Accuracy Assessment And Merging Calibration For IMERG Precipitation Products Over Mainland China

Precipitation is a key part of the global water cycle,connecting the atmospheric and surface processes,and plays an important role in the material and energy cycle.Accurate precipitation data and its spatio-temporal distributions are of great significance to the related fields,such as hydrology,meteorology,climatology and agricultural science.Observations based on rain gauges and weather radars are conventional approaches to obtain precipitation information.However,due to the limited number and uneven spatial distribution,the ground stations cannot reflect the spatio-temporal distributions and variations of large-scale precipitation.Currently,retrieval technology based on satellite-based remote sensing has become the only mean of access to obtain wide range of spatio-temporal continuous precipitation information.The successful launch of the Global Precipitation Measurement(GPM)core observatory has made the GPM level-3 precipitation product Integrated Multi-satellit E Retrievals for Global Precipitation Measurement(IMERG),which is produced based on the dual-frequency precipitation radar and various satellite estimates and algorithms,widely concerned and applied.And IMERG provides high-quality precipitation information with fine resolution for the new generation of global precipitation observation.The latest version of IMERG,named V06 B,was produced in June 2019.Multiple steps have been optimized and adjusted in the retrieval algorithm.Before indepth application of the new generation IMERG precipitation product,it is necessary to evaluate the accuracy in specific areas to reveal the spatio-temporal distribution characteristics,error structures,and the ability to capture precipitation events of the product.Based on ground observations,this study evaluated the accuracy of the latest generation of IMERG and Fengyun-2G(FY2G)quantitative precipitation estimates from Chinese meteorological operational satellite over mainland China.According to the characteristics of the precipitation data reflected in the accuracy assessment and the deficiency of the IMERG retrieval algorithms,IMERG precipitation data was calibrated using Asian Precipitation Highly Resolved Observational Data Integration(APHRODITE)with high precision and SM2RAIN-ASCAT with the variation of soil moisture,and finally obtained IMERG-Pixel based adjustment using APHRODITE and SM2RAIN-ASCAT(IMERG-PBAAS)with high precision and fine spatio-temporal resolution.The major works and conclusions are described as follows:(1)The research on the accuracy assessment of the spatio-temporal distribution characteristics of precipitation products over mainland China.In this study,the spatial distributions of the amount of precipitation and indicators of IMERG and FY2 G QPE were studied at different time scales.The spatial trend of the two precipitation products was similar compared with the ground observations,showing a decreasing trend from the southeast to the northwest.IMERG had complete spatial coverage,while FY2 G QPE had incomplete spatial coverage.The spatial distributions of the indicators showed that the indicators of IMERG and FY2 G QPE in northwestern China and southwestern China were generally inferior to other regions;except for individual areas,the hourly and daily scale indicators of FY2 G QPE were better than the indicators of IMERG;IMERG generally overestimated the precipitation at meteorological scale,and had significant tendency of false negatives and false alarms.In this study,the temporal variations of the amount of precipitation and indicators of IMERG and FY2 G QPE were studied at different time scales.The curves of IMERG and FY2 G QPE had similar variation characteristics in each sub-region,while IMERG showed overestimates in 2018 and FY2 G QPE exhibited underestimates to a lesser degree.IMERG had good accuracy performance at monthly scale.The indicators were even better than FY2 G QPE in some months,indicating that IMERG had good application prospects at climatological scale.Hourly indicators of FY2 G QPE were better than those of IMERG in all periods of 24 hours.(2)The research on the accuracy assessment based on numerical characteristics of indicators and precipitation events.In this study,the numerical distribution characteristics of IMERG and FY2 G QPE at hourly and daily scales and the accuracy of the precipitation products under different precipitation events were investigated.The average value of the two types of precipitation products at daily scale was better than the average value at hourly scale;the mean indicators of FY2 G QPE at hourly and daily scales were better than the mean indicators of IMERG.IMERG had a large room to improve in reducing the overestimates and underestimates,as well as false negatives and false alarms at meteorological scale.In heavy precipitation events brought by typhoon and continuous precipitation events caused by plum rains,IMERG and FY2 G QPE accurately captured the spatial distribution of the amount of precipitation,while IMERG had significant overestimates and underestimates in some areas;even though there were deviations in certain periods,the two types of precipitation products had similar temporal variations compared with the ground observations,and the deviations of IMERG were more significant than the deviations of FY2 G QPE.(3)The research on the multi-rule precipitation data merging and calibration algorithm.According to the characteristics of IMERG and FY2 G QPE revealed by the accuracy assessment,the feasibility of merging and calibration of IMERG precipitation data was explored.IMERG had good accuracy performance at monthly scale,but the accuracy performance at meteorological scale was unstable.However,IMERG had excellent spatio-temporal coverage and resolution.In addition,FY2 G QPE showed excellent estimation accuracy and precipitation capture ability at meteorological scale due to the fusion of meteorological ground observation.Therefore,merging and calibration of IMERG precipitation data was feasible.This study developed a multirule precipitation data merging and calibration algorithm based on spatial pixel matching by identifying the spatio-temporal resolution characteristics of the reference datasets.The algorithm finally output high-precision calibrated precipitation data IMERG-PBAAS with the spatial resolution of 0.1 ° and the temporal resolution of 0.5h.The spatial distribution trend of daily average precipitation of IMERG-PBAAS was almost the same as the original IMERG,while IMERG-PBAAS had lowered the estimates of the amount of precipitation in the study area except the northwestern region and the spatial distribution variations became more detailed.By comparing the temporal variations of the precipitation datasets before and after applying the calibration algorithm,it could be found that: compared with the value of the original IMERG,the average precipitation of IMERG-PBAAS in the entire year showed significant degree of decline;according to the comparison of typical time periods and typical sites,IMERG-PBAAS greatly improved the overestimation and underestimation of the original IMERG,and became more consistent with the ground observations.Based on the ground observations from meteorological stations,the accuracy evaluation indicators were calculated and it could be found that the correlation coefficient and detection probability of IMERG-PBAAS improved to a certain extent but not significant at hourly scale,while the performance of the remaining indicators improved significantly;the performance of indicators at daily scale improved more significantly;the increase of RMSE and FAR at hourly scale was obvious,indicating that the reference data played an important role in the control of total amount of precipitation.By comparing the performances in the heavy precipitation event brought by typhoon,it could be found that IMERG-PBAAS had the better ability to capture the spatial distribution of precipitation than the original IMERG,making it more consistent with the ground observations.IMERG-PBAAS significantly reduced the underestimates of the original IMERG during the typhoon heavy precipitation event.Meanwhile,the value of correlation coefficient and root mean square error also improved significantly.This study conducted a preliminary exploration to reveal the spatio-temporal distribution characteristics,numerical characteristics,and error structure of the new generation of IMERG precipitation data over mainland China.In addition,it provided the reference for source data selection and retrieval algorithm calibration of IMERG precipitation data in the next generation.