Multi-dimensional evaluation of four monthly long-term and high-resolution gridded precipitation datasets on the Qinghai-Tibet Plateau

Abstract

The reliability assessment of gridded precipitation datasets (GPDs) is vital for various precipitation-related studies, particularly in poorly gauged regions as the Qinghai-Tibet Plateau (QTP). However, existing studies predominantly focuses more on mainstream GPDs with low spatiotemporal resolution, making it challenging to meet the requirements for widespread applications. There is an urgent need for a systematic assessment of long-term and high-resolution GPDs. In this study, we considered four GPDs (HRLT, TPHiPr, PENG, and CHIRPS) with a spatial resolution above 0.05° during 1981–2019, evaluating their monthly performance across three dimensions: spatial accuracy, temporal accuracy, and precipitation event detection ability. The results revealed the following: (1) TPHiPr exhibited the highest spatial accuracy (R2 = 0.81), and the four GPDs consistently performed the worst in arid areas (R2 = 0.59 ± 0.15) and the best in semi-humid areas (R2 = 0.80 ± 0.02). (2) PENG attained the highest temporal accuracy (NSE = −0.56), and the four GPDs performed the worst in winter (NSE = -1.06 ± 0.9) and the best in autumn (NSE = 0.08 ± 0.16). (3) HRLT and TPHiPr demonstrated superior precipitation event detection abilities compared to others. The four GPDs consistently exhibited high detection abilities for light monthly precipitation events (0–10 mm) and low detection abilities for heavy monthly precipitation events (150–200 mm). This study underscored the potential advantages of GPDs in the three dimensions, and also highlighted their limitations in arid areas, winter, and heavy precipitation events, thus providing directions for further GPDs' accuracy improvement.