Random forest-based rainfall retrieval for Ecuador using GOES-16 and IMERG-V06 data

A new satellite-based algorithm for rainfall retrieval in high spatio-temporal resolution Ecuador is presented. The relies on the precipitation information from Integrated Multi-SatEllite Retrieval Global Precipitation Measurement (GPM) (IMERG) and infrared (IR) data Geostationary Operational Environmental Satellite-16 (GOES-16). It was developed to (i) classify area (ii) assign rate. In each step, we selected most important predictors hyperparameter tuning parameters monthly. Between 19 April 2017 30 November 2017, brightness temperature derived GOES-16 IR channels ancillary geo-information were trained with microwave-only IMERG-V06 using random forest (RF). Validation done against independent not used training. validation results showed technique (multispectral) outperforms IR-only IMERG product. This offers multispectral can improve performance compared single-spectrum approaches. standard verification scored a median Heidke skill score of ~0.6 rain delineation R between ~0.5 ~0.62 rate assignment, indicating uncertainties Andes’s elevation. Comparison RF rates 2 km2 daily gauge measurements reveals correlation = ~0.33.