USGS: Science for a changing world

"Reprinted" from Geocarto International, December 1995, v. 10, no. 4, p. 69-80.

Validation of National Land-Cover Characteristics Data for Regional Water-Quality Assessment

Ronald B. Zelt
U.S. Geological Survey
Cheyenne, WY 82001 USA

Jesslyn F. Brown
Hughes STX Corporation
EROS Data Center,
Sioux Falls, SD 57198 USA

Michael S. Kelley
Center for Advanced Land Management Information Technologies
Conservation and Survey Division
University of Nebraska-Lincoln
Lincoln, NE 68588-0517 USA

Abstract

Land-cover information is used routinely to support the interpretation of water-quality data. The Prototype 1990 Conterminous U.S. Land Cover Characteristics Data Set, developed primarily from Advanced Very High Resolution Radiometer (AVHRR) data, was made available to the U.S. Geological Survey's National Water-Quality Assessment (NAWQA) Program. The study described in this paper explored the utility of the 1990 national data set for developing quantitative estimates of the areal extent of principal land-cover types within large areal units. Land-cover data were collected in 1993 at 210 sites in the Central Nebraska Basins, one of the NAWQA study units. Median percentage-corn estimates for each sampling stratum were used to produce areally weighted estimates of the percentage-corn cover for hydrologic units. Comparison of those areal estimates with an independent source of 1992 land-cover data showed good agreement.

Introduction

In 1991, the U.S. Geological Survey (USGS) began full-scale implementation of the National Water-Quality Assessment (NAWQA) Program. The goals of the program include the identification, description, and explanation of major natural and human factors that affect water-quality conditions (Leahy et al., 1990). Land-cover information is used routinely to support the interpretation of water-quality data (Goolsby et al., 1991, 1993; Kellogg et al., 1992; Stamer and Huntzinger, 1994). Within the NAWQA Program, land-cover data are used in characterizing the drainage areas upstream from surface-water sampling sites and ground-water areas selected for spatially distributed sampling. The areal characteristics are useful for sorting the large number of sampled areas into groups of similar stream basins or ground-water areas for comparative analyses. Statistical models also are being developed to relate specific physical, chemical, or biological conditions of sampled water to the proportions of selected land-cover types or related explanatory variables. Knowledge of the spatial patterns and extent of land-cover types can aid water-resource managers, planners, and regulators in understanding where surface water or associated alluvial groundwater is likely to be adversely affected, for example, by certain herbicides (Stamer and Zelt, 1994).

One of the initial planning activities undertaken in each of the 60 NAWQA study-unit investigations is the development of a strategy for acquiring data sets that provide quantitative information about the ancillary variables that may affect water-quality conditions in the study unit. In developing such a plan, investigators evaluated the availability and suitability of existing data for satisfying the program objectives. During the period of time when the initial 20 NAWQA study-unit investigations were evaluating existing land-cover data, a Prototype 1990 Conterminous U.S. Land Cover Characteristics Data Set (U.S. Geological Survey and University of Nebraska-Lincoln, 1993) was produced by the USGS EROS Data Center (EDC), in collaboration with the Center for Advanced Land Management Information Technologies (CALMIT) of the University of Nebraska-Lincoln (UNL), as part of the U.S. Global Change Research Program. The purpose of the data set was to provide a national synoptic overview of land-cover characteristics to complement other sources of land-cover data.

Some properties of the EDC/CALMIT data set were determined by the satellite image data from which it primarily was developed. National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) sensors are carried aboard two NOAA satellites in high polar orbits. Most of the globe is scanned twice daily, and the resulting data are gridded using nominally 1-km2 pixels. To minimize the effects of cloud cover and to manage the large volume of data, EDC researchers derived 28-day maximum-value composite images of a vegetation "greenness" index (the normalized-difference vegetation index) computed from 1990 AVHRR spectral-reflectance data (Eidenshink, 1992). To be successful, the compositing procedure requires precise geometric registration of each daily image to a common coordinate system. A systematic geometric correction based on a satellite-platform model was followed by a precision correction using image-to-image correlation with a reference image having a verified geometric accuracy (root-mean-square error) within 1.0 pixel (Kelly and Hood, 1991). Eight composite images summarize the phenological changes in flora for March through October 1990.

