Identification of ditches and furrows using remote sensing: application to sediment modelling in the Tana watershed, Kenya

Abstract

ABSTRACTRidge-tillage is an agricultural practice where crops are planted on elevated ridges, with furrows in-between. Ridge-tillage has been shown to significantly reduce erosion from croplands, but data on the presence of ridge-tillage is sparse and challenging to collect at the landscape scale. Thus, water quality models often do not account for ridge-tillage in a spatially-explicit manner, potentially overlooking the important impacts of this practice. We have developed a novel method that exploits the spectral, radiometric and linearity shape characteristics to identify both drainage ditches and ridge-tillage furrows using remote sensing of 0.5 m satellite data. We applied the method to the Sasumua watershed in Kenya, where we had false positives in only 3% of randomly selected polygons, and we detected the majority of ditches in 59% of randomly selected polygons. We then assessed the potential value of including these data in sediment modelling, showing that representing these practices could reduce sediment export in the study area by roughly 80%. Being able to readily identify the presence of ditches and furrows could enable the development of more accurate water quality models, and help identify priority areas for intervention to improve water quality (and possibly crop yields) through changing agricultural practices or policies.