| Forest is the material basis for the survival of human society.It not only provides forest products and biological habitat for the outside world,but also has a variety of important ecological functions.Stand growth and harvest model can judge the response of specific forest types to planned management and predict the future development trend of stand.Therefore,it is often used as a technical means of forest management decision-making.As an important force in the construction of ecological civilization in China,state-owned forest farms have played an important role in providing ecosystem service functions,maintaining national ecological security and coping with global climate change.Based on the theories and methods of forest biological characteristics and ecological modeling,building a whole stand growth and harvest model with certain accuracy is helpful to improve the level of sustainable forest management in state-owned forest farms.Therefore,this study takes the arbor forest resources of Gonglongping state-owned forest farm as the research object.Using the three-phase forest resource survey data completed in 2006,2016 and 2020respectively,according to the relevant technical regulations,the arbor forest resources in the study area are classified into five forest types:Pinus armandii forest,Pinus yunnanensis forest,Cunninghamia lanceolate forest,hard broad-leaved forest and soft broad-leaved forest,After eliminating the abnormal data,442 stand division of arbor forest were retained,and 254 groups of stand division data that had not been operated during the investigation period were screened.190 groups(75%)were randomly selected as the modeling data,and the remaining 64 groups(25%)were used as the test data;In addition,58 stand division were randomly selected to conduct a sample plot per tree survey.The sample plot survey data were used as the verification data to test the accuracy of the data of the small class and the practicability of the model.The sample plot size was 20 m×30 m.On the basis of analyzing the characteristics of forest resources in the study area,using the theory and method of nonlinear mixed effect model,this paper attempts to build a whole stand growth and harvest model with forest types as random effects,to describe the growth process of stand average height,average DBH,plant number density,density index,stand sectional area and volume,and then compile the stand growth process table of each forest type.The main conclusions of this paper are as follows:(1)Mitscherlich model(Ra~2=0.658)is the optimal basic model to simulate the average height growth of stand,while the Ra~2,AIC and RMSE values of mixed effect model are 0.794,1358.79 and 1.808 respectively,which are better than the basic model,indicating that increasing the random effect of forest type level can improve the fitting accuracy of the model.(2)Schumacher model(Ra~2=0.647)is the optimal basic model to simulate the average DBH growth of stand.When considering the introduction of site index as covariate or forest type as random effect,the fitting accuracy of the model can be improved.The Ra~2,AIC and RMSE values of the generalized mixed effect model are 0.881,1538.44 and 2.447 respectively,which are better than the other three models.The fitting accuracy from large to small is the generalized mixed effect model,the basic mixed effect model Generalized model,basic model.When simulating the change of stand number density,considering the influence of site conditions and forest types can also improve the fitting accuracy of the model.(3)Constructing mixed effect model can improve the fitting accuracy of stand basal area and volume growth model.The Ra~2,AIC and RMSE values of the mixed effect growth model of stand basal area were0.835,923.129 and 4.12 respectively,and the Ra~2,AIC and RMSE values of the mixed effect growth model of stand volume were 0.981,-213.637 and 0.107 respectively,which were better than the fixed effect model,indicating that increasing the random effect of forest type level can improve the fitting accuracy of the model.(4)Using the common population survival curve to fit the stand number density model,when considering the introduction of site index as a covariate or forest type as a random effect,the fitting accuracy of the model can be improved.After the basic model is fitted,Ra~2 is 0.425.When considering the introduction of site index as a covariate to build the generalized model,Ra~2(0.443)increases,while AIC,RMSE,MAE and MAPE decrease;When the generalized mixed effect model is constructed by increasing the random effect of forest types,the accuracy of the model is further improved(Ra~2=0.468).According to the maximum density line equation,the slope of the maximum density line of each forest type can be calculated.Combined with the stand average DBH and plant number density model,the functional equation of the stand density index prediction model can be obtained.(5)The mixed effect model can improve the fitting accuracy of the prediction model of stand basal area and volume growth.The Ra~2,AIC and RMSE values of the mixed effect growth prediction model of stand basal area were 0.659,63.569 and 0.283 respectively,and the Ra~2,AIC and RMSE values of the mixed effect growth prediction model of stand volume were 0.789,56.398 and 0.265 respectively,which were better than the fixed effect model,indicating that increasing the random effect of forest type level can improve the fitting accuracy of the model.To sum up,this study explored and constructed a systematic theory and method of mixed effect model based on the whole stand growth and harvest model system of the state-owned forest farm with the forest type as the random effect,analyzed and revealed the impact of different site conditions and dominant tree species on the stand growth,and the results showed that considering the random effect of increasing the site index factor and forest type level can effectively improve the fitting accuracy of the stand growth model. |