Loveland et al. (1991) described the analytic methods by which an unsupervised clustering algorithm was used with the eight composite images to define 70 vegetation classes (spectro-temporal or "seasonally distinct" classes) that were distinguishable by their vegetation-index values through 1990. Most of the 70 preliminary classes were subdivided on the basis of ancillary geographic data (elevation, ecological regions, and frost-free period) to produce a final classification representing 159 seasonally distinct land-cover (SDLC) regions (Brown et al., 1993). A variety of existing digital and conventional maps, photographs, and images was evaluated to determine the characteristics of the 159 regions. The resulting descriptive labels (examples in Table 2) are the investigators' best estimate of the land-cover types associated with each region, as viewed from a nationwide perspective. The labels are qualitative descriptions only and, where they indicate multiple land-cover types within a region, do not indicate which types predominate or in what relative proportions they typically occur.

Merchant et al. (in press) report results from several attempts at non-site-specific validation of the 1990 SDLC region labels as encouraging but inconclusive. Site-specific verification of the SDLC regions is in progress, but results are not available yet.

The spatial resolution of the AVHRR scanner is sufficiently coarse that the field of view may include several land-cover types having heterogeneous spectral-reflectance properties. The resulting mixed pixels in the digital image do not record spectral reflectance for a particular land cover, but rather an average of reflectances from the various features. This complication is not limited to coarse spatial-resolution data. For example, using high spatial-resolution imagery, Fung and Chan (1994) examined the spatial composition of the several spectral classes that were associated with test sites representing a single land-use/land-cover type. Most of their test sites were characterized by considerable spectral heterogeneity. However, by quantifying the spatial composition at their test sites, Fung and Chan were able to identify the dominant spectral classes for each land-use/land-cover type. This suggests that by quantifying the spatial composition at test sites corresponding to AVHRR pixels, quantitative estimates of the extent of each land-cover type can be obtained for areas defined on the basis of the SDLC regions in the EDC/CALMIT data set. Moody and Woodcock (1994) also have suggested that land-cover classes based on image data having 1-km spatial resolution be defined in terms of mixtures of land-cover types for which the proportions could be estimated.

Objective and Scope

The study described in this paper primarily explored the utility of the EDC/CALMIT data set for developing quantitative estimates of the areal extent of the principal land-cover types within large areal units that are hydrologically defined. This paper presents the preliminary results from collection of land-cover data at 210 sites in 1993 in the Central Nebraska Basins NAWQA study unit. The sampling design, which incorporated stratification based on the EDC/CALMIT data set, also is presented.

Study Area

The Central Nebraska Basins (CNB) was among the initial 20 study-unit investigations begun as part of the full-scale NAWQA Program. The CNB study unit (Figure 1) consists of the area drained by the Platte River from the confluence of the North Platte and South Platte Rivers near North Platte downstream to its confluence with the Missouri River south of Omaha. The study unit encompasses about 78,000 km2 and includes the Loup and Elkhorn River Basins as well as basins of smaller tributaries to the Platte River. As one of the largest active NAWQA study units and an area of interest to the developers of the EDC/CALMIT data set, the CNB unit was a logical choice as an area in which to study methods for quantifying areal estimates of land cover based on SDLC regions. In addition, this investigation provided the EDC/CALMIT team with an improved understanding of the feasibility of refining for regional problems a data set that was developed for nationwide use.

(27 kb, gif)

Figure 1. The Central Nebraska Basins study unit is one of 60 within the USGS National Water-Quality Assessment Program.

