Ecological Environment Vulnerability Assessment Based On ANN Model In Sichuan Province

With the rapid social development and optimization of the industrial structure adjustment; global climate change; the development and utilization of natural resources, ecological environment vulnerability degree and carrying capacity are being changed accordingly. On the one hand, to meet the requirement of people's production and living, many parts of the natural resources have been exploited, the land use and structure has been changed. Environment deterioration was caused by human activities. On the other hand, some parts of the ecological environment has been improved through large ecological environment construction project, returning cultivated land to forest or pastures and carrying out ecological restoration in key ecological areas. Furthermore, due to the various types of natural environment and the obvious regional differences in our country, different regions under the pressure of the environment and development have different degrees of vulnerabilities. The ecological environment vulnerability is one of the important indicators to measure regional sustainable development ability. The study of vulnerability evaluation has attracted more and more attention and the ecological environment vulnerability problem has gradually become one of the focus issues.Located in the southwest of China, Sichuan province is the birthplace of many rivers. It is the ecological barrier of the upper reaches of the Yangtze River, and it is also one of an important region to maintain the overall stability of the ecological environment of our country. With complex terrain, abundant natural resources, the ecological environment in this area is diversity. Not only the entire Yangtze River valley but also the country's social and economic sustainable development will be impacted by the ecological environment of Sichuan.Based on the industry standards and existing research results, using 3S technology, as well as remote sensing data, elevation data, statistical yearbook and survey data, the eco-environmental vulnerability evaluation database of the study was built. On this basis, BP neural network and extreme learning machine were established and significant tested. At last, BP-ANN model which own better significant was chosen to objectively evaluate the eco-environmental vulnerability of the study area. The research results have important theoretical and practical significance on the protection of ecological environment in this area, and provided an important scientific basis for the departments of the country or provinces' policy development about a region environmental protection and ecological restoration.This article drew the main conclusions and innovations as follows:(1)In view of the complex terrain and diversity ecological environment, with reference to the industry standards and existing research results. This study built a scientific, reasonable and practical eco-environmental vulnerability evaluation index system based on the theory of “origin-result” system. With collecting and processing all kinds of index, nine factors which have larger influence on ecological vulnerability are received.(2)Using landsat images, the interpreter marks of remote sensing in west Sichuan plateau, plains, east Sichuan hilly region, were separately established. Saving information extraction time, the information of land use in the study area is extracted by using the object-oriented support vector machine(SVM). A set of suitable for large scale ecological environment vulnerability assessment method has been innovatively put forward according to quantify the influence of factors which are hard to quantify after the first classification processing(such as geological disaster factors).(3)BP-ANN model and ELM neural network for ecological vulnerability assessment were constructed. This study has separately discussed the best number of hidden layer neurons about the tow algorithms and compared the effectiveness, advantages and disadvantages of the two kinds of evaluation model. It shows that the of the tow are at about 0.7. BP-ANN model's mean square error is more stable and the results are more regular while the result of ELM has a certain degree of stability but has no regularity. ELM model saves manpower and time for its fast operation and easy to use.(4)Researches on the assessment of ecological environment vulnerability, particularly in the provinces unit using artificial neural network model is less. This study evaluated the ecological environment in the study area based on the BP-ANN model trying to use county as evaluation unit which bring beneficial exploration for the vulnerability evaluation research in provinces units. After analyzing the evaluation results and the space structure of the distribution of the results, the studies have shown that the ecological environment in eastern area is the best, and in Sichuan basin is better than mountainous area. With the best natural condition, Chengdu plain has a relatively stable ecological environment. The ecological condition around the Chengdu plain area is the second. Due to the weak economic foundation and acute ecological environment destruction, the ecological environment vulnerability is serious in Liangshan prefecture, Aba prefecture and Ganzi prefecture which need to be taken attention. According to the result of evaluation, Proposals can be put forward for the ecological environment sustainable in the study area.