Data denial experiments for extratropical transition

Extratropical transition (ET) of tropical cyclones presents a significant challenge to numerical weather forecasts. Small uncertainties in sensitive regions around ET events can lead to large forecast errors. An important source of such uncertainties lies in the insufficient data coverage. Introducing targeted observations into the data assimilation in the vicinity of an ET event or in regions sensitive to fast error growth may have a notable value for a numerical forecast.

An upper bound for the value of additional targeted observations can be gained by denying data and measure the resulting impact on the forecast without these data. This is done under the assumption that the impact of removing information is symmetric to the impact of adding information in the same region.

Data denial experiments using the ECMWF IFS are designed to investigate the value of targeted observations for historical ET cases over the Atlantic. The impact of removing data in specified locations on the forecast for the ET events is examined. Data is denied in sensitive areas determined with singular vectors (SVs) optimized over a 2 day interval for an optimization region over Europe and the deterministic forecast is rerun for 5 days. Furthermore, the influence of observations in the vicinity of the decaying tropical cyclone (TC) and associated with structures inherent to a TC on the forecast is examined by denying data in a rectangular box around the ET. For one case, Hurricane Helene (2006), targeted SVs were calculated and used as data denial regions for new forecasts. The degradation of the forecasts in the different experiments is investigated to consider which region might be most valuable for additional observations.