Evolutionary Optimization of Spatial Sampling Networks Designed for the Monitoring of Soil Organic Carbon

In this research, optimal spatial networks designed for second-phase sampling of soil organic carbon were found based on the spatial information obtained during a first-phase sampling. The study was carried out in soils of tropical crops, forests and pastures in an area of about 1,310 ha located in the foothills-situated to the east of the Colombian Andes mountains. Mean soil organic carbon content in the upper 1 m in the study area was 18.9 t ha(-1). Additional points supplementing the existing initial sampling set were allocated, in random and regular configurations, following two different approaches: sequential and simultaneous addition. The search for the optimal set of additional points was performed using an evolutionary optimization technique known as genetic algorithms. Results showed that random schemes allocated following the simultaneous addition approach were more efficient than regular schemes. Besides, the sequential addition produced suboptimal solutions, becoming less efficient than the simultaneous addition when the number of additional points to be allocated was increased. The optimization technique used in the study, the genetic algorithms, proved to be effective to find optimal spatial networks designed for second-phase sampling of the variable of interest.