Scripts used to general the GlobalSoilMap products for available water capacity for metropolitan France. The objectives and results of the scripts are explained in the publication by Román Dobarco et al. (2019):
Román Dobarco, M., Bourennane, H., Arrouays, D., Saby, N.P.A., Cousin, I., Martin, M.P., 2019. Uncertainty assessment of GlobalSoilMap soil available water capacity products: A French case study. Geoderma 344, 14–30. https://doi.org/10.1016/j.geoderma.2019.02.036
Abstract Plant available water capacity (AWC) refers to the maximum amount of water that a soil can store and provide to plant roots. Spatial predictions of AWC through digital soil mapping at high resolution and national extent provide relevant information for upscaling ecological and hydrological models, and assessment of the provision of ecosystem services like water quantity and quality regulation, carbon sequestration, and provision of food and raw materials. However, the spatial predictions of AWC are prone to errors and uncertainties. Moreover, this digital soil mapping process requires using pedotransfer functions (PTFs) due to the lack of sufficient georeferenced measurements of the upper (i.e., soil moisture at field capacity, θFC) and lower (i.e., soil moisture at permanent wilting point, θPWP) limits of soil moisture contents defining AWC. This adds an additional source of uncertainty to the final estimates of AWC. The objectives of this study were: 1) to predict AWC for mainland France following the GlobalSoilMap (GSM) project specifications on depth intervals and uncertainty, and 2) to quantify the uncertainty of AWC accounting for uncertainty of the soil input variables and the PTFs' coefficients. We first predicted the soil input properties by GSM layer (0–5, 5–15, 15–30, 30–60, 60–100, 100–200 cm), and then applied PTFs for estimating θFC, θPWP, and volumetric AWC (cm3 cm−3). The volume of coarse elements by GSM layer was subtracted before aggregating AWC to estimated soil depth for a maximum of 2 m. The uncertainty of AWC was quantified by first-order Taylor analysis. Independent evaluation indicated that clay had the lowest R2 (clay R2 = 0.27, silt R2 = 0.43 and sand R2 = 0.46) and RMSE (clay RMSE = 128 g kg−1, silt RMSE = 139 g kg−1 and sand RMSE = 172 g kg−1) from the three particle size fractions. However, the model for coarse elements had the worst predictive performance (R2 = 0.14 and RMSE = 21%) among all AWC input variables. The performance of the GSM predictions for θFC and θPWP had a R2 of 0.21 and 0.29. When the PTFs were applied to the spatial predictions of sand and clay, the RMSE for θFC and θPWP had a relative increase of 25% and 36% respectively compared to when they were applied to measured horizon data. Across the majority of mainland France, the main sources of uncertainty of elementary AWC were coarse elements and soil texture, but the contribution of uncertainty of PTFs' coefficients increased in areas dominated by very sandy and clayey textures. An advantage of the produced maps of θFC, θPWP and AWC is that the end users can incorporate associated uncertainties into ecological and agricultural modelling, and decision-making processes involved in soil and water planning.