Digital soil mapping using uncertain soil observations to support sustainable agricultural intensification in West and Central Africa

PhD research supervised by prof. dr Gerard Heuvelink (ISRIC - World Soil Information, Wageningen University), dr Jetse Stoorvogel (Wageningen University), dr Ermam Aynekulu (ICRAF) and dr Keith Shepherd (ICRAF).

The increase in the demands for spatial soil information has led to the increased use of GIS/geostatistical techniques and proximal soil sensing (PSS) methods for the development of Digital Soil Mapping (DSM) models. Many DSM models currently use soil spectral data, which are relatively inexpensive compared to soil data generated using conventional methods, but not as accurate. PSS measurement errors are usually not accounted for during the mapping process and may affect the quality of the model outputs and impair end-users’ decisions. This research aims at incorporating errors in soil measurements in state-of-the-art DSM approaches and assess their effect on DSM output accuracy. A spatially stratified and random sampling procedure has been used to collect soil samples across targeted landscapes in Cameroon and Chad and analysed using conventional laboratory methods and PSS. Uncertainties in laboratory and spectral data will be quantified and explicitly included in the covariance structure of the spatial model. The DSM modelling and prediction approaches to be used will include multiple linear regression and non-linear machine learning regression methods, such as Random Forest, combined with kriging. In cases where this cannot be done in a mathematically tractable way, a Monte Carlo simulation approach will be used to trace the effect of measurement errors. Soil property maps generated will be reported at various spatial supports to ease interpretation with various stakeholders at local and community levels. The soil information generated together with their quantified accuracies will be used to formulate agronomic recommendations using Decision Support System for Agro Technology Transfer (i.e., DSSAT), while accounting for uncertainties in soil measurements and predictions. The crop production yield analyses under two scenarios (with and without uncertainties will help in assessing the required accuracy level of soil information to support policies that enhance sustainable agricultural intensification practices.