Outliers detection of cultivated land quality grade results based on spatial autocorrelation

Scientific and accurate cultivated quality grade results are an important guarantee for applications. However, the phenomenon of abnormal cultivated land quality is always occurred due to the error of investigation or computation, so exploring effective method to inspect abnormal data of the grade results is an important problem to be solved. Cultivated land quality grade results of Wuhan Hannan district are selected as the research data, and several typical relative indexes of cultivated land quality are researched as spatial variables, including natural index, use index and economic index, etc. Global Moran's I and Local Moran's I index are used to analyze spatial structural characteristic and clustering rules of cultivated land quality, and then put forward cultivated land quality outlier detect standard based on the research of the front. The results showed that: On the whole, natural index, use index and economic index all showed strong positive spatial correlation and the aggregation degree is high; On the local, the cultivated land quality form "high - high" and "low - low" spatial cluster, accompanied with "high - low", "low - high" space isolation area, which formed a horizontal Mosaic distribution pattern of "patches" and "gaps". Further study to the "high-low", "low - high" space isolation zone with significance analysis can effectively detect the cultivated land quality abnormal data. This method improves the traditional statistics due to ignoring the spatial correlation in space anomaly inspection and provides a new train of thought for cultivated land quality data outliers detection.