Physical parameters of peat and other organic soils can be derived from properties described in the field

Knowledge about the bulk density and porosity of peat and other organic soils is of major importance, as both parameters directly or indirectly effect hydrological conditions (e.g., soil moisture, water level fluctuation), soil physical processes (e.g., shrinkage, swelling, subsidence) and biological processes (e.g., peat mineralization, peat growth). The agricultural usability (e.g., trafficability, plant growth, yield) and the rewetting and oscillation capacity of peatlands also strongly depend on soil hydraulic properties and, thus, on bulk density and porosity. Additionally, knowledge of bulk density is necessary to convert concentrations (e.g., soil organic carbon content) into volume-related quantities. Bulk density and porosity depend on the botanical origin of the peat, the degree of decomposition and other pedogenetic processes. These soil characteristics can be identified directly during soil examinations in the field. In contrast, the determination of bulk density and porosity requires volume-based sampling and subsequent laboratory analyses. Here, we present pedotransfer functions for peat and other organic soils to derive bulk density and porosity using random forest models. Based on a dataset from approximately 600 horizons from 100 peatland sites in Germany and other European countries, we built a set of different pedotransfer functions combining predictor variables determined in the field. These included the degree of decomposition, peat type (e.g., Sphagnum peat, Carex peat, amorphous peat), horizon characteristics (e.g., aggregated, oxidized, permanently saturated, ploughed), average horizon depth, rooting intensity (no roots to extremely dense, estimated from root proportion per cm²), admixture of mineral compounds and the occurrence of carbonate (estimated using 10% hydrochloric acid). Further pedotransfer functions were built, using soil organic carbon content as an additional predictor variable. The results show that bulk density and porosity can be predicted using only a few predictor variables (3-7) with a low bias and high coefficient of determination. Adding soil organic carbon content as an additional predictor variable further improved the pedotransfer functions. Depending on the combination of the predictor variables, root mean square errors (5-fold cross validation) varied between 0.069 to 0.099 g cm-3 for the bulk density and 3.8 to 4.7% for the porosity pedotransfer functions.