Identification of Spatial Outliers by Moran's Index and Evaluation of Their Effects on the Spatial Distribution of Soil Organic Matter

Soil organic matter (SOM) is a key index in evaluation of the soil degradation and soil carbon sequestration. Therefore, determination of the SOM spatial patterns is essential for developing the suitable strategies of soil and ecosystem management. Geostatistical methods such as kriging have been widely employed to investigate the spatial pattern of different soil properties like SOM. The quality of the spatial maps is significantly influenced by the statistical properties of the raw data. Thus, identification and elimination of spatial outlier data, which has rarely been considered in previous works, is an important step in preparation of accurate and suitable maps of soil organic matter. The aim of the present study was to evaluate the effect of spatial outliers on the spatial pattern of soil organic carbon at the Rozeh-Chay watershed, Urmia, west Azarbayjan province. A total of 89 surface soil samples (0-10 cm) were collected based on the stratified random sampling scheme. After the normalization of the raw SOM data, global outliers were eliminated by box plot method. Spatial outliers were identified by the global and local Moran’s I indices. Spatial correlogram of the global Moran’s I showed the highest spatial autocorrelation at distance of 900 m, which was used as a distance band for preparation of spatial clusters map by local Moran’s I index. Cluster map of the local Moran’s I index showed four spatial outliers which were eliminated to assess their effects on the accuracy of kriging map of SOM. With the elimination of the spatial outliers, the MAE and RMSE of the SOM map were decreased from 0.97 and 1.31 to 0.85 and 1.12, respectively. Therefore, the accuracy of the kriging map increased by 13.5 percent. Generally, it can be concluded that the combination of Moran’s I index and kriging method improves the efficiency of organic matter map in the study area.