Comparisons of Satellite Precipitation Estimates over the United States Affiliated Pacific Islands (USAPI)

Accurate measurement of precipitation is no trivial task. Efforts have been made to increase the density of rain gages to provide greater spatial coverage to increase our understanding of precipitation variation especially in sparse regions. However, there are still many places that are not covered, such as the United States Affiliated Pacific Islands (USAPI). To help fill these gaps, precipitation is being measured and estimated using satellite sensors. In this study, we compare precipitation estimates from several satellites and missions with each other and with in-situ station precipitation measurements across the USAPI. This data sparse region is made up of many small, topographically dynamic islands that are heavily dependent on precipitation for their fresh water resources. It is an ideal region to compare precipitation estimates and determine whether satellite sensors are able to accurately estimate precipitation in complex regions. This is especially true because many of the islands are smaller than the spatial resolution of the satellite products, thus further complicates our analyses. In this study, precipitation estimates from the NOAA Climate Data Record (CDR) of Precipitation Estimation from Remote Sensing Information using Artificial Neural Network (PERSIANN), the NOAA Climate Prediction Center (CPC) Morphing Technique (CMORPH), the Tropical Rainfall, Measuring Mission (TRMM), and the Global Precipitation Measurement (GPM) were compared both spatially and temporally with each other and in-situ station estimates. Analysis was done to determine 1) Spatial differences between precipitation estimates, 2) Temporal differences between precipitation estimates, 3) Whether precipitation estimates of smaller versus larger islands were more accurate, and 4) Whether precipitation was accurately estimated for topographically complex islands. Mean daily, monthly, and yearly precipitation estimates were compared by pixel location with each other and with the closest in-situ station. This provided a validation analysis for each precipitation estimate and an accuracy analysis based on pixel agreement.