生态学中的尺度问题——尺度上推

尺度推绎是生态学理论和应用的核心。如何在一个异质景观中进行尺度推绎仍然是一个悬而未决的科学难题,是对当今生态学家在全球变化背景下研究环境问题的重大挑战。就目前的研究,一般可分为四大类尺度推绎途径:空间分析法(如分维分析法和小波分析法)、基于相似性的尺度上推方法、基于局域动态模型的尺度上推方法、随机(模型)法。基于相似性的尺度上推方法来源于生物学上的异量关联,可将其思想延伸至空间上,研究物种丰富度、自然河网、地形特征、生态学格局或过程变量和景观指数等。基于局域动态模型的尺度上推方法需要首先确定是否进行跨尺度推绎,以及是否考虑空间单元之间的水平相互作用和反馈,然后再应用具体的方法或途径,如简单聚合法、有效值外推法、直接外推法、期望值外推、显式积分法和空间相互作用模拟法等。随机(模型)法以其它尺度上推方法为基础,根据研究的是单个景观,还是多个景观,采用不同的途径。理解、定量和降低尺度推绎结果的不确定性已经变得越来越重要,但相关研究仍然极少。以上所有有关尺度推绎的方法、途径和结果分析共同构成了尺度推绎的概念框架。

Scale issues in ecology: upscaling

Scaling means transferring information between or across spatial and temporal scales or organizational levels. Transferring from finer scale to broader scale is called as upscaling, and transferring from broader scale to finer scale is called as downscaling. In both basic ecology and its applications, scaling is the essence of predicting and understanding a phenomenon, and is at the core of ecological theories and applications. However, scaling is often very complex because scaling has to overcome constraints and critical thresholds between different systems, non-linear interactions among different components always occur within the same scale and among different scales, and especially spatial heterogeneity always exists. Therefore, scaling across heterogeneous ecosystems remains an unresolved puzzle, greatly challenging current ecologists devoting to studying environmental problems under global change. As a whole, four general scaling approaches can be distinguished: spatial analysis, similarity-based scaling, local dynamic model-based scaling, and random model approaches. Spatial analysis approach is based on spatial pattern analyzing, such as fractal and wavelet analysis method. Similarity-based scaling approach is an important approach and has been widely used in physics, earth science, hydrology, meteorology, and biology. The similarity-based biological allometry reveals the relationships between biological characteristics and body sizes, while spatial allometry reveals the relationships between landscape characteristics (such as species richness, natural river network, landform features, ecological variables, and landscape metrics) and spatial scales. The simple power law in allometric relations might be the integration of extremely complex underlying processes and mechanisms, possibly related to ubiquitous fractal structure of biological body and landscape, but the validation of this hypothesis will has to be conducted in developing scaling theory. However, allometric relations may only exist within a limited range of scales, beyond which some new processes will occur and the allometric relations at these scales cannot be extrapolated to other scales. Three key problems have to be addressed in local dynamic model-based approach: building a dynamic model at local scale, accurately defining and quantifying spatial heterogeneity of model parameters and input variables at local scale, aggregating or integrating heterogeneous information of output variables at local scale to object scale. The differences in both quantifying heterogeneity and aggregating information decide the merits and lacks of each method within this approach. Firstly, if scaling is conducted between adjacent scales, and the interactions among different spatial units can be ignored, lumping, extrapolation by effective parameters, direct extrapolation, extrapolation by expected value, and explicit integration methods can be used. Secondly, if scaling is conducted between adjacent scales, and the interactions between different spatial units cannot be ignored, lumping, extrapolation by effective parameters, and direct extrapolation methods still can be used, but extrapolation by expected value and explicit integration methods cannot be used any longer. Especially, spatially interactive modeling is another important approach to realize upscaling by developing multiple-scaled models to directly model the interactions among different spatial units. Thirdly, if scaling is conducted among multiple scales for a hierarchical landscape, scaling ladder approach (i.e. hierarchical patch dynamics strategy) can be used. Random model approach is based on the other upscaling approaches or methods. According to whether scaling is conducted in a single landscape or over multiple landscapes, different approaches may be adopted. Understanding, quantifying and reducing the uncertainty in scaling results have become more and more important, but the related studies still extremely lack. All the above methods, approaches and analysis will contribute to the conceptual framework of scaling science.