Landscape Analysis and Characterization to Support Region Environmental Assessment (LACRA)
Principal Investigators: Edward A. Martinko, Ph.D.2, Jerry L. Whistler, M.A.2, Frank deNoyelles, Jr., Ph.D.2, Jerry A. Griffith, Ph.D.2, Dana Peterson, M.A.2, Lyle Cowles1 (Project Coordinator), Don Miller, Ph.D.1 (Project Statistician), and Bruce Jones, Ph.D.1 (Project Officer)
- US Environmental Protection Agency, Region VII
- The University of Kansas
The EPA Region VII R-EMAP project seeks to determine the condition of fisheries resources in the states of Missouri, Kansas, and Nebraska. Preliminary evidence from fish tissue studies has suggested that fisheries resources in these states have been adversely impacted by chemical pollutants. The Region VII R-EMAP study consequently focuses on the following questions: 1) What is the status of the health of fisheries in Region VII?, and 2) What associations exist between selected indicators of natural and anthropogenic stresses and indicators of fisheries health? To help answer these questions, Region VII conducted systematic field sampling based on a set of randomly selected hexagons using EMAP ecological indicator protocols. Under the LACRA study, the University of Kansas developed various landscape components, including composition and pattern, that were derived from several remote sensing sources and used to characterize area sampling units. A time-series of normalized difference vegetation index (NDVI) values derived from satellite data was also developed to characterize area sampling units. Correlation and regression analyses were used to find statistical relationships between the field data and the landscape components and NDVI data.
Two hundred ninety randomly selected stream sites in Nebraska, Kansas, and Missouri were sampled once during late spring through summer of 1994 or 1995 for several parameters including: conductivity, turbidity, total phosphorus, nitrate-nitrite nitrogen, the Index of Biotic Integrity (IBI), and a habitat index (HI). Based on landscape data from watersheds that were delineated for each sampling location, interrelationships were examined between these water quality parameters and land use/land cover, the normalized difference vegetation index (NDVI) derived from AVHRR satellite data, and vegetation phenological metrics (VPMs) derived from time-series NDVI data. Statistically significant relationships were found between NDVI values and derived VPMs with the stream condition parameters (r values to 0.8, a = 0.05). The NDVI or VPMs were more highly correlated to the selected stream condition parameters than were the land use/land cover proportions. Knowledge of the general land use/land cover setting within the watersheds, however, was important for interpreting these relationships. The most common variables associated with the stream data were early spring NDVI values or VPMs associated with the date of onset of greenness. This finding suggests the possibility of forecasting summer stream conditions with spring AVHRR satellite data. Maps depicting the distribution of three sample parameters were developed that illustrate the use of these results. Such maps may be useful for targeting limited resources for further study or remediation.
In addition to NDVI-water quality relationships, the relationships between the selected water quality parameters and landscape pattern metrics (LPMs) were also investigated. The relationships between two indices, the IBI and the HI, and NDVI and VPMs were also explored. FRAGSTATS software was used to calculate landscape pattern metrics. Despite the limitations of LPMs found in this study, a combination of the interspersion and juxtaposition index, patch density, and percent forest within the watershed explained greater than 60% of the variation in levels of nitrate-nitrite nitrogen and conductivity in the Ozark Highlands. There were also strong correlations between patch density and the Habitat Index in the Mississippi River Lowlands, and between landscape diversity and conductivity in the Nebraska Sand Hills. In most cases, however, the landscape pattern metrics were not significantly correlated with the stream data. Several problems with using landscape pattern metrics were noted and include: small watersheds having only one or two patches, which prevented calculation of some LPMs; collinearity with landuse/land cover data; and counterintuitive or inconsistent results that resulted from basic differences in land use/land cover patterns among ecoregions or from other important factors that determine water quality. The amount of variation explained in water quality parameters using multiple regression models that combined land use/land cover and LPMs was generally lower than that from NDVI or VPMs. A comparison of LPMs and NDVI indicated that NDVI had greater promise for monitoring landscapes for regional water quality within the study area. Major conclusions derived from this work are (1) NDVI or VPMs better explain variation in stream water quality conditions than either land use/land cover composition or pattern, (2) knowledge of the general land use/land cover setting within the watersheds is necessary to interpret the relationships, (3) stratifying the watersheds by ecoregion yielded stronger relationships between the field data and landscape data, and (4) these results become increasingly significant when one remembers that field sampling points were established using the EMAP sampling framework (i.e., randomly selected), and that other watershed influences, such as geology and watershed size, were not held constant.