Land-Cover Sampling Design

The target population for this study consisted of all locations within the CNB study unit. The sampling frame was the set of 1-km2 cells within the CNB study unit that were classified by EDC and CALMIT (U.S. Geological Survey and University of Nebraska-Lincoln, 1993) into SDLC regions. The information desired was the extent of selected land-cover types within areal units. For this study, the cover type at any point on the land was defined as the one category from Table 1 that best described the observed vegetation covering the land surface or as one of the categories that described nonvegetated areas. Although only the principal land-cover types were of immediate interest, data were recorded for several minor land-cover types for possible use in other investigations being conducted in the study unit.

Because various types of land cover have differing spectral reflectance properties and because the spatial organization of land cover varies among physiographic and ecological regions, the homogeneity of land-cover characteristics of SDLC regions within the CNB study unit was expected to be greater than for the study unit as a whole. Therefore, the stratified-random sampling design based on the SDLC regions that was selected for this study was expected to produce more precise estimates of land cover for areal units than would simple random sampling.

Study-unit Stratification and Sample Selection

In 1993, a preliminary stratification design was developed to collect onsite land-cover observations. The design involved six sampling strata, each of which was defined as the set of all 1-km2 cells that had been assigned to a specific set of SDLC regions. The grouping of SDLC regions into the six strata was performed on the basis of the nationwide labels that had been assigned to the regions by the EDC and CALMIT researchers. Table 2 lists the groups of SDLC regions that compose the sampling strata. Stratum A is composed of SDLC regions that had nationwide labels indicating row crops; stratum B regions were labeled as mixed crops; stratum C region labels indicate mixtures of grass and crops; stratum D region labels include woods and crops; stratum E regions were labeled as grass; and stratum F region labels indicate woods and grass as typical. The areal extent of the six sampling strata (Figure 2) includes about 99 percent of the study unit.

Within each sampling stratum, sites were chosen using simple random selection. Thirty-five 1-km2 sampling sites were selected from each of the six sampling strata, yielding a total of 210 sampling sites. Because all of the 1-km2 cells from each SDLC region either were included in or excluded from each sampling stratum, each cell within a SDLC region had an equal probability of being selected.

Figure 2. Six groups of land-cover regions (Table 2) in the Central Nebraska Basins were selected for onsite-verification sampling in 1993. The location of three example sampling sites is shown.

Data Collection and Analysis Procedures

Once the set of sampling sites was selected, preparations were made for locating and gaining access to each site. Each observer was provided with a written description of data-collection procedures and 2 to 4 hours of training in the use of the procedures.

To assist the investigators in locating each sampling site with respect to observable reference features on the land, the sampling-site boundary was plotted on transparent film at a scale of 1:24,000 and registered as an overlay to the corresponding USGS 7.5-minute topographic quadrangle map.

Color-infrared aerial photography at a nominal scale of 1:40,000 was acquired by the USGS National Aerial Photography Program (NAPP) for all of Nebraska during 1987-90. A black-and-white print of a NAPP photograph of each sampling site was supplied to the observers to provide detailed spatial reference material for locating sampling sites in the field and to serve as a base for recording observed land-cover features.

The identity and location of land-cover features at each sampling site were determined by visual inspection. For the 21 sites to which access permission was not secured or which were otherwise inaccessible to the observer, land-cover data were determined by interpretation of the aerial photography. No attempt was made to compensate for the date of the photographs being prior to 1993, but all of the inaccessible sites were in rangeland-dominated areas that have exhibited minimal change in land cover.

The areal extent of the land-cover types of interest was delineated on the aerial photograph (Figure 3) and labeled with an alphabetic code corresponding to one of the land-cover categories listed in Table 1. To document the onsite observations photographically, three to five photographs generally were taken at each site using 35-mm color film. The first photograph for each site included the site number (Figures 3b, 3c) to document the site identity of each group of photographs.

