Bridging of Ground Validation between Gridded and Orbit-Level Rainfall through IMERG

The GPM IMERG product, which provides global precipitation gridded at a high resolution, will likely be widely used in climate and weather research to advance our understanding on precipitation processes. However, its reliability has to be assessed through ground validation so as to quantify uncertainties and potentially improve the product.

In this study, we evaluate IMERG with two different ground-based products. We first compare it against a dense network of dual tipping-bucket gauge platforms in Pocomoke City, MD (38.1°N, 75.6°W), that is highly reliable but limited to an area of approximately a single IMERG grid box. At the same time, we use the gauge network as a means to independently verify the Multi-Radar Multi-Sensor (MRMS) combined radar-rain gauge product, an operational QPE product which provides much wider coverage but with greater uncertainty. Collectively, the simultaneous comparisons of these two ground-based products to IMERG establish a hierarchical and consistent means to assess uncertainties over the grid box as well as larger regional scales.

Over the IMERG grid box containing the gauges, the comparisons between IMERG, MRMS and the gauges have considerable scatter. At this scale and relative to the gauges, MRMS exhibited a normalized absolute error of between 25% to 38%. Against this backdrop, an analysis of varying grid box sizes and accumulation periods using MRMS over a much larger region of the central and southern U.S. showed an expected improvement in IMERG with spatiotemporal scales, with a possible "break point" at 3 hours and spatial scales of order 0.5° to 1°. Associating the errors with individual microwave sensors used in the GPM constellation and hence IMERG suggests the existence of platform-related biases. Our approach demonstrates the use of combined radar and gauge-based QPE over a hierarchy of scales and reliability to effectively quantify the ground validation of IMERG.