Multisource Precipitation Estimation Using Artificial Neural Networks and Geographically Weighted Regression for a Hyperarid Environment

Accurate and timely monitoring of precipitation remains a challenge, particularly in hyper-arid regions such as the United Arab Emirates (UAE) characterized by complex land-atmosphere dynamics and inadequate hydrological monitoring networks. The aim of this study is to improve the accuracy of the Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement (GPM) mission’s latest product release (IMERG V06B) locally over the UAE through two distinct approaches, namely, geographically weighted regression (GWR) and artificial neural networks (ANNs). Daily soil moisture retrievals from the soil moisture active passive (SMAP) mission (9 km), terrain elevations from the Advanced Spaceborne Thermal Emission and Reflection digital elevation model (ASTER DEM, 30 m), and precipitation estimates (0.5 km) from a ground-based weather radar network are incorporated as explanatory variables in the proposed GWR and ANN model frameworks. The Nash-Sutcliffe efficiency (NSE) and Pearson correlation coefficients (PCC) are used as performance measures, while error metrics include the probability of detection (POD), false alarm ratio (FAR), relative bias (rBIAS), and root mean squared error (RMSE).

First, the performances of the daily GPM and weather radar precipitation estimates are assessed using rain gauge observations from a network of 65 gauges from January 1, 2015 to December 31, 2018. Next, the GWR and ANN models are developed with 52 gauges used for training, and 13 gauges reserved for model testing and inter-comparisons over ungauged areas for summer and winter periods. Results show that GPM estimates perform better with increasing elevation (z) and higher rainfall amounts (PCC = 0.29 z^0.12), while weather radar estimates perform better for lower elevations and light rain conditions (PCC = 0.81 z^-0.18). SMAP soil moisture retrievals compare well with the rain gauge observations with PCC values reaching 0.78 over Al Ain and the western region, and an interquartile range of 0.38 to 0.58, corroborating its use as a surface signature for observed rainfall events. Taylor diagrams show that both GWR and ANN models outperform the individual GPM and radar estimates, with the poorest correction obtained by GWR during the summer period. The incorporation of soil moisture resulted in improved corrections by the ANN model compared to the GWR during summer. Better performance is consistently recorded by the ANN compared to GWR with relative NSE improvement rates of 56% (and 25%) for GPM estimates and 34% (and 53%) for radar estimates during summer (and winter) periods. The ANN-derived precipitation estimates can be used as more reliable inputs for hydrological applications over ungauged areas across the UAE. The methodology followed is applicable to other arid and hyper-arid regions requiring improved precipitation monitoring.