利用混合模型分析地域对国内马尾松生物量的影响

开展全国森林生物量监测和评估,建立适合较大区域范围的通用性立木生物量模型是一项重要的基础工作,而分析森林生物量受不同地域的影响并保证不同尺度范围森林生物量估计值的可靠性,是必须面临的问题。以南方马尾松(Pinus massoniana)地上生物量数据为例,介绍了如何利用混合模型理论来分析地域对马尾松地上生物量的影响以及利用混合模型构建全国通用性立木生物量模型,为得到不同区域尺度范围内可靠的森林生物量评价和估计提供了有效途径。结果表明,混合模型不仅提高了模型的精度和通用性,并且模型中每个参数都有特定的数学含义,通过这些参数很容易分析出随机因子对生物量的影响程度。因此混合模型方法具有较大的灵活性和适应性,可推广到其它通用性模型(如材积方程)的建立。

Analysis the effect of region impacting on the biomass of domestic Masson pine using mixed model

It is fundamental work for monitoring and assessing national forest biomass in order to develop generalized single-tree biomass models for large scale forest biomass estimation. However, the reliability of estimating forest biomass among different scales is a real problem. Mixed model plays an important role in various industry such as pharmacokinetics, agriculture and medicine, because of its outstanding power in dealing with repeated-measures data, longitudinal data, blocked design data, multilevel data and the capacity to make accurate local prediction. In this article, we consider the influence of origin and region of Masson pine (Pinus massoniana) to biomass, because the data including origin and region could be considered as the two level blocked data, then, the mixed-effects model can be used to analysis these kinds of problem. The mixed-effects model approach is an excellent statistical technique for parameter estimation, and it has been used in many fields for nearly twenty years. However, in recent years, it starts to use mixed-effects model in forestry, and it is much less to use mixed model in building biomass model. In this paper, based on the above-ground biomass data of Masson pine, which is an important coniferous species in southern China, generalized single-tree biomass equations for national and regional forest biomass estimation were established by using linear mixed model, which provided effective approaches to solve the compatibility of forest biomass estimation among different scales. In mixed model, the random parameters could reflect difference by assuming the data distributing normally, and these parameters cancel each other out with an expected value of zero. In fact, we can consider that the difference among these samples is divided into two parts in mixed model: one oriented from the difference among types; another resulted from the random effects. The result shows that based on the above-ground biomass data of Masson pine in southern China, the generalized single-tree biomass equations for national and regional forest biomass estimation were developed by using linear mixed model, which could solve the universality of forest biomass estimation among different scales. The fitting result of subject-specific models shows that the above-ground biomass estimation of trees with the same diameter is varied by different origins and regions. To the Masson pine in southern China, the above-ground biomass of a tree with the same diameter in natural forest is larger than that in plantation; and the biomass estimation decreases from the eastern region (Jiangsu, Zhejiang, Fujian) to the south-central region (Jiangxi, Hunan, Guangdong), and then the north-western region (Anhui, Guangxi, Guizhou). If we consider origin and region together, different patterns of the above-ground biomass estimation would appear: for natural forests, trees with the same diameter in eastern region have the largest biomass; whereas, for plantations, trees in south-central region have the largest biomass. In conclusion, the linear mixed model with random parameters is better than traditional model. Especially, the linear mixed model is more flexible because all the parameters in the model have certain mathematics meaning, and which can be used to solve different ecological phenomenon. In view of these advantages, the mixed model could be applied to develop other generalized models such as tree volume equations.