GANNs-CA时空数据模型在优势树种演替研究中的应用
陈文勇1,2王颖1,3李建军1张 尹1,2

(1. 中南林业科技大学,湖南 长沙 410004;2. 湖南国防工业职业技术学院,湖南 湘潭 411100;3. 数字洞庭湖南省重点实验室,湖南 长沙 410000)

GANNs-CA;种群关系;优势;演替;时空数据

Application of GANNs-CA spatio temporal data modelin the study of dominant tree species succession
CHEN Wenyong1,2, WANG Ying1,3, LI Jianjun1, ZHANG Yin1,2

(1. Central South University of Forestry and Technology, Changsha 410004, Hunan, China; 2. Hunan National Defense Industry Career Technical College, Xiangtan 411100, Hunan, China; 3.Key Laboratory for Digital Dongting Lake basin of Hunan Province, Changsha 410000, Hunan, China)

GANNs-CA; population relationship; dominant; succession; spatiotemporal data

DOI: 10.14067/j.cnki.1673-923x.2017.12.017

备注

加强林木种群内在关系的研究,将帮助人们更多地了解森林生态系统内部深层次问题。采集了常绿阔叶林永久性观察样地的林分结构、地形、土壤等数据作为研究因子。通过2003 年和2006 年的相关数据,利用遗传算法(GA)的搜寻速度快和全局寻优效果优异的特点进行神经网络(ANNs)迭代训练和检测,得到了最精确的影响因子的阈值和权重。利用元胞自动机(CA)在时空格局演化模拟和空间局部优化方面的优势,依据影响因子的阈值和权重,确定了元胞自动机的转换规则,构建了常绿阔叶林种群种间关系演变的GANNs-CA 模型。依据该模型预测了2020 年会同常绿阔叶林6 种优势种群的生长状况。结果 表明:2013 年到2020 年间观察样地的6 种优势种群将有不同程度的变化,栲树和青冈将分别减少4.65% 和3.19%;而笔罗子和刨花楠将增加3.76%和3.81%,其它优势种群则变化不大。本次研究发现观察样地中的6 种优势种群种内竞争强于种间竞争。这可能与该样地处于演替的初期阶段,优势树种种类较多,且优势度不明显有关。

The research on the relationship between forest species will help to find the Internal problem of the forest ecological system, so as to provide the theoretical basis for improving the management level of the mingled forest. In this paper, the stand structure, terrain, soil and other data in permanent observation plots of evergreen broad-leaved forest were collected and studied. Because the genetic algorithm (GA) has a fast search speed and excellent global optimization,Relevant data from 2003 and 2006 were used to perform neural network (ANNs) iterative training and testing. Finally, the threshold and weight of the influence factor are obtained exactly. Cellular automata (CA) has advantages in space-time pattern evolution simulation and spatial local optimization. So based on the threshold value and weight of the influencing factors, the conversion rules of cellular automata are determined. and the GANNs -CA model was constructed, which was used to simulate the evolution of evergreen broad - leaved forest population relationship. According to this model, the growth status of 6 dominant species of the evergreen broad-leaved forest in 2020 was predicted. The results show that from 2013 to 2020, six kinds of dominant species in observation plots will have different degrees of changes. Castanopsis fargesii and cyclobalanopsis glauca will respectively reduce by 4.65% and 3.19%, meliosma rigida and machilus pauhoi kanehira will increase by 3.76% and 3.81%, and other dominant species will change little. Under the general conditions, the evergreen broad-leaved forest has fiercer competition with the species of single excellent community than the interspecific species. This may be related to the early stages of succession on the sample area. In this plot, the dominant species are more, but the dominance is not obvious.