Table 1 - Land-cover categories for which data were recorded
Land-cover category	Category definition
Corn	Features showing evidence that area is planted to this crop
Sorghum	Features showing evidence that area is planted to this crop
Soybeans	Features showing evidence that area is planted to this crop
Wheat, oats, barley, or rye	Features showing evidence that area is planted to one of these crops
Alfalfa	Features showing evidence that area is planted to this crop
Other hay or grass	< 50% tree closure
Forest or woodlands	> 50% tree closure
Emergent aquatic vegetation	< 50% open water
Urban or built-up	Commercial, residential, industrial, or transportation structures
Barren land	< 50% vegetated
Open water	< 50% emergent aquatic-vegetation zones

A. Nebraska Sandhills (site 63) (fig 3a button)

(292 kb, gif)
B. Platte Valley (site 184) (fig 3b button)

(265 kb, gif)
C. Glaciated Area (site 44) (fig 3c button)

(289 kb, gif)

Figure 3. Onsite verification of 1993 land cover included recorded observations on an aerial-photograph base and photographic documentation of sites in the (a) Nebraska Sandhills, (b) Platte Valley, and (c) Glaciated Area of eastern Nebraska.

Table 2 - Preliminary land-cover sampling stratification for the Central Nebraska Basins
Sampling stratum	Seasonally distinct land-cover region number and label	Area as a percentage of study unit

A	10 Soybeans, cotton, rice, corn	0.5
	11 Corn, soybeans	10.5
	17 Corn, soybeans	15.9
	27 Irrigated crops	.3
	34 Irrigated crops, mixed row crops	5.7

B	5 Small grains, mixed row crops	1.3
	6 Mixed row crops, small grains	.4
	7 Mixed crops	.9
	13 Small grains, mixed row crops, pasture	.6
	14 Mixed crops	12.9
	16 Corn, soybeans, alfalfa, flax, wheat	.1

C	35 Grass, small grains	20.8
	36 Grass, wheat, sorghum	.2
	37 Small grains, sorghum, grass	< .1
	38 Grass, wheat, peas, lentils	< .1
	39 Grass, wheat, sorghum	7.0

D	40 Woods, irrigated crops, grass	.4
	42 Corn, soybeans, sorghum, irrigated crops, mixed woodlots	3.5
	46 Woods, corn, soybeans, pasture	2.1
	47 Woods, soybeans, corn, pasture	.6
	52 Forage crops, hay, woodlots	.6

E	58 Grass	.4
	59 Grass	7.4
	61 Grass	.1
	64 Grass	3.0
	65 Grass	3.6

F	91 Woods, pasture	.3
F	93 Woods, pasture	< .1

To estimate the percentage of the site covered by each land-cover category, a grid planimeter (10 squares per 25.4 mm) was overlain on the aerial photograph on which the land-cover units had been delineated. The observer recorded the number of grid cells covered by each land-cover category. Each such cell count was converted to a percentage estimate by dividing by the total number of grid cells that equaled the area of the entire site. The result was rounded to the nearest whole percent.

In addition to quantifying the land-cover composition for each sampling site, descriptive and inferential statistics were computed to summarize the data for each sampling stratum. Stephens' approximation (D'Agostino and Stephens, 1986, chapter 5) of the Shapiro-Wilk test for normality was applied to the sample data from each stratum. The median percentage of site coverage was computed for each land-cover category by stratum. Nonparametric confidence intervals for the medians were computed at the 95-percent confidence level.

Areal land-cover estimates were computed for selected land-cover categories for areal units corresponding to the USGS hydrologic cataloging units (Seaber et al., 1986). The estimates were areally weighted sums of the summary statistics for the strata that composed each hydrologic unit. Three estimates were made using three summary statistics: the median, the lower limit of the confidence interval for the median, and the upper limit of the confidence interval. As a preliminary check on their validity, these estimates were compared to an independent source of land-cover data for hydrologic cataloging units--the 1992 National Resources Inventory (U.S. Department of Agriculture, Soil Conservation Service, 1994). Results from the 1992 National Resources Inventory are sample-based estimates that have a margin of error of 5% or less at the substate level and larger margins of error for smaller areal units.

To support the interpretation of water-quality data, the percentage of nine selected surface-water drainage areas covered by row crops (corn, sorghum, and soybeans) was estimated using areally weighted sums of the median row-crops percentage for each sampling stratum. Those estimates were then compared with differences between the streams in terms of water-quality metrics, such as the index of biotic integrity (IBI) for fish (Karr, 1981). The IBI values were averages for fish samples collected during 1993-94 at the downstream end of each of the nine selected drainage areas.

