Russ Fiedler
未读,2017年1月23日 01:43:362017/1/23登录即可回复作者
登录即可转发
您无权在此群组中删除帖子
请登录以举报消息
要么此群组的电子邮件地址为匿名状态,要么您得查看成员电子邮件地址权限才能查看原始帖子
收件人 mom-...@googlegroups.com
Hi,
I'd just like to alert the community that there are some very strange
choices for the valid ranges of common variables. These choices
determine the values (scale_factor and add_offset) used for packing
variables into 16 bit integers and can result in massive degradation of
precision.
Unfortunately, some of these choices changed from mom4p1 to mom5 and we
got caught using some old diagnostic tables. In particular, various sea
surface height variables had their ranges changed from +/-10m to
+/-1000m. This resulted in an effective resolution of 3cm and severely
compromised what we could do with the output. Ouch! Also mixed layer
depths have a range from 0-1000km resulting in a resolution of about 15m!
In short, if you are considering the use of the packing option in mom5,
be very careful and check the default ranges in the code since it is
rather likely that they haven't been sanity checked.
I guess it would be nice if the valid range could be overridden by the
diagnostic table but that may be difficult.
Russ