Modelling soil bulk density at the landscape scale and its contributions to C stock uncertainty

[摘要] Soil bulk density (D_b) is a major contributor to uncertainties inlandscape-scale carbon and nutrient stock estimation. However, it is timeconsuming to measure and is, therefore, frequently predicted using surrogatevariables, such as soil texture. Using this approach is of limited value forestimating landscape-scale inventories, as its accuracy beyond the samplingpoint at which texture is measured becomes highly uncertain. In this paper,we explore the ability of soil landscape models to predict soil D_busing a suite of landscape attributes and derivatives for both topsoil andsubsoil. The models were constructed using random forests and artificialneural networks.

Using these statistical methods, we have produced a spatially distributedprediction of D_b on a 100 m × 100 m grid, which was shown tosignificantly improve topsoil carbon stock estimation. In comparison tousing mean values from point measurements, stratified by soil class, wefound that the gridded method predicted D_b more accurately, especiallyfor higher and lower values within the range. Within our study area of theMidlands, UK, we found that the gridded prediction of D_b produced astock inventory of over 1 million tonnes of carbon greater than thestratified mean method. Furthermore, the 95% confidence intervalassociated with total C stock prediction was almost halved by using thegridded method. The gridded approach was particularly useful in improvingorganic carbon (OC) stock estimation for fine-scale landscape units at whichmany landscape–atmosphere interaction models operate.

[学科分类] 地球化学与岩石