Optimization of sampling design for total organic carbon assessment using spatial simulated annealing: comparison of different variogram models performances.

Assessment of soil organic carbon is of primary interest for evaluating soil quality and its variation as effect of agronomic management. Appropriate sampling strategy and data analysis play a crucial role to take into account variability that occurs at a scale smaller than the block size, assess spatial dependence between observations and residuals, avoiding in this way erroneous conclusions about treatment significance (Littell et al., 2006). However, an over-sampling of the investigated field could be time-consuming, labor-intensive and costly without a consequent significant knowledge gain. To avoid this pitfall, optimization of sampling schemes allows reducing the number of sampling points with a negligible impact on the accuracy of the estimate of the investigated attribute (Barca et al., 2015). In defining optimal sampling schemes, important issues and decisions concern, among others, the choice of the optimization approach to be used (model-based or design-based), the optimal variogram model when a model-based approach is considered, the use of covariate information. In this preliminary study, spatial simulated annealing was used as a method to optimize a TOC sampling scheme. Two theoretical variogram models were used in order to reduce a previously defined experimental design and to assess the impact of model selection on the optimal configuration.