Spatial interpolators for improving the mapping of carbon stock of the arboreal vegetation in Brazilian biomes of Atlantic forest and Savanna

Abstract

The aim of this study was to map aboveground carbon stock of arboreal vegetation in the Savanna and Atlantic forest biomes in Minas Gerais State, Brazil, in order to assess the best spatial technique for mapping. The dataset was obtained from 148 forest fragments of these biomes. The best form of mapping was based on statistical criteria and mapping quality. The exponential semivariogram model was selected for conducting the study. The geographical model developed in this study for regression-kriging application was fitted having as input longitude and biome variables, and, globally, has presented good spatial behavior of the carbon stock distributed along the Minas Gerais State. From mapping carbon stock by different techniques, it was found that regression-kriging mapping was the most efficient. In addition, as the semivariograms were fitted for each biome using kriging and co-kriging, it is possible to stand out the flexibility for using regression-kriging, including biome as a categorical variable in the geographical model. Another result was the high correlation found between different forms of mapping, which adds reliability for this study. Thus, it was concluded that the carbon stock distribution in the arboreal vegetation of these two biomes is spatially structured. Ordinary kriging and co-kriging have presented satisfactory results, however, regression-kriging has been more reliable for mapping and estimating carbon stock distribution, in the Minas Gerais State.