peer reviewed by: Paul Berrisford, Gianpaolo Balsamo
date of peer review:
I retrieved the land-sea mask from ERA5 or ERA-Interim on its archived grid (grid=av), or on a reduced Gaussian grid. The data has fractional values in the range 0 (sea) to 1 (land).
I also retrieved the same land-sea mask on a regular lat/lon grid (e.g. grid=2.0/2.0). This data is binary, with values 0 and 1.
Why is this?
The short answer
When you retrieve data in any other than the archived grid, the data is on the fly interpolated to your chosen grid, using ECMWF's interpolation tool EMOSLIB. EMOSLIB by design interpolates a land sea-mask to a reduced Gaussian grid as fractional values, but to regular lat/lon as binary values.
From early 2019 the new interpolation tool MIR will be used by default, which by default interpolates the land-sea mask to fractional values, including for regular lat/lon grids.
The long answer
In the data archive at ECMWF, the land-sea mask is stored on a reduced Gaussian grid (N320 for ERA5 and N128 for ERA-Interim). For both datasets (ERA5 and ERA-Interim) the data values of the land-sea mask are fractional values in the range 0 (sea) to 1 (land).
When you retrieve land-sea mask data from the ECMWF data archive:
|your grid specification||result|
|regular lat/lon grid (e.g. grid=2.0/2.0):||the land-sea mask in interpolated by EMOSLIB on the fly from the original Gaussian grid to regular lat/lon, and the output values are rounded to 0 or 1|
|Gaussian grid (e.g. grid=N256)|
the land-sea mask is interpolated by EMOSLIB on the fly from the original Gaussian grid to the specified one, without rounding.
|archive (original) grid (grid=av)||no interpolation takes place|
Related articles appear here based on the labels you select. Click to edit the macro and add or change labels.