Classification and Mapping of Agricultural Land for National Water-Quality Assessment

By Robert J. Gilliom and Gail P. Thelin

U.S. Geological Survey Circular 1131

CLASSIFICATION OF AGRICULTURAL LAND BY CROP DISTRIBUTION

A new approach to classifying and mapping the distribution of agricultural land use for water-quality assessment was developed by combining the information on general land-use distribution from the U.S. Geological Survey national LULC data, based on the Anderson classification system, with the county-based agricultural census data. The LULC data is used to locate agricultural land within each county, and the classification of agricultural land in each county is determined from census data on crop distributions.

Data on harvested crops from the 1987 Census of Agriculture were used to classify agricultural land into two primary categories that correspond to the Anderson Level II categories of (1) cropland and pasture, and (2) orchards, groves, vineyards, nurseries, and ornamental horticultural areas, hereafter referred to as "orchards, vineyards, and nurseries" (table 1). Separate classification of agricultural land within these two general categories defined in the Anderson system retains an organizational link to the Anderson system and enables county-based crop statistics and classifications to be associated with actual geographic areas of agricultural land within each county. Anderson's "cropland and pasture" class includes three broad categories of agricultural land as reported in the Census of Agriculture: row crops (including alfalfa), grass hay (all hay except alfalfa), and pasture (rangeland is separate). Grass hay and pasture are widespread and are managed in a relatively similar manner among the different areas of the nation that they are grown, although irrigation is more common in some areas than others. Row crops, in contrast, include a wide variety of individual crops with strong regional tendencies and unique management requirements, including widely varying water demand, tillage, fertilizers, and pesticides. The classification systems developed focus on row crops for land in the "cropland and pasture" Anderson class and on fruit, nut, and nursery crops in the "orchards, vineyards, and nurseries" Anderson class.

Classification of crops in both row-crop areas and orchard, vineyard, and nursery areas was based on the acreage of individual crops harvested, rather than income production or a specific management characteristic, such as cultivation method or chemical use. Conceptually, the goal is for the classification of cropland in a particular county to yield a realistic depiction of the relative areal extent of dominant crops. Use of the census data on harvested acreage causes some land that is harvested more than once to be counted more than once in contributing to acreage. In this sense, the classification is based on annual average acreage of crops for an entire county, including within-year rotation.

Classification based upon the acreage of individual crops has the advantage for water-quality studies of representing conditions prevalent for most of the land that may contribute runoff or recharge for a particular area. This approach has the potential disadvantage of not adequately accounting for small areas of specialized crops that may have unique management practices, such as heavy irrigation or chemical use, and thus a disproportionate effect on water quality. These unique situations, which are not areally extensive, are difficult to evaluate from national county- based data, and generally, any national system will need to be refined using higher resolution data in order to characterize these patterns.

The acreage of each crop in each county, as reported in the census data base, was divided by the sum of individual row-crop acreages or orchard, vineyard, and nursery acreages, as appropriate, in the county. Thus, the classification was developed based on the proportional representation of crops within row-crop and orchard, vineyard, and nursery areas, but is independent of the total amount of row-crop or orchard, vineyard, and nursery land in the county.

Two general approaches were considered for developing the classification system from these normalized data on proportional areas of individual crops in each county. Cluster analysis was considered, but not used. The results from cluster analysis would change with different data from different census years and from different spatial scales. In addition, cluster definitions are interpreted general tendencies rather than clear, quantitative measures of conditions observed in the field. Furthermore, cluster analysis is well-suited to identifying relatively few broad categories of conditions and becomes subjective when used to identify minor categories, which may be very important to water-quality assessment.

The second approach considered, and the one selected, is to use simple percentage criteria to classify each county according to the proportional land area of each crop harvested. This approach has the advantages of being simple and directly related to what is observable in the field. It is also flexible in allowing identification of minor categories in a hierarchical manner without disrupting major categories. The approach is described below and the advantages and disadvantages evaluated.

Row Crops

Row crops used in developing the classification system are listed in table 2 along with 1987 Census of Agriculture (U.S. Department of Commerce, 1995) data codes and national harvested acreage. For each specific crop with a census-item number, the census contains acreage totals for counties with 10 or more farm operations from which $1,000 of agricultural products were produced or sold during the census year. Total acreages for some crops with several subcategories, such as "corn," "wheat and other grains," and "sorghum" were summed from multiple census items for each county to simplify the number of crops to classify. The crops included in table 2 account for almost 100 percent of national cropland devoted to row crops.

