A study was carried using semivariogram analysis to quantify spatial autocorrelation of carbon stock (CS) in soil, grass and weed species biomass in pastures of Brachiaria brizantha, Marandu grass with low, medium and high degradation, and grown an entisol. The sampling of plants and soil were carried out in regular grid with distances of 5x5 m in an area of 900 m². Grassland CS was assessed through descriptive statistics, comparison of averages through the test Kruskal-Wallis at 5% level of significance, geostatistics and ordinary kriging analysis. The spatial variability of CS was observed within and between pastures with low, medium and high degradation. Pastures with low levels of degradation had less spatial continuity due to smaller ranges in CS in grass biomass and total biomass (grass + weed species), in soil carbon and soil versus grass (total biomass + soil). The grid of 5x5 m was adequate to characterize the spatial variability of pasture with low and high levels of degradation. Areas of Marandu grass with average degree of degradation has high coefficients of variation (CV) between the CS values, which negative by affected the spatial modeling. High CV may also be due to the low number of samples taken (n = 36). Sampling points in grid lower than 5 m can improve the accuracy of the adjustment of semivariograms.
CO2; degraded pasture; geostatistics; inventory of carbon; carbon sequestration
