Topsoil porosity prediction across habitats at large scales using environmental variables

Soil porosity and its reciprocal bulk density are important environmental state variables that enable modelers to represent hydraulic function and carbon storage. Biotic effects and their ‘dynamic’ influence on such state variables remain largely unknown for larger scales and may result in important, yet poorly quantified environmental feedbacks. Existing representation of hydraulic function is often invariant to environmental change and may be poor in some systems, particularly non-arable soils. Here we assess predictors of total porosity across two comprehensive national topsoil (0-15 cm) data sets, covering the full range of soil organic matter (SOM) and habitats (n = 1385 & n = 2570), using generalized additive mixed models and machine learning. Novel aspects of this work include the testing of metrics on aggregate size and livestock density alongside a range of different particle size distribution metrics. We demonstrate that porosity trends in Great Britain are dominated by biotic metrics, soil carbon and land use. Incorporating these variables into porosity prediction improves performance, paving the way for new dynamic calculation of porosity using surrogate measures with remote sensing, which may help improve prediction in data sparse regions of the world. Moreover, dynamic calculation of porosity could support representation of feedbacks in environmental and Earth System Models. Representing the hydrological feedbacks from changes in structural porosity also requires data and models at appropriate spatial scales to capture conditions leading to near-saturated soil conditions.