HRES T1279
- Differences between EMOSLIB and MIR are due to
- Use by MIR of an intermediate full (regular) Gaussian grid at a resolution corresponding to the target lat-lon resolution for the spectral to grid transformation followed by grid-to-grid interpolation
- MIR uses a more accurate truncation of the spectral series to correspond target grid resolution to avoid ripples from aliasing effects
- For transformations to high resolution target grids for which no spectral truncation occurs
- differences are at most +/-3K
- differences are smaller than +/-1K for more than 99% of grid points
- For transformations to lower resolution target grids
- the maximum and minimum differences are significantly larger.
- differences are smaller than +/-2K for more than 99% of grid points
- larger differences are due to sub-optimal default truncation of the spectral series used by EMOSLIB:
- EMOSLIB truncation=T799 for 0.25x0.25 where theoretical ideal used by MIR is truncation=T719
- EMOSLIB truncation=T319 for 0.5x0.5 where theoretical ideal used by MIR is truncation=T359
- EMOSLIB truncation=T213 for 1.0x1.0 where theoretical ideal used by MIR is truncation=T179
- The use of the intermediate grid with MIR generally produces smoother fields.
Target resolution | Statistics of differences | Histogram of differences | Global | Europe |
---|---|---|---|---|
0.1x0.1 | Maximum difference = 1.693253 | -1.693253 : -1.354602 - 28 ( 0.000432%) | ||
0.125x0.125 | Maximum difference = 2.510147 | -2.510147 : -2.008118 - 22 ( 0.000530%) | ||
0.25x0.25 | Maximum difference = 8.004074 | -8.004074 : -6.403259 - 1 ( 0.000096%) | ||
0.5x0.5 | Maximum difference = 13.379105 | -13.379105 : -10.703284 - 0 ( 0.000000%) | ||
1.0x1.0 | Maximum difference = 9.762299 | -10.036530 : -8.029224 - 14 ( 0.021486%) |