Prediction of soil hydraulic properties for the extent of Austria | Österreichweite Regionalisierung bodenhydraulischer Eigenschaften
2022
Zeitfogel, Hanna | Feigl, Moritz | Schulz, Karsten
In the light of a changing climate, access to high quality soil information is gaining immense significance. Soil parameters are fundamental for modeling hydrological processes, which is crucial for solving problems like identifying zones at risk of pluvial floods. Currently, there is no single data product available which covers the whole study area and still displays the variability of local soil observations. Thus, the challenge is the combination of soil data from different sources and resolutions and, at the same time, the preservation of the characteristically high spatial variability of soil properties. For the Austrian wide spatial prediction of soil parameters two machine learning (ML) models (XGBoost and FNN) were trained with all available soil data sources and environmental raster datasets. The soil parameters sand, silt, clay and humus were predicted at three different depth levels and a resolution of 1 × 1 km² for the area of Austria. The resulting maps are able to largely reproduce the original data variability. Two approaches were tested for deriving the saturated hydraulic conductivity (ks): Firstly, ks was determined by applying existing pedotransfer functions (PTFs) on the previously regionalized soil parameters. Secondly, ML models were directly trained with available soil hydraulic datasets to predict ks. The prediction of ks includes high levels of uncertainties. The derived soil maps help to reduce current gaps in soil data availability for Austria and act as a basis for identifying zones at risk of pluvial floods.
显示更多 [+] 显示较少 [-]