A Study Of The Forewarning Model Of Grassland Desertification Based On Matlab

Grassland is not only an important renewable resource in the terrestrial ecosystem, but also an imp ortant natural shelter zone. It plays a significant role in protecting the regional and national ecological e nvironment. However, with the high-speed development of the human society, the ecosystem of grassla nd is experiencing degradation and desertification to different degrees. The degradation of grassland has impacts on the ecological civilization strategy of building a beautiful China. Therefore, it is urgent to ta ke effective measures timely to stop the grassland degradation, restore grassland vegetation, and give ful1play to the ecological functions of grassland.The study chooses typical sample areas based on five degrees of desertification, namely potential, mild, moderate, severe and very severe, to carry on field surveys and monitor dynamic changes of envir onmental factors, such as physical and chemical properties of soil, and then establishes forewarning mo del of grassland desertification by Bayes classifier and support vector machine(SVM).The conclusions of the thesis are as follows:(1) With grassland desertification getting worse, soil nutrient content, as a whole, is on the declin e. The content of total nitrogen, total phosphorus, rapidly available phosphorus and rapidly available p otassium in soil gradually reduces as the desertification degree increases.The total nitrogen content di ffer greatly among grasslands of different degrees of desertification. The a diversity index of plant co mmunity shows a trend of decline with the worsening of desertification degrees. As for potential deser tification grassland, the index is the highest; as for extreme desertification grassland, the index is the1owest.(2)The correlation analysis between the grassland soil and vegetation index shows that there is si gnificant positive correlation between rapidly available potassium and rapidly available phosphorus. T he enzymes activity of sugar is significantly or extremely significantly correlated to the soil nutrients, except that the enzymes activity of sugar and the content of rapidly available phosphorus are not signi ficantly correlated. There is obvious correlation between the a diversity index of plant community and the soil indexes.(3) When seven indexes of meadow soil and vegetation, namely organic matter, total nitrogen, alk ali solution nitrogen, available phosphorus, sucrase, dominance index and evenness, are used in buildi ng the model, the prediction accuracy of grassland desertification forewarning models based on Bayes classifier, is62.5%,66.67%,86.67%,73.33%and87.5%respectively for the five different degrees, namely potential, mild, moderate, severe and very severe, and the overall predictive accuracy is77.27%.(4) When seven indexes of meadow soil and vegetation, namely organic matter, total nitrogen, alk ali solution nitrogen, available phosphorus, sucrase, dominance index and evenness, are used in buildi ng the model, the grassland desertification warning model based on SVM has the highest accuracy. T he overall prediction accuracy is83.33%and the prediction accuracy of the grassland desertification f orewarning models based on SVM is75%,83.33%,80%,80%and93.75%respectively according to t he five different degrees, namely potential, mild, moderate, severe and very severe. (5)By contrast, the predication accuracy of SVM model is higher than the Bayes classifier model. The predication results highly accord with actual situations, and the predication accuracy of more seri ous desertification grassland is higher than that of the potential and mild degree desertification grassla nd. SVM model has promotional value in the grassland desertification warning and its application fiel ds have been widened.