Advancing remote sensing of soil moisture: exploring novel spatial scales with innovative sensors and platforms

Soil moisture content (SMC) is an important parameter that regulates a variety of processes in the soil-plant-atmosphere continuum. A spatially and temporally comprehensive observation of SMC is therefore important for many applications such as agriculture or climate modeling. Each application has specific requirements for the spatial and temporal resolution as well as the depth of SMC measurements, which current measurements and remote sensing methods only partially meet. In-situ measurements, which only represent the SMC of a very small soil volume, are widely used. However, they are not sufficient to map the heterogeneity of SMC patterns on larger spatial scales. In contrast, remote sensing provides global SMC products at a coarse spatial but high temporal resolution. However, approaches for an area-wide SMC monitoring with high spatial resolution are still needed. Furthermore, the derivation of SMC using remote sensing observations is impaired by their scale difference to in-situ SMC measurements. This work addresses these two challenges by exploring possibilities to infer SMC at the missing spatial scales using new measurement methods, such as Cosmic Ray Neutron Sensing (CRNS), or new sensor platforms, such as drones. The high spatial resolution of drone-based remote sensing significantly reduces the discrepancy between remote sensing observations and in-situ measurements. The widely used thermal infrared (TIR) cameras offer a suitable solution for the drone-based derivation of SMC. However, the camera technology used for drones poses challenges in the processing and acquisition of data. With an evaluation of 56 flights, this work comes up with guidelines for TIR data acquisition and processing to ensure successful image alignment. Additionally, it develops a method to quantify the time-varying relative uncertainty of the data. This work therefore takes a first step towards standardizing the acquisition and processing of TIR data and outlines further necessary steps that need to be clarified before their application for the derivation of SMC. Other widely used sensors for drone-based remote sensing fall into the area of spectral remote sensing. Due to its high spatial resolution, spectral remote sensing is also very interesting for satellite-based derivation of SMC. It uses the fluctuating brightness of soils and plant characteristics as proxies for the derivation of SMC. However, the extent to which spectral remote sensing can map SMC across hydro-climatic gradients and management practices is an ongoing debate. In this context, using drone- and satellite-based remote sensing and a machine learning model, this work establishes that spectral remote sensing can only infer SMC in the VIS-NIR range under water-limited conditions. An additional SWIR band, however, extends the range of applicability of spectral remote sensing along hydro-climatic gradients. The performance of models for the derivation of SMC is impaired in particular by grazing and their generalizability by mowing. A data-driven derivation of SMC should therefore always be critically examined with regard to the underlying relationships in order to ensure the generality of the models. Data-driven models require large amounts of data, which are limited, in particular for UAS-based remote sensing. Therefore, this thesis presents a physics-based method for the derivation of SMC. Since current radiative transfer models are based on soil reflectance and do not represent the dynamics between plant physiological processes and SMC, their applicability is constrained to areas with low vegetation cover. Hydro-climatic conditions, on the other hand, have no influence on the quality of the results. Another way to reduce the scale difference between remote sensing observations and in-situ measurements is to increase the support volume represented by the SMC measurement. CRNS is a measurement method that fulfills these requirements. The support volume represented in the CRNS measurement varies in depth and extent depending on the degree of wetness of the environment. This work investigates the possibility of remote sensing observations to capture this spatio-temporally variable measurement using random forest regression. With this approach this thesis provides a better understanding of the characteristics of remote sensing signals with respect to penetration depth and their dependence on hydro-climatic conditions. The results show that spectral remote sensing can follow the temporal-spatially varying SMC measurements well and indicate an equally fluctuating SMC signal within the spectral observation data. Synthetic Aperture Radar (SAR), on the other hand, can only detect SMC during wet conditions due to its low penetration depth. Both types of remote sensing observations strongly benefit from additional information on soil properties and topography. The results of this dissertation provide important implications for the derivation of highresolution SMC using remote sensing. It clearly shows that synergies between disciplines must be utilized to derive SMC with high spatial and temporal resolution across different land use types. This requires interdisciplinary cooperation to advance satellite- and drone-based remote sensing of SMC.