IMPACT OF CLOUDY SCENE SATELLITE OBSERVATIONS ON A

IMPACT OF CLOUDY SCENE SATELLITE OBSERVATIONS ON A CLOUD-FREE INITIAL MODEL STATE Curtis J. Seaman, Manajit Sengupta, Steven J. Fletcher, Scott Longmore and Thomas H. Vonder Haar Do. D Center for Geosciences / Atmospheric Research, Cooperative Institute for Research in the Atmosphere (CIRA) Colorado State University, Fort Collins, CO 80523 -1375 Introduction Experiment Results Figure 2: Models typically underestimate mid-level cloud occurrence as this figure shows Objectives: Improve cloud forecasting by assimilating IR radiances in cloudy scenes using 4 D-VAR Investigate assimilation of cloudy radiances into a cloud-free initial model state Compare results with observations from the 2 November 2001 altocumulus case from CLEX-9 Corresponding author email: seaman@cira. colostate. edu No Assimilation Cloud Top Height Relative Humidity Vertical Motion Surface Winds Imager 3 & 4 Figure 1: Operationally, most forecast centers only assimilate clear-sky radiances, ignoring most of the globe Sounding Surface Temperature Sounder 7 & 11 Relevance: Clouds cover more than half of the Earth’s surface (e. g. Rossow and Duenas 2004) Most operational forecast centers assimilate only clear-sky infrared temperature and water vapor information (if at all) Cloudy scene radiances have been assimilated successfully in cases where the initial model state contained clouds (Vukicevic et al. 2006) What happens if no cloud exists in the initial model state? Brightness Temperature Matching Decorrelation Lengths RAMDAS uses assumed values for decorrelation length (right). Doubling and halving these values has a significant impact on the results (below). Figure 3 (below). Comparison between model-simulated and observed brightness temperatures for the GOES Imager experiment. A – B) Modeled brightness temperatures for channels 3 and 4 before assimilation. C – D) Modeled brightness temperatures for channels 3 and 4 after assimilation. E - F) Observed brightness temperatures from GOES Imager channels 3 and 4. Conclusions Control Variable: rl value: pressure (pert. Exner function) 150 km ice-liquid potential temperature 100 km u, v winds 150 km ice water mixing ratio 50 km snow water mixing ratio 50 km total water mixing ratio 50 km GOES Imager experiment: Cooled the surface Increased upper-tropospheric humidity Produced fog Not close to producing mid-level cloud GOES Sounder experiment: Produced subsidence inversion at 2 km AGL Increased mid- and upper- tropospheric humidity in center of domain Set up donut-shaped wind field in lower troposphere Closer to producing mid-level cloud Decorrelation lengths: Increasing decorrelation length increases impact of observations Figure 4 (above). Comparison between model-simulated and observed brightness temperatures for the GOES Sounder experiment. A – B) Modeled brightness temperatures for channels 7 and 11 before assimilation. C – D) Modeled brightness temperatures for channels 7 and 11 after assimilation. E - F) Observed brightness temperatures from GOES Sounder channels 7 and 11. Figure 5. Soundings from the Imager experiment with values of rl halved (A) and doubled (B). Soundings from the Sounder experiment with values of rl halved (C ) and doubled (D). Dashed lines indicate the soundings using the default values of rl. Overall conclusions: RAMDAS minimizes cost function by any means necessary Model state is only modified where observations are sensitive to model variables If there is no cloud in the initial model state, the system makes adjustments based on the assumption there is no cloud Addition of physical constraints is necessary to achieve proper cloud