In order to reveal the spatial variations of soil nutrients and to guide the reasonable management of soil resources, the water-source forest in the mountainous areas and Beijing-Tianjin sandstorm source area in western Beijing were taken as the study object, on the basis of field sampling and by using the method of combining geostatistics and GIS, the spatial distributions of soil available nitrogen, available phosphorus, available potassium, organic matter and total nitrogen were studied. Fuzzy pattern recognition theory was used to evaluate soil nutrient levels analyzed the influential factors of the spatial variation characteristics combined with the actual regional situation. The results showed that the spatial distribution of soil nutrients were in the normal distribution (P<0.05), the spatial variabilities of available phosphorus, available potassium and organic matter were stronger, the variation coefficients of available nitrogen and total nitrogen were lower, and soil nutrient gradually reduced with the increase of soil depth. The indexes of soil nutrients well fitted to the semivariogram model, organic matter and total nitrogen had good spatial autocorrelation, available nitrogen, available potassium and available phosphorus showed moderate autocorrelation, the spatial autocorrelation level of soil nutrient distribution pattern was strong. Available nitrogen was the highest in medium proportion, its area accounted for 27.87%, available potassium and available phosphorus were in lack states, but organic matter was very rich, and 73.77% of the soil texture types were rich in organic matter, total nitrogen was in abundance state with a low variation, the regional soil fertility was in comprehensive IV level. The research results can provide scientific basis for the reasonable division, rehabilitation and sustainable utilization of the regional soil resources.