The Influence of Spatial Resolution on the Prediction of Soil Organic Matter Distribution in a Mollisol Watershed of Northeast China

Geostatistics, traditional statistics and landscape indicators were used to analyse the influence of sampling resolution on the prediction of spatial variability of soil organic matter (SOM) in a typical Mollisol of northeast China. Gaussian models were recognized as the best to predict SOM spatial distribution in all resolution groups. Spatial autocorrelations as influenced by structure factors were moderate for groups 0.025, 0.037b, 0.074a, 0.074c and 0.074d, and strong for 0.015, 0.037a, 0.074b and 0.074e. The relatively shorter autocorrelation distances (A₀) in data groups were all close to 7 km. Means and standard deviation (SD) of 0.025 resolution was close to 0.015. TA (Total area), LPI (Largest patch index) and COHESION (Patch cohesion index) were similar between resolutions 0.015 and 0.025. Generally, a sample-grid ≤ 0.025 km² was recognized as a better resolution to predict SOM spatial variability by ordinary kriging interpolation if a sample-grid method was adopted in the black soil region of northeast China. The accurate prediction of soil nutrient heterogeneity by interpolations (Geostatistics) is mainly determined by the representative of soil sampling which should reflect through resolution the entire environmental factors in the research area.