Mapping soils in arid regions with Sentinel-1

Abstract

Soils are vital for life on Earth, and accurate soil data is crucial for the preservation of this vital resource. Unfortunately, mapping the quality of soils with in situ techniques is arduous due to the large heterogeneity of soils, and suitable monitoring methodologies are expensive or non-existent. Therefore, there has been much interest in using remote sensing data for soil mapping. So far, most of the efforts have focused on using multispectral and hyperspectral images that provide crucial information about the chemical composition of the soil and can be used as indirect proxies in predictive soil mapping. In combination with Artificial Intelligence (AI) techniques, these optical data have enabled a revolution in soil mapping, yielding high-resolution soil maps at continental to global scales. Still, the disadvantage of sensors working in the visible and infrared portion of the electromagnetic spectrum is that they cannot sense subsurface soil properties and are blocked by vegetation, meaning that there is no way that optical data can provide direct information about soil profile properties. This shortcoming can be overcome by using Synthetic Aperture Radar (SAR) sensors that penetrate soil and vegetation to some extent. The penetration depth into the soil is in the order of a wavelength, i.e., a few centimetres to decimetres at commonly used SAR frequencies. However, the penetration varies strongly with the water content of the upper soil layers. Particularly in arid regions, penetration depths may be quite large, allowing to map subsurface features such as bedrock, gravel or distinct soil layers situated several decimeters below the soil surface. In this contribution, we present a novel algorithm for detecting such subsurface features from SAR time series. The algorithm focuses on detecting the distinctive backscatter response of subsurface scatterers in response to varying soil moisture conditions. We will show high-resolution subsurface scattering maps generated from Sentinel-1 backscatter time series over multiple desert regions globally. These maps will be compared with existing soil maps, digital terrain models, and land cover data. Moreover, we will share findings from a recent on-site visit to a desert region near Fossil Rock in the United Arab Emirates. This visit yielded valuable insights into soil properties that are discernible in the subsurface scattering maps derived from Sentinel-1 data.