基于多源遥感数据的疏勒河上游山区流域VIC-CAS模型积雪模拟效果评估

积雪积累和消融过程是冰冻圈水文模型的重要组成部分，利用多源遥感数据对水文模型模拟的积雪分布和深度进行评估是进一步增强融雪过程模拟的物理基础，也是提高模拟可靠性的重要手段。基于2002—2013年疏勒河上游山区流域MODIS地表反射率数据集和中国雪深长时间序列数据集，对VIC-CAS模型模拟的逐日积雪覆盖度和雪深进行了综合评估。结果表明：从不同降雪年份来看，VIC-CAS模型可以较好地模拟多雪年（2008年）疏勒河上游山区流域积雪的覆盖度，在平雪年（2004年）和少雪年（2013年）模型模拟精度相对较低。从不同海拔的模拟结果来看，在流域占比最高的4 000~5 000 m高程带精度最高，2 000~3 000 m高程带精度最低；对比模拟雪深与中国雪深产品发现，多雪年的一致性较高，平雪年和少雪年的一致性较低。这表明VIC-CAS模型对疏勒河上游日尺度的积雪覆盖度和雪深的模拟精度相对较低，特别在低海拔处和薄雪情况下，其原因可能是对积雪再分布和风吹雪过程的模拟算法和参数化存在较大的不确定性，需要进一步改进。

Evaluation on snow coverage and snow depth simulated by VIC-CAS model based on multi-source remote sensing data in mountainous upper reach of the Shule River basin

The accumulation and melting processes of snow are the important parts of the cryospheric hydrological model. Generally， the simulated snow cover and snow depth time-series by distributed hydrological model were difficult to evaluate， which mainly calibrated and validated by observed runoff data. The multi-source remote sensing snow cover and snow depth products give a good choice for evaluating the spatial-temporal patterns of snow cover and snow depth of simulation， which probably help to enhance the physical basement of model. However， due to the cryospheric components include glacier， snow cover， permafrost always distributed together in one basin， the evaluation on snow cover in one basin should base on that there are enough precipitation observation data in alpine areas of the basin， and glacier meltwater were reasonable simulated， which indicated that the large uncertainties of simulated cover were removed. There are more than 20 precipitation observation instruments were installed since 2009 in the upper reach of the Shule River basin （URSRB） with average elevation above 4 000 m a.s.l.， and the annual precipitation gradient were obtained with 14.654 mm·（100m）^-1. The river runoff of Changmabao in URSRB was successfully simulated by VIC-CAS model， which coupled with glacier modules with VIC-3L model， and the simulated single glacier area changes were compared with that observed from multiple temporal remote sensing data. The simulation suggested that it well represent the glacier meltwater and glacier change. Thus， the simulated snow coverage and snow depth by VIC-CAS model were evaluated by remote sensing products in URSRB. Based on 4 383 daily MODIS surface reflectance datasets and the Long-term Snow Depth Dataset of China from 2002 to 2013， in which the snow depth algorithm developed by Che and Dai of Cold and Arid Regions Environmental and Engineering Research Institute， Chinese Academy of Sciences， the daily snow cover and snow depth simulated by VIC-CAS model were comprehensively compared in URSRB. All the forcing data， parameters， and the calibration and validation processes of VIC-CAS model in URSRB are the same with previous published literature. The comparison of stats in monthly average snow coverage and spatial pattern of annual average snow coverage in 182 sub-basins suggested that VIC-CAS model can better simulate the snow cover a in more snow year such as 2008， which has the higher relative coefficient （r） by 0.67 and lower root mean square error （RMSE） by 0.12. The simulation accuracy of VIC-CAS model is relatively lower in normal year such as 2004 with r by 0.37 and RMSE by 0.13 and less snow year such as 2013， which has r by 0.52 and RMSE by 0.09. The spatial distribution of annual average simulated snow cover has the similar patterns with that from remote sensing data， especially in more snow year， although the simulated snow coverage is less than observed by remote sensing data. The best simulation of snow coverage is located at 4 000~5 000 m a.s.l. altitude zone， which has high r by 0.44， 0.66， 0.60， and RMSE by 0.15， 0.12， 0.11 in normal year， more snow year， and less snow year， respectively， while that is the worst at 2 000~3 000 m a.s.l. altitude zone， which has r by -0.1. The consistency of snow depth between simulated and observed is high in 2008， while is low in other years. The simulated annual average snow depth is less than observed by remote sensing data. The spatial distribution of simulated snow depth suggested that it has more relationship with altitude than snow depth product， due to the later has coarse resolution. These results indicate that VIC-CAS model has a lower accuracy in the low altitude area or with thin snow， which probably comes from the related algorithm and parameterizations in snow redistribution and wind-blown snow process， which need to further enhance the observation and simulation in the future. This study provides some clues for further improving the simulation ability of hydrological model in alpine cryospheric basins.