| As a basic tool to study the dynamic law of stand growth change and the prediction of forest growth and yield,forest growth and yield model is highly valued and is an important part of modern forestry work.Stepwise regression is widely used in modeling the relationships.However,this method ignores the model uncertainty due to the variable selection process.Bayesian model averaging method(BMA)selects all possible models and uses the posterior probabilities of these models to perform all inferences and predictions.Data for this study were sampled from Chinese fir(Cunninghamia lanceolata)stands in Weimin,Shaowu city,Fujian province.The plantations were established at five initial planting densities(A:1667,B:3333,C:5000,D:6667,and E:10,000 trees/hm~2).Each treatment level established using a randomized block design in each plantation plot was replicated three times.From 1984 to 2010,there were17 instances of field sampling over 20 years.BMA and stepwise regression were used to analyze the effects of stand endogenous factors(individual tree size factors,competition factors,stand factors)and climatic factors on the growth of Chinese fir.The following four studies were conducted:(1)Analysis of driving factors and uncertainty of individual tree mortality;(2)Model development and uncertainty analysis of height-diameter allometry model;(3)Analysis of factors and model development of individual tree annual diameter increment;(4)Analysis of driving factors and model development of stand volume.Results showed that:1.In general,stand endogenous factors and climatic factors are all the factors affecting the mortality,but these factors are different in different initial planting densities.Competition between trees was small when stands are in medium and low density(1667-5000 trees/hm~2).The mortality increased with the increasing stand basal area per ha and dominant height,and decreased with the increasing stand density(living number of trees per ha),DBH,age and winter mean minimum temperature.Both endogenous and climatic factors significantly affected the mortality in high density stands(6667-10000 trees/hm~2)(posterior probability≥0.75).The mortality increased with the increasing living number of trees per ha,stand basal area,mean diameter,dominant height,age and mean annual temperature.Tree mortality decreased with the increasing DBH,basal area for all trees greater than the subject tree(BAL),mean coldest month temperature,mean annual precipitation,annual heat-moisture index,degree-days below 0°C,summer mean maximum temperature,winter mean minimum temperature and spring mean temperature.Generally,the mortality was positively correlated with competitive stress and decreased with increasing temperature and precipitation.2.Tree height growth was influenced by both endogenous and climatic factors.But at medium and high density(5000～10000 trees/hm~2),climate factors have little effect on tree height growth(posterior probability≤0.75).Tree height increased with the increase of dominant height and age,and decreased with the increase of mean DBH,mean annual precipitation,annual heat-moisture index and winter mean minimum temperature.Trees in stands planted with higher density were higher than in lower density stands.In general,the increase of competition will promote the growth of tree height,and too much water will inhibit the growth of tree height.Furthermore,except for the D density(6667 trees/hm~2),the scale of diameter is about 0.6.3.Both endogenous and climatic factors affect individual tree annual diameter increment.However,annual diameter increment was less affected by climate factors.Competition factors and individual tree size factors were the main factors affecting annual diameter increment(posterior probability≥0.75).Diameter increment decreased with increase of density,quadratic mean diameter,BAL,ln(A)and winter mean minimum temperature,and increased with increasing diameter at the beginning of growth period,stand basal area,dominant height,mean coldest month temperature,mean warmest month temperature and mean annual precipitation.Overall,the diameter increment decreased with increasing competition and increased with increasing temperature and precipitation.4.Both endogenous and climatic factors affected stand volume.Stand volume increased with increasing basal area,quadratic mean diameter,dominant height,age,mean annual precipitation,annual heat-moisture index,degree-days below 0°C,summer mean maximum temperature,winter mean minimum temperature and spring mean temperature.Stand volume decreased with increasing initial planting density,mean annual temperature and mean warmest month temperature(posterior probability≥0.75).Higher stand quality,temperature and precipitation will promote the volume increase,but too high temperature and excessive competition will inhibit stand growth.5.For most of the treatments of the four models,the posterior probability of the model obtained by stepwise regression method was less than the best model obtained by BMA(the highest posterior probability).Or some stepwise regression models are not in the first few models with higher posterior probability selected by the BMA.Therefore it leads to model uncertainty.In addition,stepwise regression method is easier to select redundant variables.But there is no significant difference in the evaluation indicators of model prediction between two methods,which may because the large sample. |
