|Title||The value of targeted observations - Part I: Data denial experiments for the Atlantic and the Pacific.|
|Year of Publication||2007|
|Authors||Kelly, GA, Thepaut, J-J, Buizza, R, Cardinali, C|
|Secondary Title||Technical Memorandum|
In the past, there has been a number of targeting experiments where extra observations were added in a small region to an existing operational system. Generally speaking, these experiments have shown that the impact of these 'targeted observations' is small, adding on average only a few hours of forecast skill. One of the weaknesses of these studies lies in that they did not provide any (reference) value of what the impact of removing all data over a larger region is, e.g. the ocean, and thus they were not able to interpret the impact of the targeted observations using these reference results. This work is the first of three companion papers that discuss the value of targeted observations: its focus is to examine the effect of removing all types of observations from most of the Pacific and Atlantic within the ECMWF operational system as of June 2005. In other words, this work discusses results from reference experiments necessary to properly interpret the impact of targeted observations. In this work, experiments using both 4D- and 3D-Var data assimilation systems designed to provide a set of baseline experiments necessary to interpret the value of targeting experiments within these oceanic basins are discussed. These experiments indicate that if a 4D-VAR data assimilation system is used, the impact of removing observations remains confined to the denial region and does not progress very far down stream as the forecast period increases. For example, the Pacific denial experiment (covering a total period of six months) shows a degradation of the 48 hour forecasts over North America, while the effect in the medium range over Europe remains very small. The Atlantic denial experiment shows a downstream short range forecast degradation over Europe, although the magnitude of the short range impact is clearly smaller than when denying observations over the Pacific Ocean.