Indicator-Based Soil Moisture Monitoring of Wetlands by Utilizing Sentinel and Landsat Remote Sensing Data

Abstract

Precise and region-wide information about soil moisture with a high spatial resolution is required for river flood plains to fulfil statutory provisions from the EU Water Framework Directive. Influenced by constraints like precipitation and vegetation cover, this complex parameter is often measured point-by-point. Continuous soil moisture products derived from remote sensing are currently available only at coarse spatial resolutions. In this context, plants are of special interest, as their growth and distribution is determined by soil properties. Plant characteristics can be utilized as indicators or proxies of mean soil moisture. This study focuses on the derivation of top soil layer moisture content based on these indicator values by synergistically utilizing different methods and active as well as passive sensors. Subsequently, a joint model was developed to derive a high spatial resolution soil moisture product for the investigated wetlands. This model reached a coefficient of determination of 0.93. The result is valuable to many different user groups like water or nature conservation authorities of all administrative levels. The implementation is mainly based on Sentinel-1 and -2 along with Landsat data sets. The high repetition rate of the Sentinel sensors can increase the classification accuracy of the bio-indicators by a precise mapping of the phenological stages. The results indicate that the method is transferable to other riparian areas to a certain degree. In addition, a time series analysis covering a period of 16 years based on a soil moisture index derived from Landsat data sets was realized. Changes in soil moisture, following restoration measures, emerge clearly and thus allow for a remote sensing-based monitoring of restoration success.