EPA Assistance Agreement Final Report Completed December 2000. Conversion to EPA Report - In Progress
Landscapes can play an important role in determining water quality which in turn impacts the ecological health of systems dependent upon that water. It has been observed that landscapes and watersheds are large-scale systems that limit the dynamic processes of subsystems such as grasslands, forests, etc. (Kepner 1995). Furthermore, the amount and spatial arrangement of land cover types in the landscape are related to ecological processes and to the effect a specific stress may have on a resource within a region, such as fisheries health. Thus, landscape characterization is an essential component for determining the status and trends in the condition of ecological resources (Norton and Slonecker 1990).
Previous work has demonstrated that water quality is influenced by landscape composition and pattern (Hunsaker et al. 1992; Hunsaker et al. 1995). Landscape attributes have also been examined to determine their relative contribution to nonpoint source pollution (Haith 1976; Omernik 1977). Within Region VII, previous work has shown that the percent total cropland for watersheds is correlated with chemical water quality measurements (Meador 1990) and stream insect diversity (Anderson 1990). Typically, these studies have been limited in scope to small and medium watersheds (<10,000 hectares) and have utilized land use and land cover information from remote sensing sources, either aerial photographs or satellite imagery.
The normalized difference vegetation index (NDVI) is a commonly used vegetation index in studies utilizing remote sensing data because it is roughly correlated with green plant biomass. The NDVI is based on the relative reflectance values in the red and near infrared (NIR) wavelengths:
NDVI = (NIR - Red)/(NIR + Red)
Vegetation indices are commonly used to reduce effects of atmospheric conditions or different soil backgrounds on spectral reflectance values. The amount of red solar energy reflected by vegetation cover depends primarily on chlorophyll content, whereas the amount of near infrared energy reflected by vegetation is affected by the amount and condition of green biomass, leaf tissue structure, and water content (Jensen 1996).
Evaluating the potential of NDVI and NDVI-derived metrics for watershed monitoring and water quality studies is important in gaining an increased understanding of landscape-water quality relationships. NDVI has been shown to be sensitive to biophysical characteristics of vegetation such as leaf area, net primary production and levels of photosynthetic activity (Stoms and Hargrove 2000; Paruelo and Lauenroth 1995; Spanner et al. 1990; Box et al. 1989; Goward et al. 1985; Tucker et al. 1983), as well as LULC (Loveland et al. 1995). Because of its ability to integrate both land cover and biophysical conditions, NDVI can be helpful in assessing regional watershed conditions that affect water quality and stream condition. The NDVI has become a standard form of band ratio for vegetation indices and is widely used (Lauver and Whistler 1993). Other forms of ratios have been used (Goetz et al. 1975, Kriegler et al. 1969), but have been shown to be functionally equivalent (Perry and Lautenschlager 1984).
Jones et al. (1996) evaluated the potential of NDVI to assess watershed health and hypothesized it could indicate losses in productivity, increased erosion, and losses of the buffer capacity along riparian corridors. They suggested examining NDVI patterns and change, as well as comparing observed versus expected NDVI based on soils, topography, vegetation and climate. Whistler (1996) explored NDVI values derived from Landsat Multi-Spectral Scanner (MSS) imagery as a surrogate for biomass, and hypothesized that NDVI values would have stronger relationships with water chemistry parameters than with land cover proportions derived from the same imagery. He found significant relationships between NDVI and selected water quality parameters which, in fact, were stronger than relationships to LULC in most cases.
