Analysis, Lateral Boundary, and Observation Impacts in a Limited Area Model

Impacts from analysis and lateral boundary updates as well as the assimilation of observations are investigated using the forecast sensitivty to observation impact framework in a limited-area atmospheric model. High temporal frequency estimates of forecast error are produced using aircraft observations for validation. Using these estimates, forecast error reduction between background and analysis trajectories is shown to decrease through the first 24 h of forecast time. The increasing importance of lateral boundary updates in decreasing forecast error with forecast lead time is presented. However, the ability of the adjoint forecast model to attribute forecast error reduction to analysis and lateral boundary updates decreases as forecast length increases. The relative distributions of the largest observation impacts for different lead times are similar. This means that impacts for shorter forecast lengths are a good proxy for impacts on longer forecasts, thereby mitigating some of the problems in long adjoint model integrations. Finally, a metric that measures forecast error against radiosondes is introduced and produces different distributions of observation impact importance.