General
- The different producing centres share many of the same characteristics in
the plots.
- ECMWF model best-fit pressures are generally very similar to Met Office.
- The CO2 slicing height assignment method generally shows better
statistics at high level than EBBT method.
- Meteosat-9 IR is worse below about 400 hPa in the extra-tropics where
there is a high height bias and increased standard deviation, but the
height biases for the WV channels appear to be slightly better than IR
at mid level. The zonal plots from ECMWF show more variation at mid
level, particularly over sea.
- GOES IR also shows a high height bias below about 400 hPa over sea, but
is slightly better over land. CO2 slicing appears to be used more for
GOES down to lower pressures than MSG and also tends to have slightly better
mid level statistics. The GOES plots from ECMWF are similar but again show
increased variation at mid level compared to the Met Office. ECMWF shows a
more pronounced low height bias over land in the tropics (about 300-400 hPa)
in the GOES cloudy WV plots compared to the Met Office.
- The WV intercept height assignment method also shows poorer statistics
at mid level.
- For Meteosat-9 IR WV intercept heights this is worse over sea than over
land with a high height bias and increased standard deviation at mid level,
but also in the extra tropics at high level. Meteosat-9 WV winds appear to
have less of a height bias at high level than the IR winds.
- GOES WV intercept heights show similar characteristics at mid level but appear
to have better statistics than MSG at high level. WV intercept is also lower
in the atmosphere than MSG.
- For MTSAT-1R the WV intercept method also shows worsening standard deviation at
mid level but unlike MSG and GOES has no high height bias. MTSAT-1R IR winds show
a more significant low height bias versus ECMWF at mid level.
- Low level winds are generally assigned lower in the atmosphere than the model
preferred location.
- The Meteosat-9 low level winds exhibit a pronounced low height bias that is
worse over land than over sea and is located in the tropics at around 10-30N.
This is a feature of the MSG winds that has been described in detail in previous
NWP SAF analysis reports (e.g. see feature 2.6 of 4th analysis report) and for
many cases it is likely to be a result of semi-transparent clouds being assigned
too low due to temperature contributions from below the cloud over the hot African
land surface. For the VIS winds the O-B best-fit pressure standard deviation tends
to increase with height. The low level height bias tends to be more pronounced in
the ECMWF plots meaning the best-fit pressure for the ECMWF model is higher in the
atmosphere than for the Met Office model. This could also be a result of the
different way in which multiple minima are handled.
- Met-7 IR low level statistics also show a low height bias but even more pronounced
than Met-9 and with higher levels of O-B best-fit pressure standard deviation than other
satellites and channels. Investigations in the NWP SAF analysis reports have pinpointed
anomalously fast winds assigned to low level that are associated with the sub-tropical
jet in the NH winter months as the likely cause of this (e.g. see feature 2.7 of the 4th
analysis report).
- MTSAT-1R low level IR winds similarly show a low height bias in the tropics. The zonal
plots show that the MTSAT winds are assigned to a much narrower band around 850-950 hPa
compared to the other producers.
- GOES low level VIS winds show a significant high height bias at around 600-800 hPa which
is most pronounced over sea. This has been linked to problems assigning heights in the
stratocumulus inversion regions of the Pacific and Atlantic (see feature 2.1 in the
3rd analysis report).
- MSG WV EBBT has low bias at mid level and the WV 7.3 winds have higher best-fit pressure
standard deviations than the WV6.2 channel.
- High level MSG WV winds assigned above 200 hPa tend to be assigned too high in the extra-tropics.
- There are relatively few MTSAT-1R IR winds assigned to high level using the EBBT method.
- MODIS pressure biases are generally quite small but the best-fit pressure variation increases
below about 600 hPa.
|
|
