Estimating Organic Surface Horizon Depth For Peat And Peaty Soils Across A Scottish Upland Catchment Using Linear Mixed Models With Topographic And Geological Covariates

In order to evaluate and protect ecosystem services provided by peat and peaty soils, accurate estimations for the depth of the surface organic horizon are required. This study uses linear mixed models (LMMs) to test how topographic (elevation, slope, aspect) and superficial geology parameters can contribute to improved depth estimates across a Scottish upland catchment. Mean (n = 5) depth data from 283 sites (representing full covariate ranges) were used to calibrate LMMs, which were tested against a validation dataset. Models were estimated using maximum likelihood, and the Akaike Information Criterion was used to test whether the iterative addition of covariates to a model with constant fixed effects was beneficial. Elevation, slope and certain geology classes were all identified as useful covariates. Upon addition of the random effects (i.e. spatial modelling of residuals), the RMSE for the model with constant-only fixed effects reduced by 24%. Addition of random effects to a model with topographic covariates (fixed effects = constant, slope, elevation) reduced the RMSE by 13%, whereas the addition of random effects to a model with topographic and geological covariates (fixed effects = constant, slope, elevation, certain geology classes) reduced the RMSE by only 3%. Therefore, much of the spatial pattern in depth was explained by the fixed effects in the latter model. The study contributes to a growing research base demonstrating that widely available topographic (and also here geological) datasets, which have national coverage, can be included in spatial models to improve organic horizon depth estimations.