Recalibration of existing pedotransfer functions to estimate soil bulk density at a regional scale

Abstract

AbstractSoil bulk density (ρb) is an important indicator of soil quality, productivity, compaction and porosity. Despite its importance, ρb is often omitted from global datasets due to the costs of making many direct ρb measurements and the difficulty of direct measurement on rocky, sandy, very dry, or very wet soils. Pedotransfer functions (PTFs) are deployed to address these limitations. Using readily available soil properties, PTFs employ estimator equations to fit existing datasets to estimate properties like ρb. However, PTF performance often declines when applied to soils outside those in the training dataset. Potentially, recalibrating existing PTFs using new observations would leverage the power of large datasets used in the original PTF derivation, while updating information based on new soil observations. Here, we evaluate such a recalibration approach for ρb estimation, benchmarking its performance against two alternatives: the original, uncalibrated PTFs, and novel, local PTFs derived solely from new soil observations. Using a ρb dataset of N = 360 total observations obtained in West Azerbaijan, Iran, we varied the local dataset size (with N = 15, 30, 60, and 360) and recalibrated four existing PTFs with these data. Local PTFs were generated based on stepwise multiple linear regression for the same datasets. The same PTFs (original, recalibrated, and local) were also applied to the study area, and the resulting ρb estimates were compared with the global SoilGrids dataset. Recalibration of PTFs reduced errors relative to the original uncalibrated PTFs; for instance, the NSE increased from −22.07 to 0.30 (uncalibrated) to 0.20–0.41 (recalibrated), and RMSE decreased from 0.12 to 0.60 Mg m−3 (uncalibrated) to 0.10–0.13 Mg m−3 (recalibrated). The recalibrated PTFs performance was comparable to or better than local PTFs applied to the same data. Recalibration of existing PTFs with local/regional uses provides a viable alternative to the use of global datasets or the development of local PTFs in data‐scarce regions.Highlights Existing global PTFs were calibrated and tested using a small dataset for local utilisation. Several new local PTFs were also developed using the same datasets. Recalibration of existing global PTFs is comparable to or more accurate than developing new PTFs.