Digital soil mapping for modelling of transport pathways for pesticides to surface water
2015
Piikki, Kristin | Söderström, Mats | Wetterlind, Johanna | Stenberg, Bo | Jarvis, Nicholas
Detailed spatial information of soil texture and SOM content are needed for the modelling of pesticide fate. As part of the SLF project Transport pathways for pesticides to surface water (SLFH1133108) such maps were derived for three depth layers of the soil profile in two sub-catchments with a total area of 450 ha. Data from proximal sensor measurements were combined with national datasets (a digital elevation model and a Quaternary deposit map) and calibrated against soil sample data. Different modelling strategies were tested and the best strategy for each layer was selected for production of the final maps. For topsoil texture, Marsplines modelling with a reduced number of covariables gave the best results (the final models included topographic variables and gamma radiation). In the upper subsoil, the texture was best predicted by an empirical relationship to the topsoil predictions. This is not surprising since the number of calibration samples in the subsoil was very small (n = 18) compared to the number of calibration samples (n = 85). In the deep subsoil, the best strategy was again to use Marsplines modelling with a reduced number of covariables. The SOM content in the topsoil and was mapped by Marsplines modelling using a limited number of predictors and allowing interactions. In the subsoil, however, all soil samples but one had SOM < 3 % and the entire area was classified as as SOM class u. Organic soils according to the SGU QD map (peat areas) were not included in the mapping. Instead they were directly assigned as SOM class t and the texture was not determined.
显示更多 [+] 显示较少 [-]