The Xiaojiang watershed of Yunnan Province, Southwest China, covers an area of about 1034 km2, which extends 24°12′-24°45′N and 103°32′-104°00′E. The watershed is mainly underlain by Triassic system with carbonate rocks of 617 km2 or about 60%. Understanding the spatial variation and distribution pattern of SOM content is needed for sustainable development in the karst region. The spatial variation of SOM content in Xiaojiang watershed was determined using geostatistics and GIS to provide information for understanding rock desertification mechanisms and preventing soil degradation.
Based on different stratum, landform, and land-use type, 177 soil samples ( 0-20 cm ) were collected from the watershed. The SOM content was analyzed using theWalkley Blackwet combustion method. The geostatistical characteristics, spatial trend and azimuth of anisotropic axle of SOM content were analyzed using the Geostatistical Analyst, ArcGIS 9.2. And a contour map of SOM content for the watershed using Kriging interpolation was obtained. The results are as follows:
(1) The distribution of SOM content was highly variable in Xiaojiang watershed. Statistical analysis showed differences in SOM content among different stratum, landform and land-use type. The coefficient of variation (CV ) of SOM content was greatest in carbonate (CV =63.1 %) and smallest in Quaternary soil (CV=40.1 %) among stratum; and greatest in karst groovy (CV=61.6 % ) and vale (CV=61.6 % ) and smallest in hill soil (CV=32.3 % ) among land form;.and greatest in unused land (CV=60.6 % ) and smallest in dry land soil (CV=52.9 % ) among land-use types. For all soil samples (n= 177) , the mean, minimum, maximum, standard deviation and coefficient of variation of SOM content were 29.00g/kg, 1.90g/k g, 93.64g/kg, 17.50 and 60.5%, respectively.
(2) The spatial variation of SOM content in the watershed was anisotropic, which showed that the SOM semivariogram depended on both the distance and the direction of soil sampling. The spatial correlation distances ( ranges) were large (39.8 km on the long axle and 22.0 km on the short axle). The nugget variance (0.28) indicated a large disconnection of the semivariogram value from the origin, and the sill (0.52) showed a large semivariogram value for distance beyond the ranges; and the integrative comparison of the prediction errors from the trend effects indicated that the 1 order trend effect was preferable.
(3) There is a different in the spatial distribution of SOM content in watershed. The Kriging spatial interpolation showed that SOM decreased gradually from the east to the west and southwest in the watershed. The lowest level band of SOM content was in the southwest of the watershed, but the high level band of SOM content was in the east of the watershed. The spatial distribution of SOM content was related to variation in stratum, land form and land-use type in the watershed.