空间尺度转换与跨尺度信息链接:区域生态水文模拟研究空间尺度转换方法综述

空间尺度转换是近年来区域生态水文研究领域的一个基本研究问题。其需要主要是源于模型的输入数据与所能提供的数据空间尺度不一致以及模型所代表的地表过程空间尺度与所观测的地表过程空间尺度不吻合。综述了目前区域生态水文模拟研究中常用的空间尺度转换研究方法,包括向上尺度转换和向下尺度转换。详细论述了2种向下尺度转换方法: 统计学经验模型和动态模型。前者是通过将GCM大尺度数据与长期的历史观测数据比较从而建立统计学相关模型, 然后利用这个统计学经验模型进行向下的空间尺度转换. 然而动态模型并不直接对GCM数据进行向下尺度的转换,而是对与GCM进行动态耦合的区域气候模型(RCM) 的输出数据进行空间尺度转换. 通常后者所获得的数据精度要比前者高,但是一个主要缺点就是并不是全球所有的研究区域都有对应的RCM。还详细论述了2种向上尺度转换方法: 统计学经验模型和斑块模型。前者是建立一个能代表小尺度信息在大尺度上分布的密度分布概率函数, 然后利用这个函数在所需的大尺度上进行积分而求得大尺度所需的信息。而后者是根据相似性最大化原则将大尺度划分为若干个可操作的小尺度斑块,然后将计算的每个小尺度斑块的信息平均化得到大尺度所需的信息。通常在计算这种斑块化的小尺度信息的时候,对每个小尺度也会采用统计学经验模型来计算代表整个斑块小尺度的信息。建议用斑块模型与统计学经验模型相集合的方法来实现向上的空间尺度转换

Spatial scaling links the information across scales

Spatial scaling has been one of the fundamental problems in eco-hydrological modeling over the past two decades. The input parameters for regional climate change impact research operating at a 1～50 km resolution cannot be directly derived from GCM (general circulation model) operating at a 100～500 km resolution. Conversely, large scale eco-hydrological models can only simulate the grid-based integrated response instead of directly parameterizing the small-scale earth surface processes. Therefore, upscaling methodology is required to scale up the information derived from the small-scale to the information required for the large-scale models;and downscaling methodology is required to scale down the information from large scale model, i.e. GCM, to the information required for regional eco-hydrological models, which operate at a much smaller scale than the one for a GCM. Several methodologies have been developed over the past two decades to undertake these non-trivial upscaling and downscaling tasks. In this paper, we discuss how the upscaling and downscaling schemes have been implemented in eco-hydrological modeling. Two primary downscaling schemes are reviewed in this paper. The first is empirical statistical downscaling, which disaggregates the information through establishing the empirical statistical relationships that link the information between small scale and large scale by comparing the large-scale values with long-term historical observation. The second scheme is dynamic downscaling, which disaggregates information by downscaling the information generated from dynamically coupling a RCM(regional climate model) with a GCM. Two upscaling schemes, empirical statistical upscaling and mosaic upscaling, are examined in this paper. Empirical statistical upscaling is achieved by assuming that the sub-grid variability of environmental variables can be represented by a probability density function (PDF), such as the VIC (variable infiltration capacity) model and the gamma distribution model. Mosaic upscaling scheme subdivides a big grid into several patches and the environmental variables are evaluated separately for each patch, and then averaged. We suggest an approach that combines the mosaic and PDF scheme for upscaling the modeling outputs from catchment to global scales.