In addition to raw NDVI values, phenological metrics derived from NDVI may also have potential for watershed assessment. Vegetation phenological events such as emergence, maturity, and senescence are important for assessing the condition of agricultural vegetation (Lee 1999; Samson 1993). Reed et al. (1994) defined 12 metrics using AVHRR NDVI bi-weekly composites that can be categorized into three groups:1) temporal (based on the timing of a phenological event), 2) NDVI-based (the NDVI value at the time of a phenological event), and 3) metrics derived from time-series characteristics. These vegetation phenological metrics (VPMs) have been successfully used to assess crop condition and potential yield (Lee 1999), characterize crop phenological variability (Reed et al. 1994), separate grasslands by photosynthetic pathway (C3 or C4) (Tieszen et al. 1997), and map LULC (Loveland et al. 1995). While previous work by Whistler (1996) has shown the advantage of using NDVI over land cover proportions for relating landscape variables to water quality, the utility of VPMs for applications in water quality monitoring has yet to be fully explored.
The primary objective of this study is to analyze, on an ecoregion and watershed basis, landscape composition and pattern metrics derived from different remote sensing sources (e.g., AVHRR and Landsat TM) and evaluate the utility of these data for use in the R-EMAP fisheries project. This is achieved primarily through the use of correlation and regression analysis between the REMAP field data and various landscape metrics. Another objective is to test and evaluate the ability of different remote sensing data sources to provide a set of landscape metrics that can be used as analytical tools to support regional environmental assessments.
Within these overarching objectives are the specific goals to:
- expand the use of land use/land cover (LULC) composition and pattern
metrics to examine existing environmental conditions at regional scales;
- examine the suitability of existing sources of regional land use/land
cover data sets that differ in their age, source data (i.e., aerial
or satellite), method of compilation (i.e., manual or digital), attribute
resolution, and spatial resolution for use in R-EMAP and other regional
- introduce the use of a vegetation index derived from digital satellite
data to characterize landscapes. Vegetation indices are highly indicative
of vegetation properties (e.g., leaf area, green biomass) and have been
shown to better explain variation in stream water chemistry than summary
area statistics for land use/land cover (Whistler 1996);
- demonstrate a multi-temporal characterization of landscapes using
several dates of a vegetation index that reflects the dynamic nature
of land cover over the course of a growing season;
- expand the characterization of landscapes using a coarse resolution
data set to include metrics that are defined by the spatial arrangement,
or pattern, of land cover types.
Although EPA Region VII includes the four states of Iowa, Kansas, Missouri, and Nebraska, field data for Iowa were not collected. Commonly perceived as a homogenous area, the landscapes of Nebraska, Kansas and Missouri are surprisingly varied. Geology in the area consists of limestones and shales in central and eastern Kansas originating from shallow Paleozoic seas. The Nebraskan and Kansan glaciations deposited glacial drift across northern Missouri, eastern Nebraska, and northeastern Kansas. The Precambrian strata of the Ozark Uplands remained a non-glaciated area with steeper and more rugged terrain. Loess soils cover much of Nebraska, while stream sediments from the Rocky Mountains cover the western edges of Kansas and Nebraska (Williams and Murfield 1977).
Precipitation ranges from 38 - 45 cm in westernmost Kansas and Nebraska, to 90 - 100 cm in eastern Kansas, to nearly 120 cm on the Mississippi River in southeastern Missouri (Goodin et al. 1995, Schroeder 1982). Native vegetation consists of shortgrass prairie in westernmost Kansas and Nebraska, tallgrass and mixed-grass prairie in the Nebraska Sand Hills and central Kansas, a mosaic of bluestem prairie and oak-hickory forest in eastern Kansas and northern Missouri, and dense oak-hickory forests in the Ozark Highlands. The central human transformation of the Great Plains region has been conversion of grassland to cropland. Currently, 90% of the area is in farms or ranches and 75% of the land is in cultivation (Riebsame 1990). Hydrological impacts stem from tillage, cropping, runoff change, water impoundment, groundwater depletion, and changes in soil structure and chemistry.
