Begin Main Content

Yanqiu Zhu and Ronald Gelaro

Computer forecast models used in weather and climate prediction incorporate information from millions of atmospheric observations to provide an accurate estimate of the current state of the atmosphere.  This snap-shot of the current state of the atmosphere is used as the starting point of the computer model forecast.  Generally speaking, the greater the number and accuracy of the input observations, the more accurate will be the estimate of the current atmospheric state, as well as the computer forecast generated from it.  However, because atmospheric observations, like all measurements, have some degree of error, some observations are more helpful to the forecast than others, and some observations may even make the forecast worse.  As the number and types of available observations continues to increase dramatically, especially from new satellites, it becomes increasingly important to develop efficient tools for estimating the impact of these observations on our weather and climate forecasts, and to improve their usage.

Data assimilation is the process by which information from observations---which tends to be irregular in space and time---is made available to a forecast model which requires this information in a regular or “gridded” form.  This process is far more complex than a simple interpolation of information from one point to another, as it depends on aspects such as the errors of the various observation types, imperfections in the computer forecast model and the physical relationships between different atmospheric variables (e.g., pressure and wind speed).  As a result, it is not straightforward to track the impact of a given observation, or subset of observations, on the resulting atmospheric analysis or forecast.

In this study, we present an efficient technique for assessing the impact of observations on atmospheric analyses and forecasts.  The technique is applied to the Goddard Earth Observing System (GEOS-5) atmospheric data assimilation system developed recently at the NASA Global Modeling and Assimilation Office (GMAO).  The technique uses a new formulation of the GEOS-5 data assimilation system, referred to as an adjoint formulation, to produce not only the usual estimate of the current atmospheric state, but also a measure of how sensitive this estimate is to any or all input observations used to produce it.  The sensitivity to, or impact of, any single observation or group of observations, whether from a single balloon measurement of wind at a particular height in the atmosphere, or a large swath of temperature measurements from an orbiting satellite, can be estimated simultaneously with a single execution of the adjoint system.  It is found, for example, that satellite observations have the largest impact on estimates of atmospheric temperature over the North Pacific, while observations from balloons and commercial aircraft strongly influence the estimated wind field over the continental United States.  This tool is being used extensively by the GMAO to understand and improve the way observations are used in weather and climate forecasts.

Figure caption: Limited-area view of (a) the squared analysis increments of temperature at approximately 550 hPa and (b) the sensitivity of JTNP  with respect to channel 5 brightness temperatures on NOAA-16 AMSU-A for 5 August 00Z. The contour interval is 0.2 K2 in (a) and the units are K in (b). The box outlines the area in which JTNP is defined.

Reference

Zhu Y. and  R. Gelaro, 2007:   Observation sensitivity calculations using the adjoint of the Gridpoint Statistical Interpolation (GSI) analysis system. Mon. Wea. Rev. (submitted).