青藏高原上湖泊众多， 其中大多数被季节性湖冰覆盖。湖冰对气候变化响应敏感， 其生消过程会显著改变湖-气交换通量。而现有的高原湖冰长时间观测数据较少， 需要利用湖冰再分析资料进行研究。但目前对ERA5-Land湖冰资料在高原的适用性及改进方法还不甚清楚。因此， 本文首先利用2010 -2022年青海湖和鄂陵湖的湖冰观测数据， 评估了ERA5-Land再分析资料对青藏高原典型湖泊湖冰特征的模拟能力。结果表明： ERA5-Land资料对青海湖和鄂陵湖的冰厚平均高估0.54~0.62 m， 对封冻期天数高估约68 d·a^-1。其次对再分析湖冰数据误差来源进行分析， 通过对比ERA5-Land和鄂陵湖观测资料及其各自驱动模式的模拟结果， 得出误差主要来源于输出ERA5-Land湖冰数据的FLake一维湖泊模式。最后基于2010 -2022年青海湖和鄂陵湖的MCD43A3地表反照率产品， 利用其多年平均反照率和动态日均反照率改进了FLake模型， 模拟冰厚平均偏差可分别减小85%和90%， 封冻期天数的模拟偏差减小了约6 d·a^-1和8 d·a^-1。两种方法均可以改进FLake模型对湖冰的模拟效果， 特别是对积雪覆盖时间较长的湖泊， 动态反照率方法改进效果更明显。本研究揭示了ERA5-Land湖冰特征的主要误差来源为FLake模式中的湖冰反照率， 并对该参数进行了改进， 提高了模型对湖冰的模拟效果， 可为提高ERA5-Land再分析湖冰数据在青藏高原典型湖泊的模拟精度提供参考。

The Qinghai-Xizang Plateau， distinguished by its vast array of lakes， exhibits marked seasonal lake ice coverage， which is highly responsive to climatic shifts.This ice coverage plays a crucial role in the dynamic interchange of fluxes between the lake surfaces and the atmosphere.Despite the significance of these ice phenomena， the limited availability of extensive， long-term observational data on plateau lake ice has led to a reliance on reanalyzed ice datasets， particularly ERA5-Land.This study aims to rigorously evaluate the effectiveness and potential enhancements of ERA5-Land's lake ice data in the distinct environment of the Qinghai-Xizang Plateau.Focusing on data collected from 2010 to 2022 for Qinghai Lake and Ngoring Lake， this research meticulously examines the ERA5-Land reanalysis data's ability to accurately capture the intrinsic characteristics of plateau lake ice.The study uncovered that ERA5-Land tends to overestimate the ice thickness by about 0.54~0.62 m and erroneously prolongs the freezing period by roughly 68 days per year for these lakes.This notable discrepancy necessitated an in-depth error analysis， which synthesized ERA5-Land data with direct observational data from Ngoring Lake， revealing that inaccuracies primarily originated from the FLake one-dimensional lake model within the ERA5-Land system.In an effort to address these inaccuracies， the study employed the MCD43A3 surface albedo product for both Qinghai Lake and Ngoring Lake over the same period.This innovative approach significantly refined the FLake model by incorporating both a multi-year average albedo and a dynamic daily average albedo.These methodological improvements led to a substantial reduction in the average bias of ice thickness， by 85% and 90% respectively， and narrowed the deviation in the modeled freezing period by about 6 and 8 days per year.The enhancements were particularly notable in lakes with longer periods of snow cover， where the dynamic albedo adjustment proved to be highly effective.This research has successfully identified the albedo parameter within the FLake model as a key source of error in ERA5-Land's lake ice characterizations and has implemented practical adjustments to rectify this.These enhancements have markedly increased the model's precision in simulating lake ice， thereby significantly improving the accuracy of ERA5-Land reanalyzed lake ice data.This advancement is particularly pertinent for the unique climatic and geographical conditions of Qinghai Lake and Ngoring Lake on the Qinghai-Xizang Plateau and offers invaluable insights for future research and practical applications in this domain.The findings of this study contribute profoundly to our understanding and modeling of lake ice phenomena in high-altitude regions and have broader implications for climatological research and environmental monitoring on the Qinghai-Xizang Plateau.