Water quality and stream condition data were collected by EPA Region VII during the late spring and summer of either 1994 or 1995 (streams were sampled once) as part of its R-EMAP Project. Two hundred ninety stream sites were randomly selected in Kansas, Nebraska and Missouri to assess fisheries health and stream condition, as well as to establish baseline data and methods usable for assessing long-term trends throughout the region (EPA 1997). Of the more than 30 water quality parameters examined, four water quality parameters that are important determinants of water quality and which integrate across the entire watershed were selected for analysis in the LACRA study: total phosphorus (TP), nitrate-nitrite nitrogen (NO2NO3), turbidity, and conductivity. In addition, the Index of Biotic Integrity (IBI) and a Habitat Index (HI) were examined. Analytical techniques used to determine the values for the water quality parameters and methods for calculating the IBI and HI are detailed in EPA (1997). Table 1 lists the variables used in calculating the IBI and HI.
The LACRA study evaluates four existing spatial data sets derived from three different remote sensing systems for their ability to provide meaningful landscape composition and pattern metrics. The remote sensing systems are NOAA Advanced Very High Resolution Radiometer (AVHRR), Landsat Thematic Mapper (TM), and aerial photography. The criteria determining selection of these existing data sources were, in order of importance, availability, universal coverage for Region VII, was the data set contemporaneous with the field data, spatial resolution, and the availability of multi-temporal coverage. None of the selected data sets met all five criteria.
Agricultural Land versus Cropland - Many different classification schemes for land use and land cover have been utilized by agencies at various levels of government that has often resulted in collection of data that is incompatible. In the mid-70s, the U.S. Geological Survey (USGS) proposed a classification system for use with remote sensor data that would standardize classification, yet would be flexible enough to allow various levels of class detail (Anderson et al. 1976). While most recent land use/cover mapping efforts utilize this scheme, classification incompatibilities still exist due to different mapping objectives and goals established by the agency collecting the data.
This study utilizes data sets that variously define the categories agriculture, rangeland, cropland, and grassland. For example, the category agriculture is often defined as including cropland and pasture. This definition is focused more on land use. On the other hand, pasture is also included in other classification schemes as the category grassland, a definition that focuses more on land cover. For this study, the impact of land cover is given more value. Therefore, whenever possible, all tiled land is categorized as cropland, while pasture and rangeland are categorized as grassland.
1:250,000 Land Use and Land Cover (USGS LULC) - Advantages of this data set are its availability and universal coverage. The USGS LULC data were compiled under the Land Use Data and Analysis (LUDA) Program of the U.S. Geological Survey in the 1970s and 80s (USGS 1986). The data set contains 37 land cover types using a Anderson Level II classification system (Anderson et al. 1976). The age of the data varies considerably (15-25 years old) depending on the acquisition date of the photography. Although the USGS dataset is dated and has coarser resolution than other datasets derived from Landsat TM or MSS, Herlihy et al. (1998) found no major differences in LULC-stream water chemistry relationships when using either the USGS dataset or more recent Landsat TM data.
The USGS LULC data are available in both vector and raster digital format and were obtained off of the USGS Global Land Information System (GLIS) web site. This study uses the Composite Theme Grid (CTG) raster format that has a grid-cell resolution of 200 meters (4 ha). The CTG formatted files (ASCII x,y pairs) were initially imported as point coverages using the ArcInfo Generate command and then converted to an ArcInfo grid format. The individual grids were merged and projected from UTM to Albers Equal Area using EPA national projection parameters. The composite grid was then clipped to the four-state region.
A final step was to recode the Level II data to Level I general land cover classes to facilitate later comparison with other data sets. The Level I classes are urban, cropland, grassland, woodland, water, and barren.
AVHRR Level II Land Use and Land Cover (AVHRR LULC) - The advantages of this data are availability, universal coverage, and it is relatively contemporaneous with the field data. The AVHRR LULC data contain 159 land cover classes using a modified Level III classification system (Loveland et al. 1991). This raster data set was digitally classified and compiled from AVHRR satellite imagery collected between March and October of 1990. AVHRR data have a 1km spatial resolution and therefore the minimum size area depicted with this data is 100 ha. AVHRR LULC data were read from CD and converted into an ArcInfo grid format. The file was then projected from Lambert Azimuthal Equal Area to Albers Equal Area using EPA national projection parameters. The conterminous US data set was clipped to the four-state region. This data set was also recoded to Level I general land cover classes.
