Multi-Scale Distributed Sensitivity Analysis of Radiative Transfer Model

Abstract:

Amidst nature’s great variability and complexity and Soil Moisture Active Passive (SMAP) mission aims to provide high resolution soil moisture products for earth sciences applications. One of the biggest challenges still faced by the remote sensing community are the uncertainties, heterogeneities and scaling exhibited by soil, land cover, topography, precipitation etc. At each spatial scale, there are different levels of uncertainties and heterogeneities. Also, each land surface variable derived from various satellite mission comes with their own error margins. As such, soil moisture retrieval accuracy is affected as radiative model sensitivity changes with space, time, and scale. In this paper, we explore the distributed sensitivity analysis of radiative model under different hydro-climates and spatial scales, 1.5 km, 3 km, 9km and 39km. This analysis is conducted in three different regions Iowa, U.S.A (SMEX02), Arizona, USA (SMEX04) and Winnipeg, Canada (SMAPVEX12). Distributed variables such as soil moisture, soil texture, vegetation and temperature are assumed to be uncertain and are conditionally simulated to obtain uncertain maps, whereas roughness data which is spatially limited are assumed a probability distribution. The relative contribution of the uncertain model inputs to the aggregated model output is also studied, using various aggregation techniques. We use global sensitivity analysis (GSA) to conduct this analysis across spatio-temporal scales.

Keywords: Soil moisture, radiative transfer, remote sensing, sensitivity, SMEX02, SMAPVEX12.