Authors: Brodie Alexander – University of Alabama
Title: MoltenFIL: Memory-efficient, parallelized Fill Sinks to enable large-scale terrain processing
Presentation Type:
Abstract: Advancements in resolution of available Digital Elevation Models (DEMs) have highlighted the need for faster and more memory efficient terrain algorithms to extract information from them. Fill Sinks (FIL) is one such algorithm which is a prerequisite for many other hydrologic terrain analysis techniques.
To help fill this need, we have prepared MoltenFIL: a memory-efficient Fill Sinks algorithm which parallelizes solving by dividing the study area into individual sinks, solving Flow Direction (FDR) for each, and then tracing backwards to ensure elevation never decreases going upstream.
The resulting outputs are similar to those achieved by other Fill Sinks implementations while offering a considerable speed and memory advantage by parallelizing processing per-sink and only keeping the currently processing sinks in memory.