Low-level wind error in the tropical East Pacific
During the Northern Hemisphere summer of 2001, the NCEP global operational model produced better forecasts of hurricanes than any other model; however, in standardized skill scores in the tropics it continued to lag other models, particularly at 850 hPa. The RMS vector error is especially large in the MRF at one day. This study examines a large time-mean error in the one-day forecasts of meridional wind (verified against its own analyses) in the tropical East Pacific south of the equator below 500 hPa. The error occurs during most, but not all forecasts, and has been present for at least a year. The error is large enough to contribute substantially to the tropical root-mean-square error in one-day forecasts of the meridional wind; eliminating the mean error statistically reduces the tropical mean RMS error by 10%.
The error occurs near two profiler stations, one in the Galapagos Islands at 1S, 90W and the other on the coast of Peru at 5S, 81W. The one in the Galapagos appears to measure only the lower atmosphere to 4 or 5 km; the other appears to measure the atmosphere to 15 km. The winds in the latter are blacklisted above 200 hPa. The profiler winds were previously blacklisted. The error in the one-day forecasts is associated with changes in the divergent circulation in the eastern portions of the Inter-Tropical Convergence Zone. The error persists in the two-day and three-day forecasts, but does not grow. It becomes more rotational with forecast length. One hypothesis is that the error is caused by either inaccurate profiler data or an incompatibility between the model and the profiler data. The error appears to be present in the parallel x test of a higher resolution model; removing the profiler data from the parallel x would test whether the error is associated with the profilers.
R. Kistler and R. Treadon did remove the profiler data from the parallel x for a few weeks; the results show a dramatic decrease in error in the immediate vicinity of the Galapagos profiler. However, the larger east Pacific error remains.
Pressure-longitude cross-section of v at 2.5S
Winds at 800 hPa
The model's clouds in the region appear to have substantial differences from independent estimates, reflecting difficulties in the modeling of low-level stratus. Whether these problems in cloudiness force the errors described above or are partly forced by them is not obvious.
Longitude-pressure cross-sections of the time-mean meridional wind at the equator
Analyses and 12 hr forecasts valid at 1200UT (East Pacific)
6hr and 12 hr forecast errors valid at 1200UT (East Pacific)
Longitude-pressure cross-sections of the time-mean meridional wind at 2.5S (East Pacific)
Latitude-pressure cross-sections of the time-mean meridional wind at 95W 30S-30N
Time-mean meridional wind at 850 hPa
Time series of meridional wind at 2.5S
RMS error in meridional wind 20S-20N
1 day forecasts
3 day forecasts
Time-mean wind in the East Pacific
Time-mean 750 hPa stream function in East Pacific
Time-mean errors in absolute vorticity
Time-mean horizontal divergence
Time-mean vertical velocity
Surface short wave radiation