Deriving regional pedotransfer functions to estimate soil bulk density in Austria

Abstract

Soil bulk density is a required variable for quantifying stocks of elements in soils and is therefore instrumental for the evaluation of land-use related climate change mitigation measures. Our motivation was to derive a set of pedotransfer functions for soil bulk densities usable to accommodate different levels of data availabilities. We derived sets of linear equations for bulk density that are appropriate for different forms of land-use. After introducing uncertainty factors for measured parameters, we ran the linear models repeatedly in a Monte Carlo simulation in order to test the impact of inaccuracy. The reliability of the models was evaluated by a cross-validation. The single best predictor of soil bulk density is the content of soil organic carbon, yielding estimates with an adjusted R2 of approximately 0.5. A slight improvement of the estimate is possible when additionally, soil texture and soil depth are known. Residual analysis advocated the derivation of land-use specific models. Using transformed variables and assessing land-use specific pedotransfer functions, the determination coefficient (adjusted R2) of the multiple linear models ranged from 0.43 in cropland up to 0.65 for grassland soils. Compared to pedotransfer function, from the literature, the performance of the linear modes were similar but more accurate. Taking into account the likely inaccuracies when measuring soil organic carbon, the soil bulk density can be estimated with an accuracy of +/− 9 to 25% depending on land-use. We recommend measuring soil bulk density by standardized sampling of undisturbed soil cores, followed by post-processing of the samples in the lab by internationally harmonized protocols. Our pedotransfer functions are accurately and transparently presented, and derived from well-documented and high-quality soil data sets. We therefore consider them particularly useful in Austria, where the measured values for soil bulk densities are not available.

Keywords

• land-use
• regional calibration
• multiple regression
• monte carlo
• cross validation
• landnutzung
• regionale kalibrierung
• multiple regression
• monte carlo
• kreuzvalidierung