Objective The spatial variability and scale effect of soil available nitrogen content in maize propagation base were investigated, and reasonable sampling interval was determined to provide a theoretical basis for accurate soil sampling design.

Method Taking the perennial southern maize propagation plot as the test plot, soil samples were collected at 10 m sampling interval with original scale, and sampled successively at 20, 30 and 40 m to form different sampling interval data sets. Spatial autocorrelation Moran index, semi-variogram and fractal dimension were used to analyze the spatial structure changes of soil available nitrogen content under different sampling interval conditions.

Result The average content of soil available nitrogen in the test site was 90.24 mg/kg, belonging to the medium level, and the coefficient of variation was 19.77%, belonging to the medium range of variation. At the sampling intervals of 10, 20, 30 and 40 m, the data structure of soil available nitrogen samples did not change notably, only the coefficient of variation was different greatly. The coefficient of variation at the sampling interval of 20 m was the largest 29.37 %, while that at the other sampling intervals was little different. With the increase of sampling interval, the spatial autocorrelation Moran index of soil available nitrogen content decreased and its spatial aggregation became weaker, while the nugget value, block base ratio and fractal dimension showed an increasing trend, indicating that the spatial variation caused by random factors gradually increased, while the spatial variation caused by random factors gradually weakened the spatial variation caused by structural factors, making the spatial correlation gradually weakened. The spatial distribution of soil available nitrogen content with a 20 m sampling interval was consistent with the original scale of a 10 m sampling interval, and the root mean square error value of spatial prediction was the smallest, indicating that the spatial prediction accuracy was highest.

Conclusion With the increase of sampling interval, the spatial autocorrelation of soil available nitrogen content gradually weakened, indicating that the spatial variability caused by random factors gradually increased, and the spatial variation caused by structural factors gradually weakened. The spatial variation characteristics and distribution of soil available nitrogen content at 20 m sampling interval were similar to those of 10 m sampling interval with the original scale and the spatial prediction accuracy was highest, which was a reasonable sampling interval in this study. It provided technical support for precise application and management of nitrogen fertilizer in maize propagation base.