Evaluating global snow water equivalent products for testing land surface models
Evaluating global snow water equivalent products for testing land surface models
复制标题
DOI：
10.1016/j.rse.2012.10.004
复制
发表时间：
2013-01
影响因子：
13.5
通讯作者：
S. Hancock;R. Baxter;J. Evans;B. Huntley
中科院分区：
文献类型：
--
作者：
S. Hancock;R. Baxter;J. Evans;B. Huntley
本文比较了三种全球雪水当量（SWE）产品SSM/I（NSIDC）、AMSR-E（NSIDC）和Globsnow（v1.0，ESA）、积雪覆盖面积（SCA）、积雪深度地面测量和气象数据，试图确定哪种产品最适合于陆面模式的测试和开发。特别注意的是，它给出了最准确的峰值积累，季节性SWE变化和第一次和最后一次积雪的日期。SSM/I和AMSR-E是纯粹的地球观测（EO）衍生产品，而Globsnow是EO和地面数据的组合。结果表明，纯EO产品可以饱和在较深的雪（SWE>80- 150毫米），可以显示虚假的功能在融化，并可能高估SWE由于强大的热梯度和错误的森林覆盖校正因子。沿着与地面数据（仅一个点）的比较，这表明Globsnow是确定峰值累积和季节性SWE循环的更准确的产品。雪的开始和结束日期的三个SWE产品进行了比较，以光学派生SCA（MOD 10 C1，作为真理），并发现雪的开始日期给出了很大的误差（均方根误差为20+天，虽然SSM/I是正确的平均）。降雪结束日期的误差较小（偏差为1- 6天），尽管差异仍在三周左右。研究期间，偶尔观察到Globsnow的突然变化（在v1.0和v1.2每日一次和v1.2每周一次产品中）。这些只发生在大约1%的病例中，似乎是假的。如果使用Globsnow验证陆面模式或同化方案，应注意避免这些跳跃。
This paper compares three global snow water equivalent (SWE) products, SSM/I (NSIDC), AMSR-E (NSIDC) and Globsnow (v1.0, ESA) to each other, snow covered area (SCA), ground measures of snow depth and meteorological data in an attempt to determine which might be most suitable for testing and developing land surface models. Particular attention is paid to which gives the most accurate peak accumulation, seasonal SWE changes and first and last dates of snow cover. SSM/I and AMSR-E are pure earth observation (EO) derived products whilst Globsnow is a combination of EO and ground data. The results suggest that the pure EO products can saturate in deeper snow (SWE>80–150mm), can show spurious features during melt and can overestimate SWE due to strong thermal gradients and erroneous forest cover correction factors. Along with the comparison to ground data (only a single point) this suggests that Globsnow is the more accurate product for determining peak accumulation and seasonal SWE cycle. The snow start and end dates of the three SWE products were compared to an optically derived SCA (MOD10C1, taken as truth) and found to give large errors of snow start date (root mean square error of 20+ days, though SSM/I was correct on average). The snow end dates had lower errors (a bias of 1–6days) although the spread was still on the order of three weeks. During the investigation, occasional abrupt changes in Globsnow were observed (in the v1.0 and v1.2 daily and v1.2 weekly products). These only occurred in around 1% of cases examined and seem to be spurious. Care should be taken to corrector avoid these jumps if using Globsnow to validate land surface models or in an assimilation scheme.