| Desertification is one of the world’s most serious environmental problems. In addition, global warming and climate change that caused by human activities have become the hot issue concerned governments and scientists. Revegetation is an effective way to combat land desertification. Artificial sand-fixing shrub forest has a variety of ecological service function. Carbon sequestration is a very important ecosystem service function. Ecological function and carbon sink function produced a significant difference under different planting densities and different planting years of sandy artificial shrub forest in large scale. At present, the results of the measurement and estimation methods of forest carbon density had a greater difference. It is not able to put multiple processes of ecosystem together to measure forest carbon density because of the complexity of ecosystem. Artificial neural networks has characteristics of distributed parallel processing, non-linear mapping, fault-tolerant and adaptive learning, which reflects a strong advantage in calculating ecosystem carbon density when the mechanism of high-dimensional nonlinear system is unclear.This study took two typical artificial shrub forest ecosystems of Caragana microphylla and Salix gordejevii at wind sandy land as research object in Aohan county, we investigated different planting densities and different forest ages comprehensively to analyze the carbon sink function of two artificial shrub forests in this area. We collected the measured data of this research area and established two shrub forest ecosystem carbon density prediction models based on the best parts of artificial neural networks-BP Neural network.It can provide a new method for carbon density estimation of artificial shrub forest ecosystem in arid and semi-arid zones of China. At the same times, it could provide scientific basis for evaluating carbon sink of sandy land ecosystem comprehensively.We draw the following conclusions:1. The shrub layer carbon density increased firstly then decreased as the increasing of forest age under different planting densities of two shrub forests. The herbaceous layer, litter layer, soil layer and ecosystem carbon density presented the trend of increasing gradually, then becoming steady as increasing forest age. Annual average carbon sequestration rate of shrub layer presented a trend from fast increasing to slow decreasing as increasing forest age. Shrub forest need proper artificial tending when the age of Caragana microphylla>6 and the age of Salix gordejevii>10.2. In the initial stage of growth, carbon density of different organs of Caragana microphylla shrub layer were expressed as:branch>root>leaf; As forest age increasing, carbon density of different organs of Caragana microphylla shrub layer was expressed as: root> branch>leaf. Carbon density of different organs of Salix gordejevii shrub layer in each age stage was expressed as:branch>root>leaf. Carbon density of different organs of herbaceous layer in different forest age was all expressed as:aerial part underground part.3. In the vertical direction, root carbon density of two shrub forests of different forest ages all presented the trend of increasing firstly then decreasing with increasing soil depth under different planting density; At the same forest age, root depth of shrub that had large planting density was larger than that had small planting density. In the horizontal direction, the greater the distance from the center of the standard wood root, the less root distribution of shrub layer of different ages under different planting densities. The horizontal distance of root distribution presented the trend of increasing as forest age increasing. At the same forest age, horizontal distribution distance of root expressed as large spacing greater than small spacing.As the root was dominant in the shrub layer, we could utilize it’s distribution characteristic as one of bases to choose a reasonable planting density of artificial shrub forest.4. Annual average carbon sink rate of two shrub forest ecosystems were decreasing as increasing forest age. Carbon sink two shrub forest ecosystems were expressed as: Caragana microphylla> Salix gordejevii. Among different parts of ecosystem carbon stocks, carbon density of soil layer was the largest, showing more than 60 percent of ecosystem; followed by shrub layer, carbon density of herbaceous layer and litter layer was the least, holding about l%-2% of ecosystem respectively.5. When the planting density of two shrub forest was 1m×4m in research area,carbon density of ecosystem of shrub were all best. Carbon stocks of Caragana microphylla artificial forest at wind sandy land in Aohan county was 519876.15t, carbon stocks of Salix gordejevii artificial forest was 325708.43t.6. The value of R2 was the largest when hidden layer was in 18th neuron point of BP Neural network model of Caragana microphylla(R=0.9943), the value of R2 was the largest when hidden layer was in 22nd neuron point of BP Neural network model of Salix gordejevii (R2=0.9936), two models presented the fittest results and the best stability,the smallest error and could reflect the actual situation. Therefore, we could ensure the network topology structure of ecosystem carbon density prediction model of Caragana microphylla and Salix gordejevii was 15-18-1 and 15-22-1 respectively. The accuracy of the two BP neural network models of shrub forest carbon density was up to about 99%, the results was very good. Therefore, BP neural network could be regarded as a new method to estimate shrub ecosystem carbon density.Shrub forest was the main artificial forest species in this region, planting density of lm X 4m should be prioritized when planting Caragana microphylla and Salix gordejevii. |
PDF Full Text Request