Results And Discussion

Quantitative land-cover data were recorded for 210 sampling sites during June 21 through August 11, 1993. The null hypothesis of normality in the frequency distribution of land-cover percentages was rejected with 95-percent confidence for most land-cover categories in all sampling strata. Skewness and outliers were apparent in most of the data distributions. Therefore, appropriate summary statistics for point and interval estimates of the center of the data for each land-cover category are the median and nonparametric interval estimate for the median (Helsel and Hirsch, 1992). The results of these computations are listed in Table 3. The large degree of variability in land-cover composition for strata B and D are reflected in 95-percent confidence intervals that are more than 20 percent in width for the principal land-cover categories (corn and grass). One possible source of variability in all six strata is the temporal difference between the 1990 AVHRR imagery and the 1993 sample observations. However, given that most of the SDLC regions in the CNB study unit correspond either to rangeland or to mixtures of consistently rotated field crops, it seems plausible that using a 1993 version of the SDLC regions would not significantly decrease the observed degree of variability in the sample data. A likely source of some variability in the data is the effect of the small residual errors in geometric registration of the daily AVHRR images. The geometric correction techniques applied, although precise, are certainly imperfect.

The simple random selection of 35 sampling sites within each stratum has a limitation that is evident in the results shown in Table 3; the sample-based estimates of percentages of minor land-cover categories are too small. Either a much larger sample size or a more sophisticated sampling design would be needed to provide better estimates for minor land-cover categories.

Table 3 - Summary statistics for 1993 land cover by sampling stratum
Sampling	Percentage of site covered by indicated land-cover category
stratum	Corn	Sorghum	Soybeans	Alfalfa	Wheat, etc.	Grass
Upper limit of 95-percent confidence interval for stratum medians
A	61	0	23	0	0	25
B	41	0	0	2	0	84
C	0	0	0	0	0	100
D	24	0	0	1	0	91
E	0	0	0	0	0	100
F	72	0	0	21	0	5
Stratum medians
A	50	0	7	0	0	14
B	22	0	0	0	0	65
C	0	0	0	0	0	100
D	6	0	0	0	0	56
E	0	0	0	0	0	100
F	67	0	0	10	0	0
Lower limit of 95-percent confidence interval for stratum medians
A	33	0	0	0	0	10
B	0	0	0	0	0	45
C	0	0	0	0	0	94
D	0	0	0	0	0	27
E	0	0	0	0	0	100
F	55	0	0	6	0	0

Figure 4 summarizes graphically the results by stratum for corn, one of the principal land-cover categories in the study unit. Again, for strata B and D especially, the large variability in the percentage of sites covered by corn is reflected in wide 95-percent confidence intervals. However, the median estimates for each stratum do provide some information that can be used to gain insight into the spatial distribution of corn in the study unit. The median percentage corn estimates for each sampling stratum were used together with the areal extent of each stratum within individual USGS hydrologic cataloging units to produce areally weighted estimates of the relative extent of corn within each cataloging unit. Figure 5 shows a comparison of those areal estimates with data from the 1992 National Resources Inventory (NRI). Given the temporal difference between the data sets and the large degree of variability observed in the sample data, the correspondence with the NRI data is remarkably good (R=0.946) for most of the cataloging units. Some of the disagreement between the two sets of estimates also may be due to the imprecision present in the 1992 NRI cataloging-unit estimates.

Figure 4. Median and 95-percent confidence interval of percentage of sampling site covered by corn, 1993, by land-cover sampling stratum (Table 3). Stratum medians were used to assign cells in the National Land-Cover Characteristics Data Set to a percentage-corn category.

Figure 5. The agreement between areally weighted estimates of percentage corn cover based on the 1993 sampling results and an independent source of land-cover data is quite close for most hydrologic cataloging units in the Central Nebraska Basins.

