The impact of targeted satellite observations on numerical weather prediction

In numerical weather prediction (NWP), large quantities of satellite data are discarded because of operational time constraints and the potential for high‐density observations to reduce the quality of the initial conditions (analysis). To retain the important observational data without significantly increasing the processing time, additional observations can be targeted in regions of the atmosphere influential on a particular verification region. In this study, Advanced TIROS Operational Vertical Sounder (ATOVS) observations were used to investigate the impact of targeted satellite data on forecast skill. The role of the temporal and spatial distribution of targeted ATOVS observations was investigated. Case‐studies were used to investigate the sensitivity of forecasts to the size of the target region and the length of time the region was targeted. The results indicate that, when targeting at the analysis time only, a target region between 1 × 106 and 2 × 106 km2 (30–60 extra observations) provides the most effective use of resources. When targeting a region for 24 hours, the impact on the forecast in the verification region was doubled (compared with the impact when targeting at the analysis time only), but targeting for longer periods than this provided little extra benefit. When targeting for more than one assimilation cycle, the size of the target region was increased accordingly to cover the migration of the sensitive region. Target regions of 6.5 × 106 km2 and larger (∼200 extra ATOVS at each assimilation cycle) were required to capture the majority of the information relevant to the forecast in the verification region.