Evaluation of PERSIANN family remote sensing precipitation products for snowmelt runoff estimation in a mountainous basin

Estimating snowmelt runoff in mountainous basins is a challenging task due to limited precipitation measurements. Satellite-based Precipitation Products (SBPs) are readily available, but still suffer from large errors in cold climate regions. This study aims to evaluate and propose post-bias corrections for uncorrected Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) and pre-bias-corrected PERSIANN-Climate Data Record (PERSIANN-CDR) daily SBPs using enriched observed data in the upper Euphrates River Basin, Turkey. SBPs are also employed in a daily multilayer perceptron (MLP) model to quantify the impact on runoff. PERSIANN-CDR outperforms PERSIANN in terms of correlation and detection, but biases substantially increase in PERSIANN-CDR in the snow accumulation season. The MLP has been trained and validated using observed precipitation data with 0.86 and 0.83 Nash-Sutcliffe efficiency (NSE), respectively. Applying post-bias corrections by a modified multiplicative linear scaling method improves runoff estimation with NSE values increasing from 0.48 to 0.61 and 0.38 to 0.68 for PERSIANN and PERSIANN-CDR, respectively.

Automated summary

This study evaluates and proposes post-bias corrections for uncorrected Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) and pre-bias-corrected PERSIANN-Climate Data Record (PERSIANN-CDR) daily Satellite-based Precipitation Products (SBPs) in mountainous basins. The study shows that applying post-bias corrections using a modified multiplicative linear scaling method can improve runoff estimation accuracy in mountainous regions.