High-Resolution Spatiotemporal Estimates of Soil Moisture on HPCs

To understand climatological effects on ground mobility, the US Army needs high-resolution, spatiotemporal estimates of soil moisture. GeoWATCH (Geospatial Weather Affected Terrain Conditions and Hazards) is a Python-based software application developed by Creare to generate high-resolution soil moisture products on-demand and in near-real-time for historical, nowcasting, and forecasting needs. GeoWATCH was designed to be deployed as an interactive web map application; the web interface hosted on a single server is not suitable for defining and generating climatological studies that require a large number of high-resolution historical soil moisture estimates. To support such climatological studies, we repackaged GeoWATCH to run on High Performance Computing (HPC) architectures.

In this presentation we will explain the process of repackaging this geospatial web application for massively parallel computation and statistical analysis on HPCs. We will outline the new user workflow, the overall architecture of the HPC-enabled GeoWATCH, and the mechanics of the parallelized computation. We will describe the Python scripts that feed the HPC queue system; the queue system scripts that run jobs on HPC with a containerized version of GeoWATCH; and the Python post-processing scripts to handle aggregation, evaluation and stitching of soil moisture product tile outputs from GeoWATCH into climatological statistical results spanning large regions and timeframes.