Investigating the spatio-temporal variation of vegetation water content in the western United States by blending GNSS-IR, AMSR-E, and AMSR2 observables using machine learning methods

Vegetation water content (VWC) is an important parameter when evaluating vegetation growth and climate change. The Global Navigation Satellite System Interferometric Reflectometry (GNSS-IR) technique has provided us with an effective approach for the monitoring of VWC. The normalized microwave reflection index (NMRI) was defined to reflect the change of VWC, based on the fact that the amplitude of the direct and reflected GNSS interferometric signal is related to the variation of VWC. However, the sparse distribution of the ground-site-based observation stations restricts the application of NMRI. Fortunately, microwave vegetation optical depth (VOD) is a dimensionless parameter that describes the rate of attenuation of microwaves as they pass through the vegetation canopy, which can provide spatially continuous vegetation information. In this study, we integrate the GNSS-IR NMRI dataset with Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E) and Advanced Microwave Scanning Radiometer 2 (AMSR2) VOD datasets by using three machine learning models: the generalized regression neural network (GRNN) model, the back-propagation neural network (BPNN) and multiple linear regression (MLR) models. This method can overcome the drawbacks of sparse distribution and spatial discontinuity of GNSS-IR site data, and obtains spatio-temporally continuous NMRI products that can more intuitively and accurately reflect the variability of VWC in the study area. The results showed that the GRNN model has the best retrieval accuracy, with model fitting and cross-validation R values of 0.76 and 0.74, and root-mean-square error (RMSE) values of 0.034 and 0.035, respectively. Based on the GRNN model, we obtained a spatio-temporally continuous NMRI product to further analyze the changes in VWC from 2007 to 2018. The results indicated that the effects of ENSO events with different intensities on vegetation and precipitation are different, and the response of VWC with different vegetation types to ENSO are also different. The anomalies of VWC show a lag phenomenon and is more stable, not as sensitive as precipitation to the ENSO events.