Numerous water-quality conditions and indicators are affected by land-cover characteristics. To illustrate, figure 6 shows results for one widely used water-quality indicator metric, the IBI for fish, in relation to the percentage of upstream land area covered by row crops. There is a strong inverse association indicated (R=-0.791). Relations between land-cover characteristics and many other water-quality metrics will be examined during the remainder of the Central Nebraska Basins study.

Figure 6. Relation between land cover and water quality is illustrated by data for nine fish-sampling stations and their upstream drainage areas within the Central Nebraska Basins.

In addition to the estimates for the six sampling strata, there were eight individual SDLC regions for which 11 or more sites had been selected and sampled during the course of the 1993 data collection. Point and interval estimates of the median land-cover percentage also were computed for those eight SDLC regions. In general, the 95-percent confidence intervals for those SDLC regions were smaller than the intervals for the sampling strata because the decreased variability that resulted from sampling a single SDLC region more than offset the effect of the smaller sample sizes.

To prepare a map displaying the spatial distribution of the median estimate of the percentage of area covered by corn, a three-step procedure was used. First, each 1-km2 cell that was classified in one of the eight SDLC regions from which 11 or more samples were collected in 1993 was assigned the percentage-corn median computed for its SDLC region. Next, the cells from the remaining SDLC regions that were included in the six sampling strata were assigned their respective stratum-median percentage-corn values. Finally, the resultant derived data set was displayed with range-graded symbology (Figure 7). The map displays spatial patterns in the density of corn production that are related to physiographic/environmental subunits of the study unit (Figure 7 inset). The Sandhills subunit has relatively little corn cover, whereas the Platte Valley and Glaciated Area subunits are characterized mostly by intensive corn production. The map contributes to a useful conceptual model of current land-cover characteristics in the study unit.

(33 kb, gif)

Figure 7. Estimated percentage of land area as corn in 1993 was largest in the Platte Valley and Glaciated Area.

Conclusions

The preliminary results from the study support the validity of using the 1990 National Land Cover Characteristics Data Set for regional studies when it is coupled with local knowledge or additional data. Regional quantification of the land-cover composition associated with SDLC regions may be most useful when performed separately for individual regions rather than for groupings of the regions, even when the region labels are similar. Simple random sampling within these regions can provide quantitative estimates of the relative extent of the principal land covers for moderately sized to large areal units, but minor land covers may be underestimated if based on only a medium-sized sample. Such quantitative estimates of land cover for areal units have demonstrated utility for regional water-quality assessment.

Additional sites in the CNB study unit are being sampled in 1994 to reduce the uncertainty about the sample median statistics for a revised set of sampling strata that are based on individual SDLC regions. Areal estimates of land-cover composition during 1993-1994 are becoming available from independent sources, such as the U.S. Department of Agriculture's National Agricultural Statistics Service. It is expected that comparisons of the final results of this study with results from independent sources will be of interest to regional investigators that require updated estimates of selected land-cover characteristics for a particular year or study area in which data from other sources may not be available.

Acknowledgments

This study was made possible through extensive assistance from CALMIT and USGS/EDC. The authors thank Drs. Thomas Loveland (USGS/EDC), James Merchant (CALMIT), and Bradley Reed (Hughes STX/EDC) for their helpful comments. CALMIT's contribution was supported by the U.S. Environmental Protection Agency (grant X007526-01), whose financial assistance is gratefully acknowledged.

References

Brown, J.F., T.R. Loveland, J.W. Merchant, B.C. Reed, and D.O. Ohlen. 1993. Using Multisource Data in Global Land-cover Characterization: Concepts, Requirements, and Methods. Photogrammetric Engineering and Remote Sensing. 59(6):977-987.

D'Agostino, R.B., and M.A. Stephens (editors). 1986. Goodness of Fit Techniques. New York: Marcel-Dekker. 560 pgs.

Eidenshink, J.C. 1992. The 1990 Conterminous U.S. AVHRR Data Set. Photogrammetric Engineering and Remote Sensing. 58(6):809-813.

