The Study Of Desert Rodents Communities Within Climate Change Lag Respones Based On BP Artificial Neural Network

Global climate changes (global warming and local cooling) strongly influence the structure of ecosystem, and the growth, reproduction and distribution of species. Climate changes led to more drought in the desert, changes in the law of rainfall and a series of meteorological factors also, a strong impact on the survival of plants and animals in desert habitats, distribution and pattern. The interference of human activities seriously affects the animals and plants to the climate change adaptation strategies, which result in the same species in different habitats changing their adaptation, and then the survival is more complicated. Neural network model is a mathematical model based on the brain of learning, setting up the model and making for the judgment. It is a nonlinear analysis method which can simulate the dynamic laws after the formation of the new system by mutually coupling the multiple systems and can predict the possible state in a period of time.In this paper, using BP neural network to simulate rodent community’s response of Alashan Desert on climate change in different interferences, and using rodent capture data in April to October from 2006 to 2014 and the meteorological data from 2003 to 2014, I study the dynamic laws of the system coupled by rodent community and climate, reveal rodent’s lag action and lag time on climate change, establish the BP neural network model and predict dominant species and diversity of rodent community. Results show:l.The models of the dominant species and the species quantity of Desert Rodent Community in Alashan established by BP neural network have differences in different time scales and the prediction effect was also different.The network models of 6 years which are better than 9 years with different interference includes 8 models.The models includes:population of rodents community with rotational grazing disturbance, Dipus sagitta forbidden grazing disturbance, population of rodents community with forbidden grazing disturbance, Dipus sagitta with the over grazing disturbance, Meriones meridianus with rotational grazing disturbance, Dipus sagitta with rotational grazing disturbance, population of rodents community in farmland, Allactaga sibirica with forbidden grazing disturbance. The goodness of fit are 0.9353>0.8895>0.8564>0.8290>0.8247>0.8121>0.7960>0.7090, the number of hidden layer are 16、18、19、18、1、20、11、7。The network models of 9 years which are better than 6 years with different interference are includes 6 models. As follows:Allactaga sibirica with rotational grazing disturbance, population of rodents community with over grazing disturbance, Meriones meridianus in farmland, Meriones meridianus with forbidden grazing disturbance, Allactaga sibirica with over grazing disturbance, Meriones meridianus with over grazing disturbance. The goodness of fit are 0.9034>0.9008> 0.8517>0.8474>0.8296>0.8046, the number of hidden layer are 24、23、11、4、8、28。2. The diversity index network models of 6 years which are better than 9 years with different interference includes the following network models. Simpson index model structure is 16-7-1 with grazing disturbance. Jaccard index model structure is 16-8-1 between rotational grazing and over grazing.Sorenson index network structure is 16-3-1 between over grazing and forbidden grazing.Cody index network structure is 16-16-1 between farmland and over grazing.Jaccard index network structure is 16-8-1 between farmland and over grazing.Sorenson index network structure is 16-3-1 bettween over grazing disturbance and forbidden grazing.The diversity index network models of 9 years which are better than 6 years with different interference includes the following network models.Shannon Wiener index network structure is 16-1-1 with grazing disturbance. Pielou index network structure is 16-7-1 with grazing disturbance;Cody index network structure is 16-17-1 between farmland and over grazing.Bray-Curitis index network structure is 16-23-1 between rotational grazing and over grazing.3. In different time scales, dominant species and the number of rodent community of different interference have lag response in the environment of climate change, and lag time is different.Under 6 year time scale:Dipus sagitta:lag time is 12 months with grazing disturbance and 1 month with grazing over disturbance; Allactaga sibirica:lag time is 24 months with grazing disturbance,24 months with grazing over disturbance and 2 months grazing prohibition disturbance; Meriones meridianus:lag time is 2months with grazing disturbance or breaking ground,12 months grazing over disturbance; population of rodents community:lag time is 2 months with breaking ground or grazing disturbance,3months with grazing over disturbance and 12 months with grazing prohibition disturbance.Under 9 year time scale:Dipus sagitta:lag time is 2months with grazing disturbance,1 month grazing over disturbance and 2 months with grazing prohibition disturbance; Allactaga sibirica:lag time is 2months with grazing disturbance; Meriones meridianus:lag time is 12 months with breaking ground, grazing over disturbance or grazing prohibition disturbance and 2 months with grazing disturbance; population of rodents community:lag time is 36 months with breaking ground, 2 months with grazing over disturbance and 36 months with grazing prohibition disturbance.4. Different interference effect the lag time of rodent community on the climate change. α βdiversity index also have lag response on it and the lag time is different with different gradients of human disturbances and habitats. At the same time, in different time scale, the lag time is different,5. In different time scales, the diversity of rodents community have lag response in climate change, and lag time is different.Under 6 year time scale:the lag time of shannon-wiener index is 3 months with rotational grazing and forbidden grazing,1 months with over grazing; lag time of pielou index is 12 months with rotational grazing disturbance,2 months with over grazing disturbance and 1 months with forbidden grazingdisturbance; the lag time of simpson index is 24 months with rotation grazing disturbance,3 months with over grazing disturbance,1 months with forbidden grazing disturbance.Under 9 year time scale:the lag time of shannon-wiener index is 1 months with rotational grazing,2 months with over grazing,12 months with forbidden grazing; lag time of pielou index is 2 months with over grazing disturbance,36 months with forbidden grazing disturbance; the lag time of simpson index is 12 months with rotation grazing disturbance,1 months with over grazing disturbance,3 months with forbidden grazing disturbance.6. Quantitative attribute β diversity index is based on the number of individuals in the habitat gradient variation trend of information, but two element attribute data exaggerate the effect of rare species, so Quantitative attribute index overcome the shortcomings of two element attribute data effectively. At the same time, quantitative attribute index P-cody, P-Jaccard and P-Sorenson of the two element attribute index cody, Jaccard and Sorenson were attempted in the lag response network model with different environmental gradients of species diversity and the climate change. In the 6 year scale and with breaking ground and over grazing gradient, the goodness of fit achieve 0.9076, 0.9258 and in 9year scale, the goodness of fit achieve 0.8973, which have ideal results.