The Karst Rocky Desertification (KRD) occurred in Southwest China has become a research hotspot and drawn great attention of the Chinese government in recent years. However, the current researches about KRD has not revealed the essential reason why the KRD phenomenon occurred in karst mountains, and has not clarified why the local farmers had undertaken unreasonable agricultural activity in karst mountain areas. This paper attempted to discover theoretically the reason why the local farmers in karst mountain areas carried out irrational activities such as the steep slope land cultivations, and to reveal essentially the occurrence mechanism of KRD through the typical case studies. Firstly, this paper assumes that the low land capacity is the genesis cause of KRD in peak-cluster depression areas. Furthermore, ecological quality of the peak cluster depression zone is determined by the relationship between the area of depressions (negative terrain) and the population of residential area. So, six typical of peak-cluster depression areas of Guizhou province were selected to compare the distribution circumstances of cropland, the characteristics of settlements and the formation of KRD with the help of Alos images in 2010 (with a resolution of 10 m×10 m). The results show that the areas such as closed peak-cluster depression of continuous distribution, shallow peak-cluster depressions and assemblages of peak-cluster depression and canyon are lack of cropland resources, but some terrain such as peak-cluster depression group embraced by flat land and not closed peak-cluster depression are rich in cropland, and the cropland distributed in spatial array of peak-cluster depression, peak-cluster valley and gully area is relatively fewer. There is a negative correlation between the percentage of the cultivated land and the percentage of KRD (including light KRD, moderate KRD and severe KRD) at peak- cluster depressions, serious KRD distributes in the peak-cluster depressions with higher percent of the settlement accounted for the cropland area. It is related to some extent, in the six study areas, between the quantity and spatial distribution of arable land and settlements, and the size and spatial distribution of KRD. The relationship could be concluded as three situations of the process of KRD, where are low, middle and upper carrying capacity of land. The serious KRD is found distribute only in peak-cluster depression areas with fewer flatland, low land capacity and high population pressure, therefore, the theory hypothesis suggested by this paper is correct, to improve the bearing capacity of land should be the primary means to desertification control in peal-cluster depressions. The theoretical models of KRD occurrence mechanism in these areas were summarized further through the comparisons of the arable land resources distribution, the characteristics of settlement and the KRD occurrence situations in these six different peak cluster depression combined topography. The ecological qualities are dominated by the harmonization between land capacity and population pressure in the peak cluster depression areas. And this finding will make contribution to understand essentially the occurrence mechanism of KRD and be guidance to the ecological restoration of KRD land.