Soil moisture dynamics and its spatial distribution represent important data of a landscape. Until today, a precise and spatially continuous measurement of these is difficult. Current remote sensing approaches focus on an upscaling of measured hyperspectral and hydrological data from laboratory setups by means of hydrological models. In this paper, we follow an alternative approach on a pedon-scale. We present a multilateral dataset from an irrigation campaign in August 2017 employing a comprehensive hydrological, geophysical, and hyperspectral sensor setup. In a first regression modeling approach, we apply machine learning methods to estimate subsurface soil moisture based on hyperspectral data and hydrological reference data. The derived results reveal the potential of estimating field- measured soil moisture. They provide the basis for further studies of modeling approaches based on this unique dataset.