Spatial stochastic model for predicting topsoil organic carbon content over a cultivated periurban region, using soil properties, a digital elevation model and remote sensing
Résumé
Though many cultivated soils in periurban areas are threatened by urbanization pressure, periurban agriculture is likely to develop through recycling organic urban waste compost on these soils. Monitoring the effects of applying organic amendments requires that topsoil organic carbon (SOC) content be spatially assessed.
This study aims at estimating SOC contents using point sampling data over the periurban surroundings of Versailles (France), covering a large periurban area (221 km²) characterized by cereal cropping systems and contrasting soil types.
It also aims at spatially quantifying prediction error uncertainties, in order to compare them with those obtained through remote sensing estimation methods (Vaudour et al., 2013).
Our data consist of common quantitative soil variables (physical and chemical properties, granulometry) and soil types at 256 point sites, and a digital elevation model with 25 m- spatial resolution. The various estimation methods must deal with some missing data.
Our estimation method is cokriging with external drifts when using only soil measures, and hierarchical bayesian spatial estimation when including multispectral SPOT4 satellite NDVI spectral bands. For each method, emphasis was put on assessing statistical estimation uncertainties. Results were cross-checked accordingly to the current year soil samples in the study area.
The estimation results and uncertainties are discussed regarding to the variables included into the models, and compared with other works about cultivated regions of equivalent area. Predictions with the best models and variable sets result in validation errors in a similar range to those obtained with satellite-based models.
Acknowledgments: this work was supported by both the Paris-Saclay campus (BASC-SOCSENSIT project) and the French National Space Agency in the framework of the TOSCA SURFAC EGC SENTINEL PLEIADES CO project.