Evaluating RapidEye and Landsat Data for Winter Wheat Mapping in Fine Scale Agricultural System of Punjab, Pakistan.

Publicly available cost-free such as Landsat data perform well in characterizing cropping systems. Commercial high spatial resolution data are often preferred in mapping fine scale agricultural systems. For landscapes such as Punjab province in Pakistan, winter wheat is the most important commodity crop. In the presented research, we characterized winter wheat by integrating within growing season 5 m RapidEye imagery with 30 m Landsat time-series data. After co-registration, each RapidEye image classified into wheat/no wheat labels were aggregated as percent cover to Landsat resolution. We produced four maps, two using RapidEye derived continuous training data (of percent wheat cover) as input to a regression tree model, and two using RapidEye derived categorical training data as input to a classification tree model. From the RapidEye-derived 30 m continuous training data, we derived percent wheat per pixel (Map 1), and binary wheat/no wheat classification derived using a 50% threshold (Map 2) applied to Map 1. To create the categorical wheat/no wheat training data, we first converted the continuous training data to a wheat/no wheat classification, and then used these categorical RapidEye training data to produce a categorical wheat map from the Landsat data. The first method used a 50% wheat/no wheat threshold to produce Map 3, and the second method used only pure wheat (≥75% cover) and no wheat (≤25% cover) training pixels to produce Map 4. The approach of Map 4 is analogous to a standard method in which whole, pure, high-confidence training pixels are delineated. We validated the wheat maps with field data collected using a stratified, two-stage cluster design. Accuracies of the maps produced from the percent cover training data (Maps 1 and 2) were not substantially better than accuracies of the maps produced using categorical training data, (overall accuracy±standard error): 88% (±4%) for Map 1, 90% (±4%) for Map 2, 90% (±4%) for Map 3, and 87% (±4%) for Map 4. Because the percent cover training data did not produce significantly higher accuracies, we posit that sub-pixel training data are not required for winter wheat mapping in Punjab. For other like landscapes, training data for supervised classification may be collected directly from Landsat images without the need for commercial high resolution reference imagery.