Kansas Land Cover Patterns (Kansas LULC) - The advantages of this data set are availability, fine spatial resolution, and it is relatively contemporaneous with the field data. The Kansas LULC database contains 10 land use/land cover types based on a modified Anderson Level I classification (Whistler et al. 1995). This data set was compiled from Thematic Mapper (TM) 30-meter data collected during the growing season between 1988 and 1990 for the state of Kansas. It depicts areas with a minimum size of 0.5 ha and thus provides finer spatial detail than either the USGS LULC or the AVHRR LULC databases.
Although derived from a raster source, the Kansas LULC data set is archived as a vector coverage with geographic coordinates. The vector coverage was projected Albers Equal Area using EPA national projection parameters and was then converted to an ArcInfo grid format. To achieve the general Level I classification, the five urban classes were recoded into one urban class.
AVHRR Normalized Difference Vegetation Index (NDVI) - The advantages of this data set are its availability, universal coverage, nearly concurrent acquisition date with the field data, and multitemporal coverage. Twenty-six bi-weekly periods for 1995 were selected for study based on their proximity in time to the R-EMAP sample data, which was collected in 1994 and 1995. As noted, AVHRR data have a resolution of 1km. Each composite is composed of the maximum NDVI value for every 1 km x 1 km pixel over a two-week period (Eidenshink 1992). This data were originally downloaded from the USGS EROS Data Center. The data were projected from their native projection, Lambert Conformal, to Albers Equal Area using EPA national parameters.
Vegetation Phenological Metrics (VPMs) - A temporal characterization of vegetation based on the 26 bi-weekly NDVI imagery was constructed for the 1995 growing season. A series of derived metrics describing vegetation phenology were developed using algorithms modified from Reed et al. (1994). Figure 1 shows the basis for their calculation and Table 2 lists the specific metrics used in this study. Loveland et al. (1995) used these VPMs to help classify LULC in the conterminous U.S.
Ancillary Data Sets - EPA Region VII designed the R-EMAP project field data to be examined at three scales of analysis; the watershed, the ecoregion, and the state. To accomplish this, four ancillary data sets were acquired or created. These data sets were used as "cookie cutters" to extract summary statistics from the landscape data sets described above. These data sets are: (1) R-EMAP sampling point watershed boundaries, (2) USGS 8-digit hydrologic unit code (HUC8) boundaries, (3) EPA level III ecoregion boundary data (Omernik 1987), and (4) USGS 1:100,000-scale state boundary data.
R-EMAP Sampling Points Watersheds - Watersheds were delineated and digitized for the REMAP sampling points. For larger watersheds, lower portions of the watershed were manually delineated and the remaining upper sections completed using digital 8-digit hydrologic unit code (HUC8) boundary files. The general procedure for watershed delineation was to plot HUC8 boundaries and the sampling points for 1:100,000 quadrangle regions on white paper. We manually overlaid these maps on the corresponding 1:100,000 USGS topographic maps. For clarity, we transferred the plotted sampling point location to the USGS map. Watershed boundaries were interpreted, usually from 10 m contour interval linework, and then drawn on the USGS maps. The drawn watersheds were then digitized.
Hydrologic Unit Code (HUC) Boundaries - HUCs are spatial units, used by the USGS, NRCS, and EPA, that generally correspond to segments of river basins. 8-digit HUCs were downloaded for Nebraska and Iowa, and 11-digit HUCs were downloaded for Kansas and Missouri from their respective state web sites. The HUC files were then imported into ArcInfo. The Kansas and Missouri HUCs were dissolved into 8-digit HUC coverages. The Kansas HUC coverage was projected from Lambert Conformal Conic to Albers Equal Area while Missouri, Iowa, and Nebraska were projected from UTM to Albers Equal Area, using EPA national projection parameters. The polygon attribute tables were modified (standardized) and the HUC coverages were then appended. The state boundaries were removed and the line work for the HUC boundaries was modified at the state boundaries to achieve a smooth transition between states. The state boundaries from 1:100,000 DLGs were added and the coverage was checked for label and topological errors. A look-up table was created and joined to the polygon attribute table to relate HUC identification numbers to HUC code numbers.
