A novel multiscale multisensor data fusion framework for massive datasets-application to soil moisture estimation

Soil moisture (SM) has been identified as a key climate variable governing hydrologic and atmospheric processes across multiple spatial scales. The global burgeoning of SM datasets in the past decade from disparate platforms such as in-situ sensors, active-passive airborne platforms, distributed temperature sensors (DTS), comic-ray neutron probes, global positioning systems (GPS) and satellite platforms such as SMOS and SMAP, holds a significant potential in improving our understanding of multi-scale SM dynamics. The primary issues that hinder the fusion of SM data from disparate instruments are:

1) different spatial resolutions of the data instruments,

2) inherent spatial dependence (or correlation) in SM distribution caused due to atmospheric and land surface controls,

3) measurement errors caused due to imperfect retrievals of instruments and

4) massive size of datasets on a continental scale.

We present a data-driven fusion scheme which takes all the above factors into account using a Bayesian spatial hierarchical model combining a non-stationary geostatistical model with a hierarchical paradigm. The applicability of the fusion scheme is demonstrated by combining point, airborne and satellite data for a watershed exhibiting high spatial variability in Manitoba, Canada. We quantify the effects of controls such as rainfall, vegetation and soil texture on the mean and the variance/correlation of SM as well as on the measurement errors of remote sensing instruments (L-band sensor). The applicability of the fusion scheme, however, is severely limited by massive datasets. Therefore, we next propose an extension of the fusion algorithm using a likelihood approximation and subsequently apply it to combine point SM networks of USCRN and SCAN with coarse satellite data from SMOS and SMAP to predict high resolution SM across contiguous US. We show that the proposed approximation is efficient both in terms of speed and storage with high predictive accuracy. The proposed scheme is data-driven, computationally efficient and can be used to combine SM (and other environmental) data from disparate platforms across scales.