Predicting the influence of observations on medium‐range forecasts of atmospheric flow

Abstract

ABSTRACT:In recent years, the Ensemble Transform Kalman Filter (ETKF) has been demonstrated to be useful for identifying a priori dynamically important locations for the placement of supplementary dropwindsonde observations aimed at improving short-range (1-3 day) forecasts of high-impact winter weather. In this paper, the ability of this strategy to predict the influence (or 'signal') of assimilating observations into the NCEP Global Forecast System for forecasts of 200 hPa wind up to 6 days is evaluated. Using a 50-member ECMWF ensemble, the ETKF was found to exhibit significantly higher skill than a seasonal 'climatology' of the important locations in predicting both (1) the spatial structure of signals within the storm track on a case-by-case basis; and (2) the variance of these signals over a 2-month period, within objectively chosen verification regions on synoptic scales. The verification region was selected by extracting the zonal envelope of the Rossby wave packet associated with the propagating ETKF signal variance. It is recommended that larger verification regions be used for longer lead times, due to the eastward expansion of the wave packet.The capability of the ETKF to predict signal variance out to 6 days was found to be dependent on the flow regime. The ETKF was most capable when the background flow was predominantly zonal, and least capable in instances where the observations were placed upstream of a blocking high over the north-eastern Pacific. Therefore, the ETKF is sometimes (but not always) able to predict when there exists significant potential for a particular group of observations to improve medium-range forecasts.

Characteristics of Ensemble Transform Kalman Filter adaptive sampling guidance for tropical cyclones

“…The results of Sellwood et al (2008) and , which demonstrated that ETKF targets in the midlatitude storm track could be traced upstream out to 7 d in zonal flow regimes, are unlikely to be corroborated for TCs except in cases of a distinct and unambiguous interaction with an approaching midlatitude trough. This higher ambiguity in ETKF targets for TCs heightens the need for careful interpretation by field programme planners.…”

“…In the future, the examination of the un-normalized value of signal variance would be useful, in order to quantitatively estimate day-to-day changes in the potential for forecast-error reduction. The ETKF has shown promise in this regard in the midlatitudes (Majumdar et al, , 2002Sellwood et al, 2008). For tropical cyclones, it is not yet clear to the authors how to interpret quantitatively the reduction in forecast-error variance, given that it is largely a function of both the TC wind field and its position.…”

“…Given that the error statistics in the ETKF rely on linear dynamics, the theoretical assumptions may be violated as the forecast time is extended beyond 2 d. Nevertheless, as demonstrated by Sellwood et al (2008) and , the ETKF guidance may be plausible out to seven days in the midlatitude storm track. No quantitative evaluations have yet been performed for TCs, and will be left for future research.…”

“…A relatively new method, the Adjoint-Derived Sensitivity Steering Vector (ADSSV; Wu et al, 2007b), seeks sensitive areas in which initial perturbations would modify the steering flow. Another technique that has proven to be effective for short-range (Majumdar et al, 2002) and medium-range (Sellwood et al, 2008) forecasts of winter weather is the Ensemble Transform Kalman Filter (ETKF), which has also recently been extended for use with TCs. While all these strategies are based on reducing error in a forecast metric, detailed intercomparison studies have shown that they disagree frequently, for TCs in the Atlantic basin (Majumdar et al, 2006;Reynolds et al, 2007) and in the western North Pacific basin (Wu et al, 2009a).…”

Predictability of an Atmospheric Blocking Event that Occurred on 15 December 2005

“…In recent years, many Observing System Experiments (OSEs) have been conducted throughout the world to investigate the impacts of observations on analysis and forecast errors in downstream areas (e.g., Langland et al 1999;Szunyogh et al 2002;Langland 2005;Fourrie et al 2006;Wu et al 2007;Chou and Wu 2008;Sellwood et al 2008;Inoue et al 2009;Yamaguchi et al 2009). In addition, TIGGE has enabled us to obtain operational medium-range ensemble forecast data quasi-operationally (;2 day behind).…”