Health Evaluation Of Forest Ecosystem In Yong’an City Based On BP Neural Network

Yongan forest ecosystem was taked as the research object,which adopts principal component analysis(PCA)and BP neural network and combines with a large number of previous studies to make a comprehensive on the Yongan forest ecosystem health evaluation.In order to obtain accurate and objective results,this study started from the three aspects------arbor forest,bamboo forest and shrub wood,and carried out the health evaluation and analysis through multi-index and multi-angle.Considering that the forest ecosystem is a dynamic system,this study would be maked a comparison on the Yongan’s forest health assessment result between 2012 and 2016 to determine the overall development trend of forest health in Yongan city and indicate the key methods of work for the future.The main conclusions were as follows:(1)The following principles were proposed in this study,which should be followed in the construction of the indicator system:scientific principle;systematic principle;dominance principle;operability principle and sustainability principle.According to the above principles and the data result of the second type survey of Yongan city,the indexes required by the research was preliminarily determined.The quantitative indexes was standardized and the qualitative indexes was quantified.By using principal component analysis,the indexes was further screened according to the factor loading matrix,and then the indexes was finally determined.Due to the large differences between the data of the arbor forest,bamboo forest and shrub forest,their indexes system would be separated in order to make sure the accuracy of evaluation results.According to the result of principal component analysis,it was ultimately determined that the arbor forest would be taken 7 principal components and 14 indexes,which were the average diameter at breast height,average height,volume per mu,canopy density,soil layer thickness,thickness of humus layer,shrub layer height,shrub coverage,herb layer height,herbaceous coverage,degree of naturalness,origin,community structure and age groups.Bamboo forest would be taken 7 principal components and 11 indexes,which were the average diameter at breast height,average height,and soil thickness,thickness of humus layer,shrub coverage,herb coverage,altitude,soil type,site quality,slope position and origin.Shrub forest would be taken 4 principal components and 8 indexes,which were canopy density,shrub layer height,shrub layer coverage,herb layer height and herb layer coverage,soil thickness,thickness of humus layer and origin.The variance contribution rate of principal component analysis was considered as the principal component weight,and the weight of each index was determined and normalized according to the factor loading and the variance contribution rate of the main city.(2)The evaluation model of this study was the health comprehensive indexes model,which divides the forest health grade into 5 categories.The health comprehensive index is[8,10],means high quality,is[6,8)means healthy,is[4,6)means sub-healthy,is[2,4)means unhealthy,is[0,2)means pathological.(3)BP neural network is an efficient and accurate evaluation method.Through MATLAB,BP neural network model was established,with the health evaluation indexes as input layer and the number of neurons was equal to the indexes number.After many experiments,the neurons of hidden layer was determined to be 6.The neurons of output layer was considered as the composite index of sublots and the number is 1.500 sublots of arbor forest were randomly selected,with 350 training samples and 150 test data;300 sublots of bamboo forest,with 210 training samples and 90 test data;120 sublots of shrub wood,with 84 training samples and 36 test data.In the BP neural network built separately,the three coefficients are nearly 1,which means the fitting effect was very good and the result was credible.(4)There were a total of 39,759 sublots with an area of 2,716,001 mu in Yongan city.Healthy and sub-healthy sublots were the majority.Among them,there were 16,817 healthy sublots,with an area of 1,204,240 mu,accounting for 42.30%of the whole sublot and 44.34%of the total area.There were 14,989 sub-healthy sublots,with an area of 928,018 mu,accounting for 37.70%of the whole sublot and 34.17%of the total area.The overall state of forest ecosystem in Yongan city is healthy and sub-healthy.The average health composite index was 6.13.There were a total of 35,620 sublots of arbor forest,with an area of 2,564,856 mu.Healthy and sub-healthy sublots were the majority.Among them,there were 15,860 healthy sublots,with an area of 1,172,037 mu,accounting for 44.53%of the whole sublot and 45.70%of the total area.There were 12,589 sub-healthy sublot,with an area of 842094 mu,accounting for 35.34%of the whole sublot and 32.83%of the total area.There were a total of 3,576 sublots of bamboo forest,with an area of 119,187 mu.Sub-healthy and healthy sublots were the majority.Among them,there were 2139 sub-healthy sublots,with an area of 72060 mu,accounting for 59.81%of the total sublot and 60.46%of the total area.There were 898 healthy sublots,with an area of 29,932 mu,which accounts for 25.11%of the whole sublot and 25.11%of the total area.There were a total of 563 sublots of shrub wood,with an area of 31958 mu.Sub-healthy and unhealthy sublots are the majority.Among them,there were 261 sub-healthy sublots,with an area of 13864 mu.There were 176 unhealthy sublots,with an area of 11,656 mu.The overall state are sub-healthy and unhealthy.The average health composite index of sublot is 4.42.(5)The overall development of forest ecosystem in Yongan city from 2010 to 2016 was good.The proportion of quality forest area was stable;healthy forest has increased;sub-healthy forest has decreased;unhealthy and pathological forest decrease were not obvious.The health change trend of arbor forest was consistent with the total trend.The change of bamboo forest in each health level was to the optimal development direction and the development prospect was very good.The change of shrub wood in each health level was to the poor development direction and the development condition was not ideal.