Fung, T., and K.-C. Chan. 1994. Spatial Composition of Spectral Classes: A Structural Approach for Image Analysis of Heterogeneous Land-use and Land-cover Types: Photogrammetric Engineering and Remote Sensing. 60(2):173-180.

Goolsby, D.A., W.A. Battaglin, and E.M. Thurman. 1993. Occurrence and Transport of Agricultural Chemicals in the Mississippi River Basin, July Through August 1993. U.S. Geological Survey Circular 1120-C. 22 pgs.

Goolsby, D.A., R.C. Coupe, and D.J. Markovchick. 1991. Distribution of Selected Herbicides and Nitrate in the Mississippi River and its Major Tributaries, April Through June 1991. U.S. Geological Survey Water-Resources Investigations Report 91-4163. 35 pgs.

Helsel, D.R., and R.M. Hirsch. 1992. Statistical Methods in Water Resources. Amsterdam: Elsevier Publishers. 522 pgs.

Karr, J.R. 1981. Assessment of Biotic Integrity Using Fish Communities: Fisheries (Bethesda). 6(6):21-27.

Kellogg, R.L., M.S. Maizel, and D.W. Goss. 1992. Agricultural Chemical Use and Ground Water Quality: Where Are the Potential Problem Areas?. Washington, D.C.: U.S. Department of Agriculture and National Center for Resource Innovations. 41 pgs., appendices.

Kelly, G.G., and JJ. Hood. 1991. AVHRR Conterminous United States Reference Data Set: Technical Papers, 1991 ACSM-ASPRS Annual Convention. 3: 232-239.

Leahy, P.P., J.S. Rosenshein, and D.S. Knopman. 1990. Implementation Plan for the National Water-Quality Assessment Program. U.S. Geological Survey Open-File Report 90-174. 10 pgs.

Loveland, T.R., J.W. Merchant, D.O. Ohlen, and J.F. Brown. 1991. Development of a Land-cover Characteristics Database for the Conterminous U.S. Photogrammetric Engineering and Remote Sensing. 57(11):1453-1463.

Merchant, J.W., L. Yang, and W. Yang. In press. Validation of Continental-Scale Land Cover Data Bases Developed From AVHRR Data. In: William T. Pecora Memorial Symposium on Remote Sensing (12th). Proceedings. Sioux Falls, SD. August, 1993. Bethesda, MD: American Society for Photogrammetry and Remote Sensing.

Moody, A., and C.E. Woodcock. 1994. Scale-Dependent Errors in the Estimation of Land-Cover Proportions: Implications for Global Land-Cover Datasets. Photogrammetric Engineering and Remote Sensing. 60(5):585-594.

Seaber, P.R., F.P. Kapinos, and G.L. Knapp. 1986. Hydrologic Unit Maps. U.S. Geological Survey Water-Supply Paper 2294. 63 pgs.

Stamer, J.K., and T.L Huntzinger. 1994. Spring Herbicide "Flush" May Deal Cities a Problem. Nebraska Farmer. 136(8):10-13.

Stamer, J.K., and R.B. Zelt. 1994. Organonitrogen Herbicides in the Lower Kansas River Basin. Journal American Water Works Association. 86(1):93-104.

U.S. Department of Agriculture, Soil Conservation Service. 1994. 1992 National Resources Inventory. Fort Worth, TX: Soil Conservation Service. 4 compact discs.

U.S. Geological Survey and University of Nebraska-Lincoln. 1993. Prototype 1990 Conterminous U.S. Land Cover Characteristics Data Set. Sioux Falls, SD: U.S. Geological Survey, EROS Data Center. 1 compact disc.

National NAWQA Page NAWQA data warehouse Nebraska related NAWQAs NAWQA Publications Page

U.S. Department of the Interior, U.S. Geological Survey
Created and Maintained by: Nebraska Webmaster Team
E-mail comments or questions to Ron Zelt, Central Nebraska Basins project chief

| Privacy Policy | | Disclaimer | | Accessibility | | FOIA