SUMMARYWe study 20 cases in which 60-hour Numerical Weather Prediction (NWP) of specific synoptic events have been markedly improved by observations made during the final 15 hours prior to the forecast initial time. For each case, the observing systems which played major r6les in improving the accuracy of the forecast mean-sea-level pressure (m.s.1.p.) field are identified using a series of Observation System Experiments. The relative utility of each observation type is then assessed according to the frequency, over all cases, with which it delivers a major beneficial impact.The method of multiple case-study employed has the advantage that observation value is assessed according to the benefit delivered to forecasts of events (e.g. cyclogenesis) which are of prime importance to operational meteorology. The method, therefore, provides a useful complement to the conventional statistical approach which typically yields an 'average' benefit, calculated for all regimes (both active and 'quiet') occurring over a given period within a specific geographical region. The cases are selected from the most significant data-impact events detected in routine operational forecasts of m.s.l.p., for Europe and North America, during the period September 1993 to December 1995.Results indicate that, for weather systems developing over North America, radiosonde and aircraft reports contribute major forecast benefits most often, confirming their key r6le in the North American network. These two observing systems contribute with similar frequency, suggesting that the effectiveness of the less abundant radiosonde reports is boosted significantly by their profile format. Surface data and cloud-track winds are the next most frequent contributors.For weather systems developing over the North Pacific and North Atlantic, aircraft winds contribute forecast benefits most frequently, and by a wide margin. Conventional surface data also play a key r6le.Comparison of the utility of wind and temperature data suggests that, on average, the benefit of wind profiles is somewhat greater than that of temperature profiles, and that, in levels above -400 hPa, wind data are considerably more valuable than temperature data.The results give insights which can be used to guide the rationalization of existing networks within the northern hemisphere mid latitudes, and the following recommendations are made with a view to improving NWP over the European area.0 Acquire more aircraft data over North Atlantic routes and over Europe (including reports during climb and descent). However, caution must be exercised if aircraft data are used to replace profile information from radiosondes.0 Deploy more surface observations (e.g. drifting buoys) and shipborne radiosonde ascents over the North Atlantic.0 Make more use of pattern-tracking techniques for deriving 'clear-air' wind data from satellite water-vapour imagery.Continue investigation of the effectiveness of observations 'targeted' on objectively defined regions where model errors are predicted to grow most rapidly.

The aim of data‐impact studies at the UK Met. Office is to investigate how observations affect the accuracy of model forecasts. Results from such experiments provide useful evidence on which to base the design of observational networks. This project, using a case study approach, investigated the relative benefit of different observation types within The Met. Office's Mesoscale Model domain on forecasts of three‐hourly precipitation accumulation over the UK up to 12 hours ahead. The method used assesses the impact of assimilating single observation types, or a limited combination of types, where impact is measured against a control forecast obtained after a dummy assimilation using no observations. In experiments for 13 case studies, the observation types that most frequently provided a beneficial impact when presented alone to the assimilation were sonde data, surface data and data from the Moisture Observation Processing System (MOPS). Copyright © 2000 Crown copyright