Assessing sampling and retrieval errors of GPROF precipitation estimates over The Netherlands

Abstract. The Goddard Profiling algorithm (GPROF) converts radiometer observations aboard Global Precipitation Measurement (GPM) constellation satellites to precipitation estimates. Analyzing the accuracy of GPROF’s estimates is vital to further improve the algorithm. Such analyses often use high-quality ground-based estimates as reference with a different spatial resolution. Often, the reference is resampled to match the satellite’s resolution. However, the implemented sampling method to simulate the satellite’s resolution varies amongst studies, which limits the transferability of conclusions. additionally, GPROF combines observations from various sensors and frequency channels, each with its own footprint size. Hence, uncertainties related to sampling are added on top of the uncertainty introduced when converting brightness temperatures to precipitation intensities. The contribution of sampling to the total amount of uncertainty remains unknown.

Here, we quantify the uncertainty related to sampling while analyzing the current performance of GPROF over the Netherlands during a four year period (2017–2020). In this area, shallow and light precipitation frequently occur. Both precipitation types are often subject to research, as both types are difficult to detect with space-borne sensors. Only GPROF estimates based on observations from the conical-scanning radiometers of the GPM constellation are used. We investigate the uncertainty related to sampling by simulating the reference precipitation as satellite footprints that vary in size, geometry, and applied weighting technique. The reference estimates are gauge-adjusted radar precipitation estimates from two ground-based weather radars from the Royal Netherlands Meteorological Institute (KNMI). Echo top heights (ETH) retrieved from the same radars are used to classify the precipitation as shallow, medium, or deep.

The method used to spatially average the reference into a satellite footprint, i.e. using Gaussian weighting or the arithmetic mean, is found to exhibit a minimal influence on the retrieved estimate. The size of the sampled area is found to be the most influential. Still, the effect of using different footprint sizes cannot explain all the differences between the ground- and satellite based precipitation estimates. Additionally, the discrepancies between GPROF and the reference are largest for low ETH, while the relative bias between the different footprint sizes and implemented weighting methods increase with increasing ETH. Lastly, our results do not show a clear difference between coastal simulations and simulations over land. We conclude that the uncertainty introduced by merging different channels and sensors cannot fully explain the errors introduced by the retrieval algorithm. Hence, retrieval errors are found to be more prominent than sampling uncertainties, in particular for shallow and light precipitation.