Four primary crops-corn, soybeans, wheat and other grains, and alfalfa-each account for 24 to 74 million acres of harvested cropland and are grown throughout large areas of the nation to varying degrees (table 2). Most of the 12 secondary crops in table 2, as aggregated, account for about 1 to 10 million acres. Most of the secondary crops require more specialized growing conditions; tend to be more restricted to smaller regions than the primary crops; frequently are high-value cash crops within the regions that they are grown; and may require intensive chemical and fertilizer use and irrigation. Some crops, such as pineapple, were not included as secondary crops because of their small acreage. The classification process for determining categories of crops from the primary and secondary crops is hierarchical, with two primary levels.

Level I

Level I classification was evaluated for all counties with reported data in the United States (all 50 States) based on a step-wise application of percentage criteria. Counties meeting criteria for each step are removed from the data base prior to applying the next step. The general logic of the level I classification is that 1, 2, or 3 primary or secondary crops that define each category account for more than 50 percent of the harvested acreage in each county classified. Criteria for each type of category are listed in order of application below:

One secondary crop accounts for more than 50 percent of the harvested acreage (12 possible categories).
Three of the four primary crops each account for more than 20 percent of harvested acreage (4 possible categories).
Two of the four primary crops each account for more than 20 percent of harvested acreage (6 possible categories).
One of the four primary national crops account for more than 50 percent of harvested acreage (4 possible categories).

Two-crop categories (criterion 3), as defined, can result in the two primary crops totaling less than 50 percent of the harvested acreage. This occurred for less than 10 percent of the counties in these categories, however, and the criterion for each crop of 20 percent, rather than 25 percent, resulted in fewer unclassified counties.

As summarized in figure 4, categories were established for all of the 26 potential categories in which there were 10 or more counties. In addition, level I categories of secondary crops (step 1) were established if the total acreage of the secondary crop exceeds 50,000 acres among the counties meeting the percentage criteria for a potential category (even if there were less than 10 counties). For example, only four counties in the nation had more than 50 percent of harvested area accounted for by potatoes, but the total acreage harvested among the four counties is more than 50,000 acres and, thus, a category was created in the classification system. This minimum acreage criteria provides an arbitrary, but consistent, basis for identifying crop patterns, particularly in the western United States, that are dominant in relatively few counties, but are great in their areal extent. The result of the level I classification process is 21 categories that collectively classify 2,686 of 3,078 counties in the nation that were analyzed. The 392 counties not classified include 65 with insignificant cropland and 327 that do not fit classification criteria.

Level II

Level II subcategories were separately evaluated for all counties in each level I category. Particularly for level I categories defined by the four primary crops, which account for most of the counties classified, it is important to identify patterns in the secondary crops, which may average smaller harvested acreage, but follow strong regional patterns and have specialized management practices that may have implications for water quality.

Each county in a level I category was classified according to the secondary crop (if any) that constitutes more than 10 percent of harvested acreage. If two or more secondary crops meet the criteria (rare), the county was classified according to the crop with the greater acreage. Level II subcategories were established for 46 of 243 possible categories for which there are 10 or more counties, or for which the total acreage of the secondary crop for the particular level II subcategory exceeds 50,000 acres among the counties meeting the percentage criteria. There are as many as 8 level II subcategories for a particular level I category. A total of 867 counties were assigned to level II subcategories.

Unclassified Counties

The 327 counties not classified by the criteria defined for level I categories were assigned to the closest category by minimizing the sum of deviations from the percentage criteria for potential categories. For example, a county with 17 percent corn, 18 percent soybeans, and 47 percent wheat and other grains would be placed in the "wheat and other grains greater than 20 percent and soybeans greater than 20 percent" category because 2 percent (20-18) is less than deviations for other possibilities, such as "wheat and other grains greater than 50 percent" which has a 3 percent deviation (50-47). Reevaluation of level II subcategories, after adding the 327 formerly unclassified counties resulted in addition of four additional level II subcategories.

Results

Results of the classification process applied to row crops in all counties in the 48 contiguous states using the 1987 Census of Agriculture data are shown in figure 5. Figure 6 shows the distribution and classification of counties that were initially not classified based on numerical criteria (these counties are also included in fig. 5). Each county is classified and color-coded on the maps in figures 5 and 6 based on the proportional areas of harvested crops, regardless of how much row crop land use occurs in the county. The county-wide classification, therefore, must be combined with information on land-use distribution to account for the extent and geographic distribution of row crops in a particular county. Table 3 summarizes the relation of each row-crop category to national row-crop production and table 4 to irrigation, chemical and fertilizer use, livestock, and poultry.