Ecoregions - Level III ecoregions were downloaded from the EPA web site and imported into an ArcInfo coverage. The coverage was clipped to the four-state region and label errors were corrected. No further processing was necessary.
State Boundaries - 1:100,000-scale DLGs containing political boundaries were downloaded from the USGS GLIS web site and imported into ArcInfo coverages. Approximately 240 files were viewed and edited, removing all political data except the state boundary data. Quads not available online were manually digitized from 1:100 000 quad sheets. The individual boundary files were appended, cleaned, labels added, and projected from UTM to Albers Equal Area using EPA national projection parameters.
ArcView version 3.1 was used to extract summary statistics from the landscape data. For each ancillary data set, GIS overlays were used to extract LULC proportions and to calculate mean NDVI and VPM values for each bi-weekly period, as well as standard deviation values of the VPMs. The U-index (human Use index), which equals the proportion of cropland plus the proportion of urban land, was also calculated and has been used to gauge the level of total anthropogenic disturbance in regional landscapes (EPA 1994). The extracted landscape data and the R-EMAP field data were then imported into SPSS. Correlation and regression analysis was conducted using SPSS version 9.0.
Allan, J.D. 1995. Stream Ecology: Structure and Function of Running Waters. Chapman and Hall, London.
Anderson, J.R., E.E. Hardy, J.T. Roach, and R.E. Witmer. 1976. A land use and land cover classification system for use with remote sensor data. Geological Survey Professional Paper 964. U.S. Government Printing Office, Washington, D.C.
Anderson, T.M. 1990. The relationship between basin cultivation and stream insect composition in small agricultural streams in Kansas. M.S. Thesis, University of Kansas, Lawrence.
Box, E., B. Holben, and V. Kalb. 1989. Accuracy of the AVHRR vegetation index as a predictor of biomass, primary productivity and net CO2 flux. Vegetatio 80:71-89.
Carpenter, S., N. Caraco, D. Correll, R. Howarth, A. Sharpley, and V. Smith. 1998. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological Applications 8(3):559-568.
Drever, J. 1997. The Geochemistry of Natural Waters. Prentice-Hall, Inc. Upper Saddle River, NJ. 436 pp.
Eidenshink, J. 1992. The 1990 AVHRR conterminous U.S. data set. Photogrammetric Engineering and Remote Sensing 58:809-813.
EPA. 1994. Landscape Monitoring and Assessment Research Plan. EPA 620/R-94/009. Environmental Monitoring Systems Laboratory, Las Vegas, NV. 64 pp.
EPA. 1997. Quality assurance project plan for measuring the health of the fisheries in EPA Region VII. U.S. Environmental Protection Agency. Environmental Services Division. Kansas City, KS. 80 pp.
GAO. 2000. Water Quality: Key EPA and State Decisions Limited by Inconsistent and Incomplete Data. U.S. General Accounting Office. GAO/RCED-00-54. Washington, DC. 76 pp.
Goodin, D., J. Mitchell, M. Knapp, and R. Bivens. 1995. Climate and Weather Atlas of Kansas: An Introduction. Educational Series 12. Kansas Geological Survey, Lawrence, Kansas.
Goetz, A., F. Billingsley, A. Gillespie, M. Abrams, R. Squires, E. Shoemaker, I. Lucchitta, and D. Elston. 1975. Application of ERTS images and image processing to regional geologic problems and geologic mapping in northern Arizona. JPL Technical Report 32-1597.
Goward, S., C. Tucker, and D. Dye. 1985. North American vegetation patterns observed with the NOAA-7 Advanced Very High Resolution Radiometer. Vegetatio 64:3-14.
Griffith, J. 2000. Interrelationships among landscapes, NDVI and stream water quality in the U.S. Central Plains. PhD. Dissertation, Department of Geography. University of Kansas. Lawrence, KS. 222 pp.
