Evaluation of Satellite Soil Moisture Datasets Using Mutual Information with Precipitation
Abstract
Multiple microwave satellite instruments and retrieval algorithms have been used to estimate surface soil moisture, resulting in soil moisture datasets with a variety of distributions and accuracies. However, evaluation of satellite soil moisture estimates is complicated by the difference in spatial scale between satellite footprints and ground-based soil moisture observations. We use mutual information to assess the accuracy of different soil moisture datasets globally and at large scales, by comparing against precipitation data. Mutual information is a measure of the mutual dependence of two variables, which represents the reduction in uncertainty about a variable given knowledge of a second variable. It is useful for comparing soil moisture and precipitation data because it can accommodate non-Gaussian distributions and non-linear relationships. Larger errors in satellite soil moisture estimates (due to instrument error, atmospheric effects, vegetation effects, or algorithm deficiencies) lead to a weaker pairwise relationship between soil moisture and precipitation, and thus lower mutual information between the two. It is possible to assess the relative accuracy among multiple soil moisture datasets by comparing the mutual information between those datasets and a single precipitation dataset. We evaluate the performance of soil moisture datasets from multiple satellite instruments globally, focusing primarily on SMAP, SMOS, and AMSR2, and using NOAA Climate Prediction Center (CPC) precipitation data. Different soil moisture datasets generated from each satellite instrument are compared, both within and across sensors. We assess the relative accuracy of each soil moisture dataset (i.e. ranking according to mutual information), determine how skill varies geographically, and provide an estimate of the uncertainty in the analysis.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.H53E..03T
- Keywords:
-
- 1843 Land/atmosphere interactions;
- HYDROLOGY;
- 1855 Remote sensing;
- HYDROLOGY;
- 1866 Soil moisture;
- HYDROLOGY;
- 4262 Ocean observing systems;
- OCEANOGRAPHY: GENERAL