Ingesting Satellite Measurements of Cloud Cover into the PSU/NCAR Mesoscale Model (MM5) and Its Impacts on Modeled Surface Precipitation

I. Yucel, University of Arizona, yucel@hwr.arizona.edu

W. J. Shuttleworth, University of Arizona, shuttle@hwr.arizona.edu

X. Gao, University of Arizona, gao@hwr.arizona.edu

S. Sorooshian, University of Arizona, soroosh@hwr.arizona.edu

 

  1. Introduction
  2. This study investigates the extent to which assimilating high-resolution remotely sensed cloud cover into the MM5 provides an improved regional diagnosis of surface radiation fluxes and precipitation. Cloud cover is a key parameter linking and controlling these terms and the specification of cloud cover in atmospheric models is among the largest sources of uncertainty. This uncertainty is exacerbated by the great disparity between the spatial scale of clouds and model grid scale. Mesoscale atmospheric models do predict clouds, but model estimate of their spatial distribution and radiative characteristics may have considerable error. Manabe et al. (1991) pointed out that cloud-related weakness in modeled surface radiation produces markedly unrealistic surface fluxes of energy and water, and weakness in simulating cloud properties is currently a major problem in numerical weather prediction models. On the other hand, measurements of the extent and optical properties of clouds using Earth satellites have the potential to supplement surface observations and to improve weather simulation by models. The present research is focused on correcting the modeled cloud cover in MM5, thus allowing the MM5 radiation and precipitation codes to calculate the actual spatial variability generated by the observed clouds. The technique presented in this research was previously applied to Regional Atmospheric Modeling System (RAMS, Version 3b) and the results were impressive (Yucel et al., 2001). There was clearly value in extending the capability to assimilate cloud cover data to other meteorological models and, in particular to the MM5 which is becoming a community standard model. The potential advantages of this approach are that, not only is the surface radiation fields better estimated, but the model may also be better able to diagnose the true spatial pattern of precipitation. It may also enhance the ability of such mesoscale modeling systems to make accurate short-term forecasts of precipitation.

  3. Cloud Assimilation Method
  4. In this study, we opted to directly replace the modeled cloud-cover field with that derived from satellite observations. This reflects the fact that errors in the horizontal position of clouds in the model calculated cloud fields are likely to be much greater than those in the satellite-observed fields. An automatic procedure was developed to derive high-resolution (4km x 4km) fields of fractional cloud cover from visible band, (GOES series) geostationary satellite data using a novel tracking procedure to determine the clear-sky composite image. It was recognized that the University of Maryland (UMD) GEWEX/SRB model provides an important relationship among cloud albedo, cloud optical depth and cloud mass. Utilizing this relationship, our research was focused on exploring the feasibility of directly assimilating cloud cover into atmospheric model by converting the measured fractional cloud cover (via cloud optical depth) into vertically integrated cloud water/ice. However, the procedure is complicated by the need realistically to distribute the cloud water/ice within the vertical profile for each modeled grid square. It was assumed that, at each modeled grid square across the modeled domain, MM5 calculates the vertical position of cloud correctly, even if it miscalculates total cloud amount in that atmospheric column. We therefore create and assimilate a cloud field whose horizontal distribution is determined using an image derived from GOES, but whose vertical distribution is that of the cloud field as calculated by MM5 in the time step immediately prior to each cloud assimilation. GOES-derived cloud fields were ingested every minute using linear interpolation between two 15-minute GOES images.

  5. Results
  6. Figure 1a shows comparison between the MM5-derived with and without cloud assimilation and observed incoming surface solar radiation with averaged hourly intervals on July 14-15, 1999 at the AZMET sites in Southern Arizona. It is apparent that the radiation physics used in the MM5 model did not capture correctly the diurnal variations of surface solar radiation caused by the overlying clouds when there is no satellite observations at each AZMET sites. However, this feature was substantially altered and improved by the cloud assimilation over the observational arrays where warm season daytime convection is common. MM5 was able to do superior job in predicting more accurate radiation fields at the surface along simulation periods and along spatial points in the horizontal model domain. The improvements are also provided in scatter plots (Figure 1b) with and without cloud assimilation. In each case, the correlation coefficient, root mean square error, and mean bias are given in the figure. In cloud sky conditions, the scatter is substantially less with ingestion of cloud cover, and statistical values are improved accordingly. Figure 2 shows a comparison between the precipitation observed at the AZMET sites and that calculated by MM5 with and without cloud assimilation for July 14-15, 1999. MM5 consistently underestimates surface precipitation at the AZMET sites when no satellite observations of clouds are ingested. However, the timing and to some extent magnitude of modeled precipitation is better captured with cloud assimilation. The storms simulated by MM5 are approximately in the right place and at the right time because the extent of cloud cover is more realistic during the day. Figure 3 shows a comparison between the forecast fractional cloud cover across a 4 km x 4 km modeled domain and the observed cloud fraction derived from the GOES satellite image for a 4.5-hour forecast period. Every 3 columns in Figure 3 indicate the observed cloud cover derived from GOES images, the forecasted cloud cover obtained by following 3 hours of 1-minute cloud cover ingestion, and the forecasted cloud cover obtained from the MM5 control run, respectively. In general, MM5 short-term forecast ability is greatly enhanced up to 2-hour forecast period with the aid of cloud cover ingestion.

  7. References

Manabe R., J. Stouffer, M. J. Spelman, and K. Bryan, 1991: Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric CO2. Part 1: annual mean response. J. Climate, 4, 785-818.

Yucel, I, W. J. Shuttleworth, R. T. Pinker, L. Lu, and S. Sorooshian, 2001: Impact of ingesting satellite-derived cloud cover into the Regional Atmospheric Modeling System, Monthly Weather Review, (in press).

Figure 1a: Incoming surface solar radiation observed at the AZMET sites compared with the equivalent modeled values with and without cloud-cover ingestion during the period July 14-15, 1999.

 

 

 

Figure 1b: Modeled hourly average daytime surface solar radiation for model simulations without and with cloud ingestion, respectively.

 

 

 

 

 

Figure 2: Precipitation observed at the AZMET sites compared with the equivalent modeled values with and without cloud-cover ingestion during the period July 14-15, 1999. Red, blue and green lines represent values (mm/hour)without ingestion, with ingestion, and observation, respectively.

 

 

 

 

 

Figure 3: Forecast fractional cloud cover compared with observed cloud fraction derived from the GOES VIS images at 15-minute intervals during the 4.5-hour forecast period. Every 3 columns in Figure 3 indicate the observed cloud cover derived from GOES images, the forecasted cloud cover obtained by following 3 hours of 1-minute cloud cover ingestion, and the forecasted cloud cover obtained from the MM5 control run, respectively.