Analysis On Potential Distribution Area Of Xinjiang Typical Poisonous Plant Based On MaxEnt And GARP Ecological Niche Model

In recent years, the serious degeneration of grassland(grassland overgrazing,blind reclamation and deforestation) and grassland insufficient infrastructure and the lag of management ultimately caused a lot of poisonous grass breeding, even forage of recession.Livestock occasionally eat poisonous grass. In drought years, livestock beggars cannot be choosers and there will even be poisoning phenomenon. So, Poison grassis called "invisible killer".they seriously affects the development of grassland animal husbandry, ecological safety, biological safety and food safety. Xinjiang is a big province of animal husbandry in our country, but because grassland poisonous weeds is becoming increasingly serious, causing great losses to the local economy.Therefore, investigation of grassland poisonous grass is very important and significant.In this paper, through the collection and finishing national open herbarium website, domestic and foreign published papers and measured Aconitum leucostomum Vorosch distribution data, received a total of 100 which do not repeat. Based on MaxEnt and GARP model, this paper adopting the 19 environmental factors of the current climate(1950-2000) and the future global climate patterns( Beijing Climate Center_Climate System Model version 1.1)which they were two typical atmospheric emission concentration(RCP2.6 and RCP8.5) during 2050 and 2070, forecasted Aconitum leucostomum Vorosch potential distribution in xinjiang and analyzed the main environmental factors affecting the distribution of it. In ArcGIS 10.0, the geographical distribution of Aconitum leucostomum Vorosch can be divided into four grades using the method of Natural Breaks. Based on method of ArcGIS spatial analysis, this paper counted distribution area of different grade and analysised the distribution regularity of it. The paper analyzes the AUC value of MaxEnt and GARP of ROC curve based on Knife analysis,the two model of AUC value were0.999 and 0.818.According to the AUC value evaluation index, two model has goodsimulation effect.Paper set 3 kinds of different proportion training samples and discussed the MaxEnt and GARP models prediction effect of the variation coefficien of AUC, the results showed that the two model by the sample distribution was less affected, the model has good stability. This paper set up a nine point distribution, according to the AUC value to verify the stability of the two models, the results showed that: MaxEnt is not affected by the threshold effect of stability, good; GARP model governed by a threshold effect, when the sample points reach a certain amount, the model exhibits a certain stability.