Table 3 shows the relations among crop groups and individual crops and table 4 shows the relations among crop groups and selected indicators of agricultural management practices. For example, examination of the third column of numbers in table 3 shows that counties in the "corn and soybeans greater than 20 percent" category (not including level II subcategories) account for about 32 percent of cropland in the United States and 56 and 64 percent of corn and soybean acreage, respectively. Also evident is that counties in this crop category account for substantial amounts of the alfalfa (14.5 percent), tobacco (18.8 percent), and vegetable (13 percent) harvest. Table 4 shows that counties in the corn and soybean category account for fertilizer and chemical expenses approximately proportional to harvested acreage, but a disproportionately large share of hog production (61 percent). This agrees with the cluster analysis of Sommer and Hines (1991). In contrast, the corn and soybean category accounts for a disproportionately small proportion of irrigated land. As one example of comparative analysis, chemical expenses for "wheat and other grains greater than 50 percent" are about half as much per unit area as the "corn and soybeans" category (compare the ratio of chemical expenses to harvested acreage for each). Major variations in chemical and fertilizer expenses reflect large-scale patterns in pesticide and fertilizer use that are important to consider for water-quality assessment. These and other types of comparisons from tables 3 and 4, combined with the map in figure 5, can be used to evaluate the significance of individual crop categories to particular individual crops and large-scale agricultural management issues.

The results of the row-crop classification are generally consistent with the cluster analysis of Sommer and Hines (1991) for their clusters dominated by individual major crops. Figure 7, for example, shows that virtually the entire area of their "corn, soybeans, hogs" cluster is classified as "corn >20 percent, soybeans >20 percent." For clusters such as "poultry", however, which are not defined primarily by crop patterns, there are a wide variety of row crop categories (fig. 8).

Orchard, Vineyard, and Nursery Crops

Orchard, vineyard, and nursery crops used in developing the classification system are listed in table 5 with 1987 Census of Agriculture data codes and national acreages. No distinction was made between primary and secondary crops, as was done for row crops, because there are no individual orchard, vineyard, and nursery crops that are grown in large quantities over broad areas of the nation to the same degree as are corn, soybeans, wheat, and alfalfa. In addition, mixtures do not occur to the same degree because most of these crops are perennial and are not rotated like row crops. The two orchard and vineyard crops with the greatest national acreage are grapes, with about 800,000 acres, and citrus, with about 1.1 million acres. By comparison, about 74 million acres of wheat and other small grains and about 65 million acres of corn are harvested.The crops included in table 5 account for almost 100 percent of national cropland devoted to orchard, vineyard, and nursery crops, including greenhouses and sod grown for sale.

Level I

Level I classification was evaluated for all counties in the United States with reported harvest of orchard, vineyard, and nursery crops by application of percentage criteria, similar to those applied to row crops. Criteria for each type of level I category are listed below:

One crop accounts for more than 50 percent of the harvested acreage (33 possible categories).
Two crops each account for more than 20 percent of harvested acreage (many possible categories).

Categories were established for all potential categories of either type in which there were 10 or more counties or greater than 10,000 acres. The result of the level I classification process is 26 categories that classify 2,098 of 3,078 counties analyzed, with 980 counties having no reported orchard, vineyard, and nursery cropland.

Level II

Level II subcategories were separately evaluated for all counties in each level I category to further characterize the distribution of orchard, vineyard, and orchard crops. Each county in a level I category was classified according to a second crop (or third) that constitutes more than 10 percent of harvested acreage (if there is a crop exceeding 10 percent). This process resulted in 19 level II subcategories with 10 or more counties or that account for more than 10,000 acres. A total of 954 of the 2,098 classified counties were assigned to level II subcategories, with the remaining counties assigned to level I categories only.

Results

Results of the classification process applied to all counties with harvested orchard, vineyard, and nursery crops are shown in figure 9. Figure 10 shows the distribution and classification of counties that were not initially classified based on numerical criteria (these counties are also included in figure 9). Following the procedure applied to row crops, each county is classified and color coded based on the proportional area of harvested crop, regardless of how much total area is included, and must be combined with information on land-use distribution to account for the extent and geographic distribution of orchard, vineyard, and nursery crops in a particular county. Table 6 summarizes the relation of each category of orchard, vineyard, and nursery crops to national production for all counties in each category. Results of the orchard, vineyard, and nursery crop classification cannot be directly compared to the highly aggregated clusters of Sommer and Hines (1991).

Next - Application to Land-Use Characterization

Back to:
Table of Contents

Pesticide National Synthesis Project