Filling depressions based on sub-watersheds in raster digital elevation models

Filling depressions is a commonly used preprocessing step for the automatic extraction of drainage networks from raster digital elevation models (DEMs). The Priority-Flood algorithm is the fastest depression-filling algorithm for floating-point DEMs. Most of the variants of the Priority-Flood algorithm processes disjoint depressions using one single priority queue, without taking advantage of the fact that the disconnected depressions can be filled independently and that the running times can be reduced accordingly. This study proposes a new algorithm to process sub-watersheds independently for the generic floating-point DEMs. The proposed algorithm draws largely on the Priority-Flood algorithm, identifies and processes each sub-watersheds independently, which provides more insight into the Priority-Flood algorithm. Its efficiency in processing small DEM datasets can be used to process each small tiles in tile-based parallel filling of depressions.