SARFDAS

SARFDAS (Synthetic Aperture Radar Forest Disturbance Alert System) is a web application from Spatial Informatics Group (SIG) for detecting and validating forest disturbances. Optical satellite imagery is often patchy in persistently cloudy regions, which delays deforestation alerts. SARFDAS instead uses synthetic aperture radar from the European Space Agency: Sentinel-1 ground-range-detected imagery, processed in Google Earth Engine, which images through cloud.

A convolutional neural network with transfer learning predicts disturbance alerts, so the model can be retrained with region-specific information where training data are limited. New alert tables are loaded into the application daily. In a web map interface, users review each predicted alert and mark it as deforestation or no deforestation, and can download the validated data. SIG positions validated alerts as the basis for reliable information dispatch to the field, resource allocation and unbiased area estimates.

SIG states that SARFDAS was created by SIG, was tested in Palawan, the Philippines, and is being adapted for use with SERVIR-Southeast Asia. The source code is published under the Eclipse Public License 2.0; the hosted application's terms of use describe it as a beta that provides experimental deforestation alert predictions.