Monday, 10 February 2003: 11:00 AM
Key-Analysis Errors of Specific Humidity
It is well known that uncertainties in the initial conditions (analysis) are a major
source of forecast errors. In this study we utilize this fact in order to determine
so-called key-analysis errors of specific humidity for the Northern Hemisphere. For
this purpose we use the nonlinear ECMWF model along with a tangent linear version
and its adjoint. It is shown that
the largest humidity key-analysis errors are found in the lower atmosphere near 700 hPa.
Their horizontal distribution is relatively homogeneous over the Northern Hemisphere.
Further, the impact of these humidity errors on subsequent forecasts is discussed. It
is shown that their influence on Z500 forecast errors is relatively small in the
very short-range, but it becomes an important part of the forecast error in the
medium-range. Finally, the linearity of the growth of humidity key-analysis errors
through the forecast range is discussed. We present evidence that the evolved humidity
key-analysis errors show notable nonlinear signatures already after a few hours.
Implications for data asssimilation and probabilistic ensemble forecasting are
discussed.
Supplementary URL: