From deterministic to stochastic: A hybrid framework coupling machine learning with Bluecat for probabilistic satellite-based precipitation estimation over Greece

Study region: This study focuses on Greece, characterized by complex topography and strong spatial and temporal variability in precipitation variability, which challenge accurate rainfall estimation and hydrological applications. Reliable precipitation data are essential for hydrological modeling, water resource management and rainfall extremes assessment, while satellite products provide extensive spatial coverage but are subject to systematic biases relative to ground observations.

Study focus: We develop a hybrid machine learning-stochastic framework for post-processing satellite-derived precipitation using ground-based observations. Daily gauge and Integrated Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG) data for the period 2000–2021 are used to evaluate independent machine learning algorithms and the proposed framework. The stochastic component provides uncertainty estimation, achieving 88.7% coverage at a 90% confidence level. We introduce a method for estimating the output distribution to generate intensity-time scale-return period (ombrian) curves using annual maximum daily rainfall from a denser station network.

New hydrological insights for the region: The proposed framework substantially improves satellite precipitation estimates over Greece. The estimated output distribution achieves Nash–Sutcliffe efficiency (NSE) and Kling–Gupta efficiency (KGE) values of 0.99 and 0.96, respectively, compared with 0.78 and 0.65 for the original satellite product. The framework provides accurate precipitation estimates and reliable uncertainty quantification, supporting hydrological modeling, flood frequency analysis and robust Ombrian curve development in regions with complex terrain. By integrating post-processing techniques within the learning pipeline, the proposed method improves accuracy over conventional merging approaches.