Validation on the content of clay and organic matter of a digital soil map across Denmark - A median-soil texture perspective

Detailed soil maps are an essential tool for water resource, land management and agricultural planning. However, due to different soil processes and variation in geology, soil textural properties vary in space at different scales challenging the development of accurate soil maps. Despite the advances in sampling and efforts to produce accurate maps, uncertainties remain at many scales. Therefore, by implementing error evaluation methods such as comparing predictions and observations, it is possible to understand the performance of digital soil maps. The aim of this work was to obtain the uncertainties of soil categories from a Danish national soil map (AC map) based on measured contents of soil texture and soil organic matter. The AC maps describes the horizon characteristics (soil texture, soil organic matter, soil horizon depths and bulk density) at three depths corresponding to the A (0-30 cm), B (30-70 cm), and C (70-120 cm) horizon. In addition, it combines information on soil classification (soil texture), geology at a depth of ca. 1.5 meters as well as the national geological region. The map has a spatial resolution of 250 meters in the A and B horizons and 500 meters in the C horizon. The analysis is based on 38,000 textural points for the A horizon, 7,000 points for the B horizon, and almost 1,700 points for the C horizon. Considering mean and medians of the content of clay and organic matter together with the observed data, the AC map was validated. The uncertainties of the AC map, statistical correlation coefficient (R²), root mean square error (RMSE), and the Nash-Sutcliffe efficiency (NSE) were used. The results showed that the AC map has an acceptable performance when predicting the clay content in the A horizon (R² = 0.97, RMSE = 1.15%, and NSE = 0.91), B horizon (R² = 0.95, RMSE = 1.85%, and NSE= 0.8), and C horizon (R² = 0.74, RMSE = 5.32%, and NSE = 0.41, respectively. The performance of the prediction of organic matter content in the A horizon (R² = 0.84, RMSE = 0.39%, and NSE = 0.65) and B horizon (R² = 0.66, RMSE = 0.44%, and NSE = 0.32) horizons was acceptable as well. Also, the result of validation showed that the highest residual errors of AC maps for clay content and organic matter content has been related loamy (>15% clay) and peat soils. In conclusion, the AC maps, with its optimal accuracy especially in the A and B horizons, can be a suitable tool for use at variable scales in the analysis of crop growth and nitrate leaching modelling studies.

Keywords: Soil texture, digital soil maps, soil variability, soil organic matter, clay, soil properties