基于遥感资料的陆面水循环模拟及检验

The Simulation and Validation of the Land Surface Water Cycle Based on Remote Sensing Data

利用遥感资料和10年逐日气象资料,采用逐步插值方法(SIA),修改VIC-3L(Three-Layer Variable In-filtration Capacity hydrological model)模式中多种覆盖类型蒸发的计算方法,计算植被对地表水循环的贡献。模拟地处南水北调工程水源区的褒河(陕南汉江支流)流域地表水的空间分布,将应用遥感资料与《中国资源环境数据库》资料的模拟结果进行比较,并用实测资料进行检验,结果表明:(1)有效的气候要素空间插值方法,是陆面水循环模拟的基础;(2)4 km分辨率DEM(Digital Elevation Model data)得到褒河流域的河网,与实际状况相吻合;(3)利用VIC-3L模拟褒河流域的陆面水文过程中,水分分量的分布合理,物理成因清晰,量值大小准确;(4)基于遥感资料模拟的地表水平衡分量的空间分布更为合理,尤其是改善了河流流量峰值的模拟;(5)遥感植被类型、叶面积指数较好地反映了实际地表覆盖状况和叶面积指数的季节变化情况。

The Yangtze River is the longest river in China and the third one in the world, with a main channel length of 6300 km and a drainage area of 1800000 km^2. The Yangtze River basin includes large urban areas and significant croplands which play important and critical roles in Chinese economy and the health of the ecological environment. There are many notable projects associated with the Yangtze River basin, such as the Three Gorges Dam project, and the water transferring project from south to north. Understanding of the natural terrestrial hydrological processes over the Yangtze River basin and the interactions between land surface and atmosphere, and the impact of the Three Gorges Dam project on its ecological environment is critical and challenging. As the first step in our effort of simulating the terrestrial hydrological process for the entire Yangtze River basin, a spatially distributed Three-Layer Variable Infiltration Capacity hydrological model （VIC-3L） is applied to the Baohe River watershed with a drainage area of 2500 km^2. Water fluxes of this watershed are simulated using VIC-3L model at a spatial resolution of approximately 4 km. The soil and land cover properties from the Resource and Environment Database （1： 4000000） of China （REDC） and vegetation and land cover data derived from MODIS data over the study area are used. Measurements from a number of weather stations are used to obtain meteorological forcings for each modeling grid based on successive interpolation approach （SIA）. The VIG-3L model is run at a daily time step. The water component differences between two sources of land cover data are compared. Simulation results are tested using daily stream flow measurements at the outlet of Baohe River basin and Jiangkou in Shaanxi Province, China, from 1992 to 2001. The results show that SIA can efficiently improve climate variable interpolation and reduce the calculating errors. VIC-3L model can simulate reliably the surface water components, such as, evapotranspiration, runoff, snow sublimation and soil moisture. The results show that these components have closer relationship with the vegetation distribution. The calculated evaporation at the meteorological station and discharge at the hydrological station possess annual and seasonal fluctuations consistent with the observations. There are evident differences of evapotranspi ration, soil moisture and runoff, resulting from Remote Sensing （RS） data and REDC data. The discharge verification indicates that the use of RS data [including Leaf Area Index （LAI） and land cover] can improve to some extent the simulation at the hydrological outlet station, because the RS data is more veritable than that of REDC. The surface water components are more accurately simulated by use of remote sensing data, verified in terms of the stream flow measurement. The simulation errors of discharge peaks are obviously reduced because the RS coverage and LAI are more reliable and veritable. All of the results show that the RS data, such as land cover and LAI, are helpful to improve the runoff simulation. The reasons are that the RS LAI reveals in fact the seasonal variation of leaf area index, and RS can provide present pattern of land cover. The REDC land cover is probably outdated because the existing six types of vegetations are represented by the REDC as massive homogenous patches. In fact, there are usually mixed forests in the Baohe River basin. The representiveless deciduous broadleaf forest of the REDC results in increase in transpiration, and decrease in runoff and baseflow, which then decreases river discharge. This is the reason that the simulated peak river discharges with the REDC are much less than the observations, and those with RS data are better approximations to the observations.