Estimation of soil texture in areas of different parent materials by using gamma ray spectrometry at local and regional scale

Good knowledge on soil texture is a base for a sound land use management as one of the driving factors for soil fertility, and thus for a worldwide sustainable food production and safe drinking water supply using groundwater resources. Furthermore, soil texture is one of the driving factors of soil fertility. In particular, in countries of the Global South but also in other regions, high quality digital information on soil properties on regional level are rather scarce. While conventional soil inventories are time consuming, digital mapping of soil properties is a promising approach to close these gaps time and cost efficiently. For this purpose, a reliable method was developed within the project “ReCharBo” (Regional Characterisation of Soil Properties) to minimize field and laboratory work by combining gamma ray spectrometry with conventional soil survey at selected study areas characterized by different soil parent materials. Data acquisition was performed by using a portable gamma ray spectrometer and soil sampling at local scale as well as by helicopter-borne gamma ray spectrometry at regional level.

For the estimation of soil texture by gamma spectral data at different soil parent materials we developed classic multiple linear regression models based on  laboratory analyses of grain size and total Potassium and Thorium contents. To consider the soil parent materials, we calculated clay-, silt- and sand-Potassium-Ratios based on laboratory data and integrated them into the models. The statistical models were validated by dividing the data set randomly fifty-fifty into a training, and a validation data set. The results on the validation data set show that soil texture can be predicted with an error (RMSE) of 5.8% (Clay), 5.5% (Silt) and 4.6% (Sand) by gamma ray spectrometry. Based on these models, soil texture can reliably be estimated by gamma ray spectrometry accompanied by a scarce soil sampling in regions with poor data.