Evaluations on gridded precipitation products spanning more than half a century over the Tibetan Plateau and its surroundings

摘要

Long-term precipitation data play significant roles in hydrological applications and climate change studies. In this study three gridded precipitation products (Asian Precipitation-Highly Resolved Observational Data Integration Towards Evaluation of Water Resources, APHRODITE; Climate Hazards Group InfraRed Precipitation with Station data, CHIRPS; and Precipitation Estimation from Remotely sensed Information using Artificial Neural Networks-Climate Data Record, PERSIANN-CDR) with the spatial resolution of 0.25 degrees were evaluated using rain gauge data at the daily, monthly and annual scales over the Tibetan Plateau and its surrounding areas (TP). R-2, bias, RMSE and MAE were adopted to assess precipitation amount; indices including the probability of detection (POD), false alarm ratio (FAR), critical success index (CSI) and frequency bias index (FBI) were used to evaluate gridded products' ability to detect precipitation occurrences. We found that: (1) APHRODITE outperformed CHIRPS and PERSIANN-CDR against observations at the annual scale with the highest R-2 (similar to 0.93), the lowest MAE (similar to 67 mm/year) and RMSE (similar to 116 mm/year), and APHRODITE also showed better performance in distinguishing rain/no rain events with the highest POD (similar to 0.9) and CSI (> 0.6), the lowest FAR (similar to 0.35), followed by PERSAINN-CDR and CHIRPS; (2) CHIRPS performed nearly to APHRODITE in precipitation amount with R-2 around 0.8 while the lowest POD (< 0.4) implied the shortage for CHIRPS in detecting rain/no rain conditions; (3) in spite of the lowest agreement with gauge measurements with the largest bias (similar to 25%), PERSIANN-CDR had a relatively higher mean POD (> 0.5), indicating that PERSIANN-CDR had better potential of detecting rainfall occurrences than CHIRPS but tended to overestimate precipitation; (4) generally, PERSIANN-CDR overestimated the precipitation greater than CHIRPS and APHRODITE data at monthly scale; and (5) the three products correlated well with each other in the interior, southern and eastern regions of TP but there were still some significant differences along the borders and surroundings of the TP by inter-comparisons. Finally, we recommended APHRODITE for climate change study for its high accuracy and long-time series, CHIRPS and PERSIANN-CDR could also be considered after correction.