Day 4

Detailed paper information

Back to list

Paper title REMOTE SENSING OF SOIL MOISTURE FOR AGRICULTURAL AREAS USING SPATIAL AND TEMPORAL HIGH-RESOLUTION SENTINEL-1 SAR TIMESERIES IN GOOGLE EARTH ENGINE
Authors
  1. David Mengen Forschungszentrum Jülich, Institute of Bio-and Geosciences: Agrosphere (IBG-3) Speaker
  2. Carsten Montzka Forschungzentrum Jülich GmbH
  3. Thomas Jagdhuber German Aerospace Center (DLR) - Microwave and Radar Institute
  4. Anna Balenzano Consiglio Nazionale delle Ricerche (CNR)
Form of presentation Poster
Topics
  • A7. Hydrology and Water Cycle
    • A7.01 Inland Water Storage and Runoff: Modeling, In Situ Data and Remote Sensing
Abstract text Agricultural systems are the main consumers of freshwater resources at global scale, using 60 % to 90 % of the total available water. While the growing demand for agricultural products and the resulting intensification of their production will increase the dependency on available freshwater resources, this sector will become even more vulnerable because of the intensifying impacts of climate change. Detailed knowledge about soil moisture, being a key parameter in the agricultural sector, can help to mitigate these effects. Nevertheless, spatial and temporal high resolution surface soil moisture data for regional and local monitoring (down to precision farming level) are still challenging to obtain. By using current as well as future Synthetic Aperture Radar (SAR) satellite missions (e.g. Sentinel-1, ALOS-2, NISAR, ROSE-L), this knowledge gap can be filled. Providing a cloud- and weather independent monitoring of the Earth's surface, SAR observations are suitable for regional and local soil moisture estimations, but with a global extent. While the increasing resolution and total number of SAR recordings will contribute to an improvement of the estimation in general, the computational costs as well as the local memory capacity on the other hand become a limiting factor in processing the vast load of data. Here, on-demand cloud-based processing services are one way to overcome this challenge. This is especially interesting as most of the severely affected regions have limited access to computational resources.
Using both VV and VH polarization for vegetational detrending as well as low pass filtering, we developed an automated workflow for estimating soil moisture using temporal and spatial high-resolution Sentinel-1 timeseries, based on the alpha approximation approach of Balenzano et al. 2011. The workflow is established within the cloud processing platform Google Earth Engine (GEE), providing a fast and applicable way for on-demand computation of soil moisture for individual time periods and areas of interest around the globe. The algorithm was tested and validated over the Rur catchment, located in the federal state of North-Rhine Westphalia in the West of Germany. With an area of 2,354 km², it comprises a great diversity in agricultural cropping structure as well as topologies. A total of 711 individual Sentinel-1A and Sentinel-1B dual-polarized (VV + VH) scenes in Interferometric Wide-Swath Mode (IW) and Ground Range Detected High Resolution (GRDH) format are used for the analysis from January 2018 to December 2020. Using all available orbits (both ascending and descending), a temporal resolution of one to two days could be achieved with a spatial resolution of 200 m. The workflow includes multiple steps: speckle filtering, incidence angle normalization, vegetational detrending and low-pass filtering. The results were validated against eight Cosmic-Ray Neutron Stations (CRNS), which are evenly distributed over the catchment, covering various types of landcover. In total, the method achieves an unbiased RMSE (uRMSE) of 5.84 % with an R² of 0.46. Looking at individual months, the highest correlation can be achieved in the months April and October with R² values range between 0.65 to 0.7, while the lowest correlation is observed in July and January, with R² values ranging between 0.15 and 0.2. Looking at individual landuse, the method achieves the best results for pastures, with an uRMSE of 0.42 and an R² value of 0.63.