DEM空间分辨率的初步分析

分布式模型的输入及其参数具有时空变异性，模型的校正也依赖于网格单元的大小，因此需要确定适当的空间分辨率来描述和控制空间变化。随着分辨率的不同， DEM的精度以及由此提取的流域特征值（如高程、坡度、地形指数、河网长度）在统计特性上也会随之变化。对50 m分辨率的DEM平均取样获得150~950 m的9组DEM，对不同分辨率下提取的流域特征值进行了统计分析，并采用信息熵度量不同分辨率的信息量。

A PRELIMINARY ANALYSIS OF DEM SPACE DATA RESOLUTION

Distributed hydrological model is used to explain the effect of information (such as terrain, soil, vegetation and climate) on every points of the study basin. The inputs and parameters of distributed hydrological model change with space and time. The model's calibration depends on the resolution of grid. In order to describe and control the space change, it is important to make sure right resolution. Distributed hydrological modeling is base on the watershed characteristics extracted from digital elevation model (DEM). Watershed characteristics extracted from different DEM resolution will be statistically different. This paper statistically analyses the watershed character values (such as elevation, gradient, length of watershed network, topographic index) extracted from various resolutions. The concept of entropy has been considered a promising method in this study as it quantitatively measures the information produced by an object (watershed). Large entropy means plenty of information. We find that the coarser the resolution is, the more smoother the terrain is. Mostly, with the DEM grid size increasing, maximal elevation and various of elevation decrease, average elevation and minimal elevation increases; maximal gradient and average gradient and various of gradient decrease; maximal topographic index and various of topographic index decrease; minimal topographic index and average of topographic index increases; length of watershed network decreases. This shows the smoothness effect of resample. With the DEM grid size increasing, entropy becomes smaller and smaller. This means DEM with coarser resolution has less information. Decrease of information is in consistent with change of watershed character values. We compare the relative change of average gradient with relative change of entropy. We find that they have approximately exponential relation. The smoothness of terrain may slower the conflux, but decrease of length of watershed network will shorten conflux time, some analyses of their effects on the velocity of flow have been done.