On the Spatial and Temporal Sampling Errors of Remotely Sensed Precipitation Products

Abstract

Observation with coarse spatial and temporal sampling can cause large errors in quantification of the amount, intensity, and duration of precipitation events. In this study, the errors resulting from temporal and spatial sampling of precipitation events were quantified and examined using the latest version (V4) of the Global Precipitation Measurement (GPM) mission integrated multi-satellite retrievals for GPM (IMERG), which is available since spring of 2014. Relative mean square error was calculated at 0.1 • × 0.1 • every 0.5 h between the degraded (temporally and spatially) and original IMERG products. The temporal and spatial degradation was performed by producing three-hour (T3), six-hour (T6), 0.5 • × 0.5 • (S5), and 1.0 • × 1.0 • (S10) maps. The results show generally larger errors over land than ocean, especially over mountainous regions. The relative error of T6 is almost 20% larger than T3 over tropical land, but is smaller in higher latitudes. Over land relative error of T6 is larger than S5 across all latitudes, while T6 has larger relative error than S10 poleward of 20 • S-20 • N. Similarly, the relative error of T3 exceeds S5 poleward of 20 • S-20 • N, but does not exceed S10, except in very high latitudes. Similar results are also seen over ocean, but the error ratios are generally less sensitive to seasonal changes. The results also show that the spatial and temporal relative errors are not highly correlated. Overall, lower correlations between the spatial and temporal relative errors are observed over ocean than over land. Quantification of such spatiotemporal effects provides additional insights into evaluation studies, especially when different products are cross-compared at a range of spatiotemporal scales.

Validation of GPM IMERG Extreme Precipitation in the Maritime Continent by Station and Radar Data

“…Several studies evaluated the uncertainties related to the sampling of precipitation measurements when estimating areal precipitation (Behrangi & Wen, 2017; Tang et al., 2018; Tian et al., 2018; Villarini et al., 2008). Here, the spatial sampling error is estimated by comparing the native resolution Subang radar precipitation (on a 0.0045° grid) against itself, but regridded onto the coarser 0.1° IMERG grid.…”

“…However, these studies were subject to potentially large spatial sampling errors, that is, errors incurred when interpolating gauge precipitation data onto the IMERG grid. By degrading the same precipitation product onto different spatio-temporal resolutions, Behrangi and Wen (2017) showed that these errors can be large, especially over land areas. Similarly, Tian et al (2018) and Tang et al (2018) found that rain gauge density has a large impact on IMERG skill metrics over China.…”

Performances of GPM satellite precipitation over the two major Mediterranean islands

“…The GPM Core Observatory was launched on February 28, 2014 by a joint effort of NASA and the Japan Aerospace Exploration Agency (JAXA). Since then, several studies have examined whether the rainfall products from the GPM mission are capable to accurately estimate global precipitation [8,9,10,11,12] and provide the order of magnitude of discrepancies between observations and satellite products highlighting the reasons for such problems [13,14,15].…”

A daily 25 km short-latency rainfall product for data-scarce regions based on the integration of the Global Precipitation Measurement mission rainfall and multiple-satellite soil moisture products

“…The optimal linear combination (OLC) approach (Bishop and Abramowitz, 2013;Hobeichi et al, 2018) provides an analytically optimal linear combination of ensemble members (rainfall estimates in this case) that minimizes the mean square error when compared to a dataset that is assumed to be accurate enough to be considered as a calibration dataset Y REF and thus accounts for both the performance differences and error covariance between the rainfall products. The optimal linear combination is therefore insensitive to the addition of redundant information.…”

“…The integration method we adopted is the optimal linear combination (OLC) approach (Bishop and Abramowitz, 2013;Hobeichi et al, 2018), which is based on a technique that provides an analytically optimal linear combination of rainfall products and accounts for both the performance differences and error covariance between the products. We tested the performance of the product (1) over four key regions, namely, India (IN), the conterminous United States (CONUS), Australia (AU) and Europe (EU), where high-quality ground-based hydrometeorological networks are available, and (2) in Africa and South America by using a triple-collocation (TC) analysis (Stoffelen, 1998).…”