Uncertainty Modeling Of Regeneration,Recruitment And Mortality For Pine-oak Stands In The Qinling Mountains Under A Bayesian Framework

Simulating forest regeneration,recruitment and mortality is crucial to the understanding of forest dynamics such as succession and diversity.It is also one of the key components of stand growth prediction and forest management planning.However,this simulation is a complex process affected by competition,site and environmental factors,and has a high degree of uncertainty.Therefore,the simulation of forest regeneration,recruitment and mortality has always been a difficult problem in forest ecosystem modeling and uncertainty analysis.The quantification and disaggregation of the predictive uncertainty can improve the understanding of the complexity and uncertainty of forest dynamics,which provides a scientific basis for simulation and management planning of forest ecosystem.Based on the above issues,this study constructed forest regeneration,recruitment and mortality models in pine-oak forests of Qinling Mountains based on a Bayesian statistical framework.We studied the uncertainty of model prediction and explained the main factors affecting forest regeneration,recruitment and mortality,especially for events with low probability.The main content and results are as follows:(1)A natural regeneration model of pine-oak forests in Qinling Mountains was developed with competition,site and climate factors using Bayesian statistics and global sensitivity analysis(GSA).Poisson,Negative Binomial(NB),Zero-Inflated Poisson(ZIP),and Zero-Inflated Negative Binomial(ZINB)models were tested as candidate model forms.The results showed that the ZINB model was superior to other model forms and was suitable to simulate forest regeneration with excessively discrete data or excessive zero data.Stand basal area,light interception,slope location,and minimum temperature during the growing season were the most critical factors that affected natural regeneration of Pinus tabulaeformis,while stand basal area,cosine of aspect interacted with the natural logarithm of elevation(CE),annual mean temperature,and precipitation of the warmest quarter were the most critical factors for Quercus aliena var.acuteserrata.The uncertainty of parameter related to topographic factors was higher than that of competition factors and climate factors in regeneration simulation of P.tabulaeformis and Q.aliena var.acuteserrata,which can be used as the main target for optimization in future research.The natural regeneration of P.tabulaeformis was positively correlated with annual mean temperature and minimum precipitation during the growing season,and negatively correlated with the mean temperature of the driest quarter.The natural regeneration of Q.aliena var.acuteserrata was positively correlated with annual mean temperature,minimum precipitation during growing season,precipitation of the warmest quarter,and negatively correlated with mean temperature of the driest quarter.The study shows that the ZINB model based on Bayesian methods may effectively quantify the major factors in forest regeneration simulation and interpret the uncertainty propagated from parameters,which is useful for predicting forest regeneration.The response of the natural regeneration number of P.tabulaeformis and Q.aliena var.acuteserrata to temperature was more sensitive than to precipitation.(2)A height increment model for young P.tabulaeformis was built using tree-level field-inventory data of pine-oak forests in Qinling Mountains.The effect of uncertainty sources on model predictions was then analyzed by disaggregating the predictive uncertainty into contributions from every single parameter.The results provide a theoretical basis for improving the reliability of young P.tabulaeformis modeling.The Markov Chain Monte Carlo(MCMC)method was used to obtain the joint posterior distribution of parameters,and to quantify the uncertainty of model outputs,in terms of the uncertainty of prediction error,measurement errors of the inputs,and the parametric uncertainty.A combination of Bayesian statistics and global sensitivity analysis(GSA)was used to quantify the uncertainty propagation for each parameter.The 95% credible interval of model prediction included 97%of observations,and sufficiently covered the ranges of random errors of observed data.The least uncertainty source was parametric uncertainty,accounting for 43% of total uncertainty.The minimum uncertainty source was the measurement errors of model input,i.e.light interception(LI)and crown competition factor(CCF),accounting for only 6% of total uncertainty.The parameter relating to CCF resulted in the largest contribution to the uncertainty of the predictions,and the propagated uncertainty attributed 64.87% of total parametric uncertainty.Parameters of LI and slope(SL)propagated 15.88% and 10.02% of total parametric uncertainty,respectively.The parameter of height accounted for only 1.78%,and the uncertainty contributed from other parameters was less than 1%.The effects of CCF,LI and SL on the 5-year height increment of young P.tabulaeformis were negative,but positive for tree height effects.The results revealed that the higher parametric uncertainty,the weaker effects of corresponding variables on predictions.(3)Stand recruitment is an important variable to describe the development of stand dynamics.Since a relatively high number of the plots have no occurrences of recruitment,the data are discrete with high uncertainty.A recruitment model of pine-oak forests in Qinling Mountains was developed under a Bayesian framework using remeasurement data.A combination of Bayesian inference and Global Sensitivity Analysis(GSA)was applied to quantify the uncertainty from different sources,including the model form,parameters,and data.The contribution of each parameter to the predictive uncertainty was quantified and interpreted.The results showed that ZINB model had the best goodness of fit and highest accuracy when simulating the recruitment of P.tabulaeformis and Q.aliena var.acuteserrata.The effects of stand basal area on tree recruitment of Pinus tabulaeformis and Quercus aliena var.acuteserrata were negative,while effects of tree density and site class index were positive.More recruitswas exhibited for Q.aliena var.acuteserrata in comparison with P.tabulaeformis,especially for the stands with higher density of Q.aliena and on fertile sites.Recruitment of P.tabulaeformis was more sensitive to the decrease in stand basal area,and the number of recruits were equal to or more than that of Q.aliena var.acuteserrata on poor site condition.The stand basal area related parameter contributed the smallest proportion of uncertainty in the simulations of P.tabulaeformis recruitment,and the parameters relating to tree species density and site class index had larger contributions to the uncertainty of the predictions for P.tabulaeformis.The parameter relating to tree density had the smallest contribution to the uncertainty of the predictions,and the parameters relating to stand basal area and site class index had larger contributions to the uncertainty of the predictions for Q.aliena var.acuteserrata.(4)Mortality models for pine-oak forests in Qinling Mountains were developed under a Bayesian framework based on competition,site and climate factors using remeasurement data.The main factors affecting stand mortality were detected and interpreted.The results showed that ZINB model had the best goodness of fit and highest accuracy in the simulation of mortality for P.tabulaeformis and Q.aliena var.acuteserrata.The effects of stand density,dominant height,SMT and MASP on tree mortality were positive,while the effects of MCMT and average diameter at breast height were negative in the mortality simulations of P.tabulaeformis.The effects of density of tree species,slope and MWMT on tree mortality were positive,while effects of MASP and average diameter at breast height were negative in the mortality simulations of Q.aliena var.acuteserrata.The parameters relating to tree species density,MCMT and SMT had smaller contributions to the uncertainty of the predictions for P.tabulaeformis,and the parameters relating to dominant height,average diameter at breast height and MASP had larger contributions to the uncertainty of the predictions.The parameters relating to specific tree species density,SIC and MWMT had smaller contributions to the uncertainty of the predictions for Q.aliena var.acuteserrata,and the parameters relating to the dominant height and MASP had larger contributions to the predictive uncertainty.The results provide a scientific basis for forest mortality prediction and management under climate change.In conclusion,the simulation of forest regeneration,recruitment and mortality was associated with a high risk of uncertainty.The Bayesian modeling approach expresses each element in the form of probability distribution in model construction.Thus,the uncertainty of the model prediction can be quantified,and the contribution of each parameter to the total predictive uncertainty can be explained.Such a Bayesian approach is capable of quantification,interpretation,parameter calibration and model modification for uncertainty analysis in simulating forest ecosystem dynamics.