It is difficult to evaluate and predict soil erosion and its related effects in mountainous areas without sufficient observation data. This study aimed to estimate and predict changes in soil erosion using137Cs

RS

and GIS techniques and propose an a pproach for soil erosion evaluation in mountainous areas. Using the erosion modulus calculating models based on137Cs concentration

the annual average soil erosion modulus of forest lands

shrub lands

grasslands

farmlands

and uncovered land were obtained. The values were 356—1531 t /（ km2·a）

330—1709 t/（ km2·a）

886—3885 t/（ km2·a）

5197—12454 t/（ km2·a）

and more than 1 5000 t /（ km2·a）

respectively. Then

erosion zoning was done by combining erosion rates and land use data by interpreting remote sense images from 1987（ Landsat TM）

1995（ Landsat TM）

and 2005（ Landsat ETM）

and these were overlaid with 1∶ 50000 DEM data. The results showed that the eroded land changed very little from 1987 to 2005

accounting for about66% —67. 3% of the total; however

the erosion intensity significantly rose from 1987 to 1995

up to 30% for some land use types. The eroded land

with the erosion modulus of 2500—5000 t /（ km2·a）

5000—8000 t/（ km2·a）

and 8000—15000 t /（ km2·a） rise 30%

23%

and 26%

respectively. The soil erosion amount in the Xiaojiang River basin was 7.51 ×106t / a

8. 19 × 106t /a

and 8. 18 × 106t /a in 1987

1995

and 2005

r espectively. Moreover

the zoning and amount of soil erosion for 2015 was predicted using a Markov-Cellar A utomata Model

which was established using the data from 1995 and 2005. The predicted result

8. 17 × 106t /a

was very similar to that from 2005. This study provides a valuable solution to evaluate and predict soil erosion for mountainous areas in southwest China.