Haith, D.A. 1976. Land use and water quality in New York rivers. Journal of Environmental Engineering. ASCE 102:1-15.
Herlihy, A., J. Stoddard, and C. Johnson. 1998. The relationship between stream chemistry and watershed land cover data in the Mid-Atlantic Region. Water, Air, and Soil Pollution 105(½):377-386.
Hunsaker, C.T., D.A. Levine, S.P. Timmins, B.L. Jackson, and R.V. O'Neill. 1992. Landscape characterization for assessing regional water quality. In: D.H. McKenzie, D.E. Hyatt, and V.J. McDonald (eds.), Ecological Indicators. Pp. 997-1008. Elsevier Appl. Sci., New York.
Hunsaker, C.T., and D.A. Levine. 1995. Hierarchical approaches to the study of water quality in rivers. Bioscience 45(3):193-203.
Jelinski, D., and J. Wu. 1996. The modifiable areal unit problem and implications for landscape ecology. Landscape Ecology 11(3):129-140.
Jensen, J. 1996. Introductory Digital Image Processing. Prentice Hall, Englewood Cliffs, NJ. 316 pp.
Jones, K.B., K. Riitters, J. Wickham, R. Tankersly Jr., R. O'Neill, D. Chaloud, E. Smith, and A. Neale. 1997. An Ecological Assessment of the United States Mid-Atlantic Region: A Landscape Atlas. PA/600/R-97/130. Environmental Protection Agency. Office of Research and Development. Washington, DC. 104 pp.
Jones, K.B., J. Walker, K. Riitters, J. Wickham, and C. Nicoll. 1996. Indicators of landscape integrity. Pages 155-168. In: J. Walker and D. Reuter, eds. Indicators of Catchment Health. CSIRO Publishing, Melbourne, Australia. 178 pp.
Kepner, W.G., K.B. Jones, D.J. Chaloud, J.D. Wickham, K.H. Ritters, and R.V. O'Neill. 1995. Mid-Atlantic landscape indicators project plan. EPA/620/R-95/003, U.S. Environmental Protection Agency, Office of Research and Development, Research Triangle Park, NC.
Kriegler, F., W. Malila, R. Nalepka, and W. Richardson. 1969. Preprocessing transformations and their effects on multispectral recognition. Pages 97-131. In: Proceedings of the Eighth International Symposium on Remote Sensing of Environment.
Lauver, C., and J. Whistler. 1993. A hierarchical classification of Landsat TM imagery to identify natural grassland areas and rare species habitat. Photogrammetric Engineering & Remote Sensing 60(10):627-634.
Lee, R. 1999. Modeling corn yields in Iowa using time-series analysis of AVHRR data and vegetation phenological metrics. PhD. Dissertation. Department of Geography. University of Kansas. 164 pp.
Loveland, T.R., J.W. Merchant, D.O. Ohlen, and J.F. Brown. 1991. Development of a landcover characteristics database for the conterminous U.S. Photogrammetric Engineering and Remote Sensing. 58:6, pp. 809-813.
Loveland, T., J. Merchant, J. Brown, D. Ohlen, B. Reed, P. Olson, and J. Hutchinson. 1995. Seasonal land-cover regions of the United States. Annals of the Association of American Geographers 85(2):339-355.
McDaniel, T., C. Hunsaker, and J. Beauchamp. 1987. Determining regional water quality patterns and their ecological relationships. Environmental Management 11(4):507-518.
McGarigal, K., and B. Marks. 1995. FRAGSTATS: Spatial pattern analysis program for quantifying landscape structure. USDA Forest Service General Technical Report PNW-GTR-351, Pacific Northwest Research Station, Portland, OR. 122 pp.
Meador, S.S. 1990. A GIS assessment of landscape/water quality relationships in small Kansas streams impacted by rural nonpoint source pollution. M.S. Thesis, University of Kansas, Lawrence.
Norton, D.J. and E.T. Slonecker. 1990. The ecological geography of EMAP. GeoInfo Systems. 1:33-43.
Omernik, J. 1987. Ecoregions of the conterminous United States. Annals of the Association of American Geographers 77:118-125.
