关于ERA5-Land数据的几个问题＜笔记自用＞

1. 关于通量方向

The ECMWF Integrated Forecasting System convention is that downward fluxes are positive. Therefore, negative values indicate evaporation and positive values indicate condensation"

ECMWF规定以向下的方向为正值（positive），向上为负值（negative），因此对于蒸散发而言，正值表示水汽凝结，负值才是蒸散发。

2. 关于scale_factor和add_offset问题

2.1 由于考虑到数据存储时占用空间的问题，ERA5-Land会采用scale_factor和add_offset来帮助减少文件大小，以下面变量为例

实际蒸散发的计算公式为
$$ET=E{T}_{read}\ast Scal{e}_{factor}+ad{d}_{offset}$$

2.2 由于下载的方式可以不同，scale_factor 和 add_offset可能会有变化。
下面的对话可以反映出差异

提问

Hello,
I downloaded snowfall data from 1979 to 2020, once for all months and once for every month separately, as netcdf files; i.e. I have one .nc file containing snowfall data from Jan to Dec for 1979 to 2020 (all_months.nc), and then I have 12 separated .nc files each containing snowfall data of a month from 1979 to 2020 (jan.nc, feb.nc, …, dec.nc).

I have noticed that the scale_factor and add_offset for snowfall data change, such that the unpacked values are slightly different despite looking at the same month (e.g. Values for October from all_month.nc are slightly different from those from oct.nc). I am calculating the yearly trends in snowfall for each month, and in some cases the trend change from having a significant trend to not having one as the p-value changes (even if the values of the slope is close).

What would be the reason for such differences? Sorry if this is a simple question, and thank you in advance.

---

回答
Hi

The scale_factor and add_offset are adjusted for each request to give an output netCDF file of a similar size to the input GRIB data size. The actual values sldo depend on the data range in the whole grib file, so I would expect some variation between different requests. How large are the differences you see?

Thanks,
Kevin

注意：如果使用matlab的 ncread函数 读取，则数据会被自动修正，无需再考虑上述过程！

请关注 大余海深 公众号，我们一起成长！