The researchers studied how to choose the best variables for analyzing land use patterns. They focused on a rural area in Linxia County and looked at terrain factors to predict patterns. They found that using global Moran's index was better than Getis-Ord general G. Local Moran's index was good at finding high-value clusters, while Local Getis-Ord G was better at finding low-value clusters. They also saw that the dispersion of residential areas matched well with the terrain features. This information can help future studies on land use patterns.