摘要
以黑龙江省孟家岗林场的长白落叶松(Larix olgensis)人工林为研究对象,基于树木生长轮宽度数据构建ARIMA模型,以解析径向生长的时间序列特征并模拟其动态过程。通过对167根树芯样本进行COFFCHA交叉定年法与年表建立,利用相关性分析量化生长轮宽度与气候要素的关系。经ADF单位根检验得到平稳时间序列后,依据AIC和BIC准则确定最优模型为ARIMA(2,1,0),并引入基于残差方差的蒙特卡洛模拟方法,构建未来50 a径向生长的多情景预测,以反映气候变化背景下的生长波动。最终保留139条有效序列,主序列年限为62 a,相关系数为0.493,平均敏感度为0.303。模型在2022—2023年预测中表现良好,均方根误差为0.1319,平均绝对误差为0.1015,平均绝对百分比误差为15.52%,所有预测值均落入90%置信区间,拟合效果优良。Ljung-Box检验进一步验证残差序列为白噪声,模型无遗漏结构。研究结果表明:该模型能够在不依赖大量外部气候数据的情况下,实现对林分未来径向生长的有效中长期预测,适用于数据相对有限的人工林经营实践。结合随机扰动模拟生成多种生长情景,有助于评估未来不同气候条件下林分生长的波动性与潜在风险,从而为人工林抚育、轮伐期确定及碳汇动态管理等提供量化支持。
Taking the Larix olgensis plantations in Mengjiagang Forest Farm of Heilongjiang Province as the research object,an ARIMA model was constructed based on tree ring width data to analyze the time⁃series characteristics of radial growth and simulate its dynamic processes.Cross⁃dating and chronology establishment for 167 tree core samples were conducted using the COFFCHA method,and correlation analysis was applied to quantify the relationship between tree ring width and climat⁃ic factors.After the stationary time series was obtained via the ADF unit root test,the optimal model was identified as ARI⁃MA(2,1,0)according to the AIC and BIC criteria.Furthermore,the Monte Carlo simulation method based on residual va⁃riance was introduced to construct multi⁃scenario predictions of radial growth over the next 50 years,so as to reflect the growth fluctuations under the background of climate change.Finally,139 valid sequences were retained,with the master chronology spanning 62 years,a correlation coefficient of 0.493,and a mean sensitivity of 0.303.The model exhibited good performance in the predictions for 2022-2023,with a root mean square error of 0.1319,a mean absolute error of 0.1015,and a mean absolute percentage error of 15.52%.All predicted values fell within the 90%confidence interval,indicating excellent fitting results.The Ljung⁃Box test further verified that the residual sequence was white noise,suggesting that the model had no omitted structures.The research conclusions show that the model can realize effective medium⁃and long⁃term predictions of the future radial growth of forest stands without relying on a large amount of external climatic data,and is ap⁃plicable to the management practices of plantations with relatively limited data.The generation of multiple growth scenarios combined with random perturbation simulation is conducive to evaluating the growth volatility and potential risks of forest stands under different future climatic conditions,thereby providing quantitative support for plantation tending,rotation age determination,and dynamic carbon sink management.
作者
贺思雨
金星姬
童茜坪
李凤日
He Siyu;Jin Xingji;Tong Qianping;Li Fengri(Northeast Forestry University,Harbin 150040,P.R.China)
出处
《东北林业大学学报》
北大核心
2026年第4期1-8,共8页
Journal of Northeast Forestry University
基金
国家自然科学基金区域创新发展联合基金项目(U21A20244)。
