Study On Growth Model And Economic Maturity Of Traditional Regression And Machine Learning

Optimizing forest management is the fundamental way to improve forest quality.The accurate estimation of DBH and accumulation is an important part in combination with site conditions,stand density and other factors.Due to the insufficient consideration of site conditions and stand density,the current stand DBH and stock models of eucalyptus and Chinese fir in guangxi are not accurate.In this paper,two methods of modified model and machine learning decision tree are used to construct the prediction model of DBH growth and accumulation in stands.On this basis,three economic maturity criteria were used to calculate the economic benefits and the corresponding economic maturity age of Chinese fir under different afforestation densities.Finally,a forest growth prediction and economic benefit simulation system was developed.The main research of this paper is as follows:(1)by using the modified model method,Richards et al.6 theoretical growth equations were used as the basic function,and the power function product combination of status index and plant density as the base number was used as the error function to construct the stand DBH and accumulation correction models of eucalyptus and Chinese fir.The results showed that: eucalyptus DBR R2=0.738,MAE=1.158,cunninghamwood DBR R2=0.782,MAE=1.751,eucalyptus accumulation model R2=0.772,MAE=9.235,cunninghamwood accumulation model R2=0.785,MAE=7.540.(2)using decision tree class 9 kinds of machine learning method,to the second class survey site factor of subcompartment data(altitude,slope direction,slope position,slope,litter thickness,thickness of humus layer,soil thickness,soil type)and stand factors(age,ha plant)as input,with an average diameter at breast height,volume as the output,construction of eucalyptus and Chinese fir stand DBH and volume decision tree model.The results show that for decision tree machine learning,integrated learning is higher than non-integration,serial integration is higher than parallel integration,and the performance of XGboost decision tree model is relatively better.For XGboost model,eucalyptus DBR R2=0.7410,MAE=0.382,accumulation R2=0.8125,MAE=0.3434;Chinese fir DBH R2=0.9320,MAE=0.3446,accumulation R2=0.9972,MAE=0.3446.For the proportion of XGboost model variables,eucalyptus accumulation was in order of age(78.0%),altitude(4.9%),soil thickness(3.8%),and density(3.2%).The age(80.2%),altitude(5.0%),soil thickness(4.1%),and density(3.6%)of China fir accumulation model were in order.(3)take the net present value,forestland expected value and forestland expected value considering the carbon sequestration benefit of forestland as the economic maturity judgment index,take a certain site condition as an example,fix the independent variable of the site,and carry out economic benefit simulation under 5 density conditions.The results show that: The economic benefit value starts with a negative value,increases with the growth of forest age,and begins to decrease to the maximum value,and decreases with the increase of density.The economic benefit value is relatively largest when the stand density is 1000 plants per hectare,corresponding to the maximum net present value(21202.71 Yuan / ha),the forest land expectation(29930.50 Yuan / ha),and the forest land expectation of carbon sequestration benefit(32423.60 Yuan / ha).The economic maturity age corresponding to the net present value is 16 a,and the expected value of forest land and the expected value of forest land considering the carbon sequestration benefit of forest land are both 15 a.