Omernik, J.M. 1977. Nonpoint source-stream nutrient level relationships: a nationwide study. EPA/600/3-77-105, U.S. Environmental Protection Agency.
Paruelo, J., and W. Lauenroth. 1995. Regional patterns of normalized difference vegetation index in North American shrublands and grasslands. Ecology 76:1888-1898.
Perry, C., L. Lautenschlager.1984. Functional equivalence of spectral vegetation indices. Remote Sensing of Environment 14:169-182.
Reed, B., J. Brown, D. VanderZee, T. Loveland, J. Merchant, and D. Ohlen. 1994. Measuring phenological variability from satellite imagery. Journal of Vegetation Science 5:703-714.
Riebsame, W. 1990. The United States Great Plains. Pages 561-576. In: B. Turner, and W. Meyer, eds. The Earth as Transformed by Human Action. Cambridge University Press, Cambridge, UK.
Samson, S. 1993. Two indices to characterize temporal patterns in the spectral response of vegetation. Photogrammetric Engineering & Remote Sensing 59(4):511-517.
Schlosser, I., and J. Karr. 1981. Water quality in agricultural catchments: impact of riparian vegetation during base flow. Water Resources Bulletin 17:233-240.
Schroeder, W. 1982. Missouri Water Atlas. Missouri Department of Conservation. Jefferson City, Missouri.
Spanner, M., L. Pierce, S. Running, and D. Peterson. 1990. The seasonality of AVHRR data of temperate coniferous forests: relationship with leaf area index. Remote Sensing of Environment 33:97-112.
Squillace, P., and E. Thurman. 1992. Herbicide transport in rivers: importance of hydrology and geochemistry in nonpoint-source contamination. Environmental Science & Technology 26(3):538-545.
Stevens, J. 1996. Applied Multivariate Statistics for the Social Sciences. 2nd Ed. Lawrence Erlbaum Associates, Hillsdale, NJ. 629 pp.
Stoms, D., and W. Hargrove. 2000. Potential NDVI as a baseline for monitoring ecosystem functioning. International Journal of Remote Sensing 21(2):401-407.
Tieszen, L., B. Reed., N. Bliss, B. Wylie, and D. Dejong. 1997. NDVI C3 and C4 production, and distributions in Great Plains grassland cover classes. Ecological Applications 7:59-78.
Tucker, C., C. Vanpraet, E. Boerwinkle, and A. Gaston. 1983. Satellite remote sensing of total dry matter production in the Senegalese Sahel. Remote Sensing of Environment 13:461-474.
Turner, M., R. O'Neill, R. Gardner, B. Milne. 1989. Effects of changing spatial scale on the analysis of landscape pattern. Landscape Ecology 3:153-162.
USGS. 1986. Land use and land cover digital data from 1:250,000- and 1:100,000-scale maps. Data users guide. National Mapping Program, Technical Instructions, Data Users Guide 4, U.S. Geological Survey, Reston, VA.
Wang, S. 1997. Nitrate dynamics of small agricultural streams in the Western Corn Belt Plains Ecoregion. Ph.D. dissertation. University of Kansas, Lawrence, KS. 298 pp.
Whistler, J. 1996. A Phenological Approach To Land Cover Characterization Using Landsat MSS Data For Analysis Of Nonpoint Source Pollution. KARS Report No. 96-1. Kansas Applied Remote Sensing Program, University of Kansas, Lawrence, KS. 51 pp.
Whistler, J., Egbert, S., Jakubauskas, M., Martinko, E., Baumgartner, D., and Lee, R., 1995. The Kansas state land cover mapping project: regional scale land use/land cover mapping using Landsat Thematic Mapper data. Proceedings of the ACSM/ASPRS 95 Annual Convention and Exposition, Charlotte, North Carolina, February 27-March 2.
Williams, J., and D. Murfield, eds. 1977. Agricultural Atlas of Nebraska. University of Nebraska Press, Lincoln, NE. 110 pp.
Yang, W., and J. Merchant. 1997. Impacts of upscaling techniques on land cover representation in Nebraska, U.S.A. Geocarto International 12(1): 27-39.