Predicting the steady-state infiltration rates of agricultural soils of the Mediterranean region from their soil surface conditions

• A classification of soil surface conditions is tested for predicting infiltration rates. • The classification is based on five easy-to-observe soil surface descriptors. • 273 farmed plots with infiltration data across southern France and Tunisia served for evaluation. • The RMSE of the soil infiltration rate prediction via classification was 5.2 mm/h. • Soil cover variables were more important to soil infiltration than were permanent soil properties. Knowledge of soil infiltration properties is essential for managing water resources. Here, we present and evaluate an expert-based classification of soil surface conditions (SSCs) for predicting steady infiltration rates in Mediterranean soils. The classification distinguishes 9 SSC classes. To identify the SSC classes, an identification chart is proposed that uses five easy-to-observe soil surface descriptors (SSDs), namely, the soil cover percentages of clods, surface crusts, rock fragments, herbaceous vegetation and litter. To evaluate the classification, we gathered a dataset including infiltration measurements by rainfall simulations and observations of 12 different SSDs at 273 plots across southern France and Tunisia. The plots represent a range of Mediterranean crops and land uses, including vineyards, wheat, barley, chickpeas, rangeland and fallow land. The expert classification was compared with a statistically derived classification based on a regression tree approach and calibrated with the observational dataset. Variance analysis revealed that the expert SSC classification and the fitted statistical classification explained 73 % and 79 % of the variance in infiltration rates, respectively. A cross-validation test showed that the expert classification is more precise and robust than the statistical classification for predicting soil infiltration rates. The expert classification exhibited a root mean square error (RMSE) of 5.2 mm/h, much smaller than the RMSE values of current pedotransfer functions for predicting soil hydraulic conductivity. Soil surface cover characteristics, proved to be more important proxies for discriminating soil infiltration capacity than permanent topsoil properties, such as soil texture and soil types. To conclude, the expert classification provides an easy and accurate way to estimate the soil infiltration capacity of agricultural fields.