One long-term problem with EMOSLIB is that there is an inconsistency between interpolated values of the land-sea mask and sea-surface temperature. This can lead to:
- land-points (i.e., points where the interpolated LSM >= 0.5) with a sea-surface temperature
- sea-points (i.e., points where the interpolated LSM < 0.5) without a sea surface temperature
This is demonstrated in the figures below for the LSM and SST from the HRES O1280 grid interpolated to a 0.5x0.5 regular lat-lon grid. The figure on the left shows the results using EMOSLIB with the default (bilinear) interpolation method and that on the right those using a nearest neighbour method for both SST and LSM.
|
|
LSM and SST from HRES (O1280) both interpolated with EMOSLIB to a 0.5x0.5 regular lat-lon grid using the default interpolation method (bilinear) and land-sea mask processing
Globally, in total, there are:
| LSM and SST from HRES (O1280) both interpolated with EMOSLIB to a 0.5x0.5 regular lat-lon grid using the nearest neighbour and default land-sea mask processing
Globally, in total, there are:
|
Repeating the interpolation with MIR produces the following results, again with the figure on the left showing the results with the default (linear) interpolation method and that on the right those using a nearest neighbour method.
|
|
LSM and SST from HRES (O1280) both interpolated with MIR to a 0.5x0.5 regular lat-lon grid using the default interpolation method (linear) and no land-sea mask processing
Globally, in total, there are:
| LSM and SST from HRES (O1280) both interpolated with MIR to a 0.5x0.5 regular lat-lon grid using the nearest-neighbour method and no land-sea mask processing
Globally, in total, there are:
|
Results are confirmed using other input and output resolutions tested so far.
Using the nearest neighbour (or nearest LSM) interpolation method for both fields with MIR produces a sea-surface temperature field consistent with the land-sea mask.