator incorporated human population from the 1990
U.S. Census Bureau data, land cover data reclassified into seven land cover types (i.e., urban, agriculture,
forest, wetland, barren, water and other), land ownership, and USGS DEMs. Block-group level population
counts were obtained from the 1990 U.S. Census Bureau. A sequence of filtering steps was applied to reas-
sign the population counted within each block group to smaller map units (Schumacher et al., 1999). Three
land cover types (i.e., water, wetland, barren) were assumed largely uninhabited and excluded from the
block groups. Publicly owned lands with the exception of U.S. Forest Service, U.S. Fish and Wildlife,
Bureau of Land Management, Bureau of Reclamation, and Agricultural Research Services lands were fur-
ther excluded from block groups, as were lands with slopes exceeding 30%. Population counts for each
block group were then redistributed into the four habitable land cover classes (i.e., urban, agriculture, forest,
and other) weighted by area and proportions reported by Schumacher et al. (1999) using the following
formula:

" / /

/v

t /

where, Pf is the population assigned to land cover i, Pj is the total population in the block group, vt»,- is the
weight assigned to land cover i, A; is the area of land cover i, and i — n, the maximum possible number of
land cover types. Assuming all four land cover classes were present in equal proportions in a block group
area, 80% of the population was allocated to urban areas, 5% to agricultural lands, 5% to forested lands and
10% to other. In reporting units lacking one or more of these land cover categories, these relative propor-
tions would be maintained. For example, if one block group contained only agriculture, forest and other,
both agriculture and forest received effective weights of 25%, while agriculture received an effective weight
of 50%.

Conclusions

Indicator results were highly skewed with many values near zero.

The Willamette Valley in northwestern Oregon, viticultural areas in northern California, and the coastal plains
of southern California had high scores for all indicators. These areas are expected to have the most degraded
aquatic resource conditions.

The low indicator scores for most assessment units in the Lower Yellowstone, Colorado Plateau and Southern
Rockies study areas appear to suggest few aquatic resource impacts. However, many potentially significant
stressors (e.g., grazing, forestry and mining) were not addressed by the indicators presented.

Assessing potential grazing and forestry impacts will require the use of additional remotely sensed imagery
and/or more sophisticated image processing techniques.

The indicator of upland erosion potential will benefit from a more sophisticated modeling approach.

Colorado Plateau

Population Density

Percent of

Percent of

(capita/km2)

Assessment Units

Area

0-0.9

40%

46%

1.0-9.9

39%

35%

10-99

12%

10%

100 - 999

6%

7%

1,000-9,999

2%

2%

Results

The dasymetric population density results show the rela-
tive distribution of human population density across the
six study areas. The table below summarizes these
results.

For the Northwestern Oregon study area, population den-
sity was highest in the Willamette Valley and Deschutes
River basin. Northern California was sparsely populated
with the highest population densities (capita/km2) occur-
ring in Sonoma County. Population density in Southern
California was concentrated in the metropolitan areas of Los Angeles and San Diego. The Lower
Yellowstone, Colorado Plateau and Southern Rockies study areas were sparsely populated with the highest
population density (10 capita/km2) occurring in the vicinity of Lame Deer in the Lower Yellowstone study

For the Northwestern Oregon study area, population density was highest in the Willamette Valley and
Deschutes River basin. Northern California was sparsely populated with the highest population densities
(capita/km2 ) occurring in Sonoma County. Population density in Southern California was concentrated in the
metropolitan areas of Los Angeles and San Diego. The Lower Yellowstone, Colorado Plateau and Southern
Rockies study areas were sparsely populated with the highest population density (10 capita/km2) occurring in
the vicinity of Lame Deer in the Lower Yellowstone study area.

Discussion

This indicator represents an improvement over the use of raw census data for displaying population distribu-
tion across the landscape. However, several modifications are required to improve the utility of this indica-
tor. The weighting factors utilized in this indicator were derived for use in western Montana, and may not
accurately reflect population distribution in other areas of the western U.S. (e.g., southern California). More
regional population distribution information is required to improve the accuracy of this indicator. In some
regions (e.g., highly urbanized southern California) block group areas were smaller than assessment units.
The dasymetric mapping of population in these cases resulted in inefficient data processing and no improve-
ment in the realism of population distribution. In such cases population distribution could be made more real-
istic by using larger census units (e.g., tracts) for dasymetric mapping. The algorithm employed also did not
account for inter-block group population transfer. This may be important in regions with significant metro-
politan areas.

References

Hillel, D. 1998. Environmental Soil Physics. Academic Press, San Diego, California, USA.

Hunsaker, C.T., R.L. Graham, G.W. Suter, R.V. O'Neill, L.W. Barthouse, and R.H. Gardner. 1990. Assessing
ecological risk on a regional scale. Environmental Management 14:325-332.

Jones, K.B., K.H. Riiters, J.D. Wickham, RD. Tankersley Jr., R.V. O'Neill, D.J. Chaloud, E.R. Smith, and
A.C. Neale. 1997. An Ecological Assessment of the United States Mid-Atlantic Region: A Landscape Atlas.
U.S. Environmental Protection Agency, Washington, D.C., EPA/600/R-94/130.

Schumacher, J.V., RL. Redmond, M.M. Hart, and M.E. Jensen. 1999. Modeling patterns of human use on
public lands in the western USA. In Proceedings of the Fourth Symposium on the Environmental Monitoring
and Assessment Program (in press).

Yuan, Y., R.M. Smith, and W.F. Limp. 1997. Remodeling census population with spatial information.
Computers, Enviromnent and Urban Systems 21: 245-258.

Western EMAP
Landscape
Assessment
Areas

Southern California
Study Area

Northwest Oregon
Study Area

Northwest Oregon
Study Area

Lower Yellowstone
Study Area

Southern Rockies
Study Area

Northwest Oregon
Study Area

Lower Yellowstone
Study Area

Southern Rockies
Study Area


-------