Distribution Prediction And Evaluation Of National Reserve Forest Species Based On Random Forest Model

In recent years,China's economy and society have developed rapidly,the population's consumption of timber has increased rigidly,the gap between supply and demand has continued to expand,and the external dependence has exceeded 50%.In addition,because the log trade has been hindered,so the allocation of resources has become more difficult,and the risk of timber safety is quite prominent.In order to ensure the co-ordination of domestic timber supply and demand and promote the development of high quality timber,China proposes to establish a national reserve forest system.In the process of establishing national reserve forest demonstration base,the suitability and suitability of reserve forest species in the base must be considered comprehensively.In this paper,Ceheng County,the Buyi and the Miao Southwest Guizhou Autonomous Prefecture,is taken as the research area.The target of research were Cunninghamia lanceolata(Lamb.)Hook.,Eucalyptus robusta Smith,and Quercus glauca Thunb.who belong to national reserve forest species.Fitting Richards tree height-DBH curve based on the average height and DBH data of dominant tree species,calculating site indices of three national reserve forest species,combining the topographic data of and soil data of Forest Management Inventory,constructing a random forest regression model was constructed to predict the suitability distribution areas of three national reserve forest species,and a suitability evaluation system of national reserve forest species in Ceheng County was established.In addition,using the second class investigation,the target tree species subcompartment distribution data and non-target tree species subcompartment distribution data under various site conditions were obtained.The sample set was established based on topographic data and soil data,and the random forest classification model was constructed.The potential distribution area of target tree species in the study area was predicted,and the interaction mechanism of environmental factors was analyzed.Main results of this research are as follows:(1)Topographic and soil factors were selected as input factors and site index as output factors to construct a random forest regression model.The results showed that the RMSE was 0.60770-1.84134 and the split attribute data was 9-10,which indicated that the fluctuation range of error curve of the model was small and the overall model was stable.According to the average value of site index calculation results,the site index simulation results are converted into suitable and unsuitable areas.The suitability evaluation system of target tree species based on Random Forest regression model was established.(2)The area suitable for Cunninghamia lanceolata(Lamb.)Hook.growth in Ceheng county is 140652.33 hm2,mainly distributed in the central,southwest and northwest areas of Ceheng county;the area suitable for Eucalyptus robusta Smith growth in Ceheng county is 135006.30 hm2,mainly distributed in the central,northwest and southeast areas of Ceheng county;the area suitable for Quercus glauca Thunb.growth in Ceheng county is 165612.82 hm2,mainly distributed in the central,northwest and southeast of Ceheng county.(3)The growth of Cunninghamia lanceolata(Lamb.)Hook.and Eucalyptus robusta Smith is mainly affected by three topographic factors:altitude,slope direction and slope.The growth of Quercus glauca Thunb.is mainly affected by three environmental factors:altitude,basement rock bareness rate and soil layer thickness.The distribution of Cunninghamia lanceolata(Lamb.)Hook.,Eucalyptus robusta Smith and Quercus glauca Thunb.is mainly affected by seven environmental factors:altitude,litter thickness,soil layer thickness,slope direction,slope,basement rock bareness rate and slope position.