Superobbing and Thinning Scales for All-Sky Humidity Sounder Assimilation

Abstract

Listen

Translate article

Abstract Humidity sounder radiances are currently thinned to 110-km spacing prior to assimilation at ECMWF and used with no averaging applied. In this paper, the thinning scale and possible averaging of all-sky humidity sounder observations into “superobs” are considered. The short- and medium-range forecast impacts of changing the thinning and averaging scales of humidity sounder radiances prior to the data assimilation are investigated separately and then together. Superobbing acts as a low-pass filter and provides smoother images of departures, decreasing the effective sensor noise and thus the standard deviation of background departures, marginally for 183-GHz channels (5%–15%) and significantly for 118-GHz channels (5%–55%). Observations are thus more representative of the model effective resolution, with a better utilization of total information content than thinning native-resolution radiances, as less information is discarded. Whether changed in isolation or combination, the additions of data via superobbing and finer thinning are both shown to markedly improve background fits to independent observations, indicative of better short-range forecasts of humidity and improved winds via the 4D-Var tracer effect. Wind forecasts in the Southern Hemisphere are slightly improved in the medium range. A final configuration is tested at the resolution of the current operational model, with humidity sounder radiances averaged into 50-km superobs with 70-km spacing. This provides about 140% more radiances for assimilation and more finely detailed maps to analyze mesoscale features. The forecast impact of this change is larger in testing with higher model and data assimilation resolutions, showing the scale dependence of such decisions and the expected performance in an operational configuration. Significance Statement This paper investigates thinning and averaging scales for humidity-sounding microwave observations in the ECMWF data assimilation system. The introduction of spatial averaging shows a positive impact, as does the assimilation of observations with finer spacing. These changes permit more total information on humidity into the system, and both are beneficial for short-range forecasts of humidity and winds in the mid- to upper troposphere. The results highlight the interplay between spatial scales of observations and those of the analysis system, with possibilities for improved utilization in this particular case. This is expected to remain a key consideration in assimilation systems going forward, given the continued increases in the resolution of assimilation systems and forecast models.