You are here:
Monitoring Agricultural Cropping Patterns in the Great Lakes Basin Using MODIS-NDVI Time Series Data
Citation:
LUNETTA, R. S. AND Y. Shao. Monitoring Agricultural Cropping Patterns in the Great Lakes Basin Using MODIS-NDVI Time Series Data. Presented at Fifth International Workshop on the Analysis of Multi-temporal Data (MultiTemp 2009), Mystic, CT, July 28 - 30, 2009.
Impact/Purpose:
Presentation
Description:
This research examined changes in agricultural cropping patterns across the Great Lakes Basin (GLB) using the Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) data. Specific research objectives were to characterize the distributions of crop types in the GLB for 2005, 2006, and 2007, and quantify any changes in rotation patterns between 20052006 and 20062007. MODIS 16-day composite NDVI product (MOD13Q) (20052007) obtained from the USGS EROS Data Center were pre-processed using the method developed by Lunetta et al. (2006). This created a continuous high quality NDVI dataset for the general cropland and crop-specific (e.g., corn, soybean, wheat) mapping. For each calendar year, the classification of general cropland versus non-cropland was conducted first. The training data points for the cropland and non-cropland were primarily derived from the 2001 NLCD (National Land Cover Data). The identifications of individual crop types were subsequently conducted within the cropland mask. Three major crop types were considered including corn, soybeans, and wheat. The training pixels were identified using visual interpretation of MODIS NDVI profiles to derive “phenology end-members”. We employed an ecoregion-stratified classification approach to improve the classification performance by dividing the study area into 12-ecoregions (Omernik, 1987) and conducting independent crop-specific classification within each ecoregion (Figure 1). Both the general cropland classification and crop-specific classification were performed using three-layer multilayer perceptron (MLP) classifiers.