A 20-meter root-zone soil moisture dataset using Earth Observations and water balance modelling

At the interface of the surface energy balance and land surface hydrology, soil moisture is a key descriptor for hydrology, ecosystem dynamics and climate variables. Monitoring of soil moisture is of high value in multiple contexts, including water resources management through irrigation detection and quantification, and in land management and climate initiatives. In Denmark, drained organic low-lying soils represent 7% of agricultural land but are responsible for around 50% of the greenhouse gas emissions from agriculture. Soil moisture monitoring is key to prioritizing the decommissioning of relevant farmland. Soil moisture can be observed from space with radar or radiometer instruments; however, applications are often limited by the sensor penetration depth, which restricts the vertical spatial resolution to the top few centimeters of the soil. Spatial sampling is often in the order of kilometers, and too coarse for field-scale screening.

In this study, we propose to use the widely applied FAO-56 soil water balance model for crop evapotranspiration (ET) estimates and use it to estimate soil moisture in the root zone through reverse modelling. Using EO estimates of ET at high resolution derived from Sentinel-2 optical images, downscaled Sentinel-3 thermal images, and the TSEB (Two-Source Energy Balance) ET model, we derive a map of soil moisture in Denmark at 20 m daily resolution. Precipitation is obtained from ECMWF Era-5 and the OpenLand soil texture and characteristics dataset is used to parameterize the soil column. Landuse information at parcel scale from the Danish Agricultural Agency is used to estimate the root zone depth along with time series of Leaf Area Index. The approach is based on optical observations which have limited applicability in cloudy regions and winter months. We therefore apply gap filling using temporal interpolation of the ET time series or assumptions on soil moisture conditions to obtain the final decadal and monthly soil moisture time series.

The map is validated against probe measurements of soil moisture at 21 different sites across Denmark. Overall, the RMSE is around 5% m3/m3 and spatio-temporal patterns are well captured. The main limitations can be attributed to the soil parameterization as well as uncertainties in the coarser climate forcing. This study presents a unique dataset at national scale using publicly available datasets and combining Earth Observations with physical and conceptual modelling to obtain a key hydrological and biophysical parameter. The approach can be extended to most farm and grasslands and can be adjusted, where more precise local parameterization is available. The dataset as presented here, is used to inform a screening and management tool for the Danish Environmental Protection Agency to evaluate the impact of decommissioning low-lying farmlands, and to support research efforts in quantifying nitrogen dioxide emissions from poorly drained soils.