IrrNet: Spatio-Temporal Segmentation Guided Classification for Irrigation Mapping

Abstract

Irrigation systems can vary widely in scale, from smallscale subsistence farming to large commercial agriculture (see Fig. 1 ). The heterogeneity in irrigation practices and systems across different regions adds to the complexity of mapping (see Fig. 1 ). Distinguishing between irrigated and non-irrigated areas is challenging due to the spectral characteristics of various irrigation systems and practices across different regions, further complicating the task of mapping different types of irrigation. For example, rainfed agriculture is prevalent in the Midwest, Southeast, and parts of the Northeast U.S., while irrigation is common in arid Western and Southwestern states. Rainfed farming can result in highly variable patterns of cultivation. Farmers may practice rainfed agriculture in some fields while irrigating others, leading to a complex mosaic of irrigated and non-irrigated areas within the same region.