The potential and uncertainty of triple collocation in assessing satellite precipitation products in Central Asia

Although satellite-based precipitation estimation has extensive application potential, validation of its reliability is challenging for areas lacking ground-based data which is particularly true for many arid and semiarid regions. The triple collocation (TC) method can be used to evaluate three independent inputs with unknown true values, and thus provides an appealing alternative for assessment of satellite precipitation products in lack of observation regions. This study is the first to utilize TC to comprehensively assess the uncertainties of various satellite precipitation products in Central Asia (CA) with a distinctive continental arid and semi-arid climate. TC requires the errors of inputs to be independent with each other, while many multi-satellite precipitation products use overlapped data sources. To address this problem, this study uses a soil moisture-based product (SM2RAIN) and a reanalysis model-based product (ERA5) as two inputs of a triplet with the last input coming from each one of the six satellite precipitation products (3B42, CHIRPS, CMORPH, GSMaP, IMERG, PERSIANN). Six independent triplets are obtained in this way. The temporal/spatial resolution is daily/0.1° and the period is 2007–2019. The results show the overall performance of GSMaP is best among eight gridded precipitation products over CA, followed by IMERG, CMORPH and PERSIANN. All precipitation products show degraded performance with increasing altitude. Moreover, the accuracy estimates are subjected to uncertainties caused by the TC method and data inputs. Overall, the study concludes the TC method can provide a new perspective for the assessment of precipitation products over data-absent arid and semiarid regions, while careful check and explanation of evaluation results are always necessary to defend the rationality of TC in specific cases.