DEFORESTED AREA OF PICEA CRASSIFOLIA IN ARID REGION RECENTLY PROTECTED: QILIAN MTS. NATIONAL NATURAL RESERVE (CHINA)

The earth is now facing the land degradation due to human disturbance, natural habitats were converted to rural and agri- cultural areas in order to fulfill the increasing de- mand of human population. The deforestation of Picea crassifolia (Qinghai spruce) forest at Qilian Mts is an example of such disturbance. P. crassi- folia is an ecologically and hydrologically impor- tant plant species in the northwestern arid area of China. However, the forests have been intensively and extensively deforested. In order to restore the human-disturbed ecosystems, the spatial distribu- tion of P. crassifolia needs to be delineated. This study employed Genetic Algorithm for Rule-set Prediction model (GARP) and Maximum entropy model (Maxent) and four environmental vari- ables (mean temperature of the warmest quarter, precipitation of the wettest quarter, annual solar radiation, topographic wetness index) to predict the potential distribution of P. crassifolia in Qil- ian Mts. Genetic Algorithm for Rule-set Predic- tion model (GARP) produces a model of species niches in geographic space based on heteroge- neous rule-sets. Maximum entropy model (Max- ent) focuses on fitting a probability distribution for occurrence based on the idea that the best ex- planation to unknown phenomena will maximize the entropy of the probability distribution, subject to the appropriate constraints. The environmental variables were spatially interpolated throughout the entire study area. We used sensitivity-specific- ity sum maximum approach to select the thresh- old value. The projected niche space for the mean temperature of the warmest quarter is between 8.5 and 18.1°C; the space for the precipitation of the wettest quarter is between 149 and 245 mm; the space for annual solar radiation is 118-1100×10 3 wh m -2 and the space for topographic wetness index is between -0.4 and 5.1. The results show that both GARP and Maxent's models produce ac- ceptable predictions, but the overall comparison shows that GARP prediction is better than Max- ent's; the comparison between the observed dis- tribution and the predicted distribution suggests that 61% (2869 km 2 ) of P. crassifolia forests have been deforested.