A two-step merging strategy for incorporating multi-source precipitation products and gauge observations using machine learning classification and regression over China

Abstract. Although many multi-source precipitation products (MSPs) with high spatiotemporal resolution have been extensively used in water cycle research, they are still subject to various biases, including false alarm and missed bias. Precipitation merging technology is an effective means alleviate this uncertainty. However, how efficiently improve detection efficiency intensity simultaneously a problem worth exploring. This study presents two-step strategy based on machine learning (ML) algorithms, gradient boosting decision tree (GBDT), extreme (XGBoost), random forest (RF). It incorporates six state-of-the-art MSPs (GSMaP, IMERG, PERSIANN-CDR, CMORPH, CHIRPS, ERA5-Land) rain gauges the accuracy of identification estimation from 2000 2017 over China. Multiple environment variables spatial autocorrelation combined process. The first employs classification models identify wet dry days then combines regression predict amounts classified days. merged results compared traditional methods, multiple linear (MLR), ML models, gauge-based Kriging interpolation. A total 1680 (70 %) randomly chosen for model training 692 (30 performance evaluation. show that (1) (MSMPs) outperformed all original terms statistical categorical metrics, which substantially alleviates temporal biases. modified Kling–Gupta (KGE), critical success index (CSI), Heidke Skill Score (HSS) improved by 15 %–85 %, 17 %–155 21 %–166 respectively. (2) plays significant role merging, considerably improves accuracy. (3) MSMPs obtained proposed method superior MLR, interpolation, models. XGBoost algorithm recommended more large-scale data owing its computational efficiency. (4) performs better when higher-density training. it has strong robustness can also obtain than even gauge number reduced 10 % (237). provides accurate reliable under complex climatic topographic conditions. could be applied other areas well if available.