County-level Estimates of Landscape Complexity and Configuration in the Coterminous US

One the most obvious difficulties in comparing the influence of landscape on crop production across studies is the choice of landscape metric. There exist countless metrics of landscape composition—the categories of land cover found on a landscape—and landscape configuration—the spatial organization of these categories. Common landscape composition metrics include measures of diversity—such as the Shannon Diversity Index or the Simpson Diversity Index—and measures of land cover composition—such as the percent of the landscape classified as natural cover. Common landscape configuration metrics include measures of patch size (contiguous areas of the same land cover) and mixing as well as edge length (linear length of patch boundaries/perimeter) and fragmentation. Even just considering diversity metrics, numerous options to select from can be found in the literature. Each one of these metrics has its own particularities in terms of sensitivity to scale, rare categories, and boundaries that can significantly alter the conclusions of studies examining the relationship between landscape characteristics and crop production. To address this challenge, we assess the sensitivity of our model results to a number of indicators of landscape composition and configuration using the USDA NASS Cropland Data Layer (CDL) as our indicator of land cover. This dataset classifies land cover at a 30-meter resolution nationwide from 2008 to present using satellite imagery and extensive ground truth data. While the 30-meter spatial resolution of this land cover data cannot accurately represent very small or narrow patches of land cover including shelterbelts and wildflower strips, given its relatively high resolution, full coverage, and historical availability, it is the best data for understanding land cover across agricultural landscapes in the U.S. We extract landscape indices from the CDL data using the landscapemetrics package in R, which considers all land cover in each county’s bounding box with the exception of open water and null categories. We measure compositional complexity using a set of six common landscape metrics associated with the number or the predominance of land cover categories across a landscape. Five of these metrics—Shannon Diversity Index, Simpson Diversity Index, Richness, Shannon Evenness Index, and Simpson Evenness Index—can be considered measures of land cover diversity. The sixth metric–Percent Natural Cover–is a simple measure of the predominance of undeveloped and uncultivated land cover classes (such as wetlands, grasslands, and forests) on a landscape. All of the compositional complexity metrics are aspatial, in that their calculation is not contingent on how land cover categories are arranged within the landscape. Configurational complexity is measured using four landscape metrics associated with the size of land cover patches (continuous areas of a single land cover category), shape of land cover patches, or mixing of land cover categories across the landscape. The metrics Mean Patch Area and Largest Patch Index are most strongly associated with patch size, the Contagion metric is a measure of land cover category mixing and strongly related to patch size, and the Edge Density metric is related to patch size and shape. Unlike the landscape composition metrics, the four landscape configuration metrics are spatially explicit and depend on the arrangement of land cover categories across the landscape.

All code used to build data can be found here: https://github.com/katesnelson/aglandscapes-what-or-how

Resources in this dataset:

• Resource Title: County-level Estimates of Landscape Complexity and Configuration in the Coterminous US File Name: landscape_panel.txt Resource Description: GEOID: State and county FIPS codes in format SSCCC YEAR: Year in which CDL data was collected VALUE: Index value INDEX_NAME: Indices with _AG were computed for the subset of agricultural lands in a county. Indices with _ALL were computed for the entire landscape (agricultural and nonagricultural lands) in a county.

  LSM_AREA_MN_AG/ALL: Mean patch area, a measure of patch structure. Approaches 0 if all patches are small. Increases, without limit, as the patch areas increase. Higher values generally indicate lower complexity. LSM_CONTAG_AG/ALL: Contagion, a measure of dispersion and interspersion of land cover classes where a high proportion of like adjacencies and an uneven distribution of pairwise adjacencies produces a high contagion value. Range of 0 to 100. Higher values generally indicate lower complexity. LSM_ED_AG/ALL: Edge density, a measure of the patchiness of the landscape. Equals 0 if only one land cover is present and increased without limit as more land cover patches are added. Higher values generally indicate higher complexity. LSM_LPI_AG/ALL: Largest patch index, a measure of patch dominance representing the percentage of the landscape covered by the single largest patch. Approaches 0 when the largest patch is becoming small and equals 100 when only one patch is present. Higher values generally indicate lower complexity. LSM_RICH_AG/ALL: Richness, a measure of the abundance of categories. Higher values generally indicate higher complexity. LSM_SHDI_AG/ALL: Shannon Diversity Index, a measure of the abundance and evenness of land cover categories. This index is sensitive to rare land cover categories. Typical values are between 1.5 and 3. Higher values indicate higher complexity. LSM_SHEI_ALL: Simpson Evenness Index, a measure of diversity or dominance calculated as the ratio between the Shannon Diversity Index and the theoretical maximum of the Shannon Diversity Index. Shannon Evenness Index = 0 when there is only one land cover on the landscape and equals 1 when all land cover classes are equally distributed. Higher values generally indicate higher complexity. LSM_SIDI_ALL: Simpson Diversity Index, a diversity measure that considers the abundance and evenness of land cover categories. This index is not sensitive to rare land cover categories. Values range from 0 to 1. Higher values generally indicate higher complexity MODE_AG : Most dominant agricultural land use type found in the data (mode of agricultural CDL categories) MODE_ALL : Most dominant land use type found in the data (mode of all land use categories) PNC : Percent natural cover

• Resource Title: Technical Validation File Name: technical_validation.txt