• Scale dependence of predictability
• Ensemble forecasting on the mesoscale
• Verification of mesoscale model forecasts based on mesoscale predictability
• Experimental Numerical Weather Prediction
• QG+1 dynamics

Rebecca Morss continued work with Roger A. Pielke, Jr. (University of Colorado, Boulder) on developing a framework for understanding and investigating the connections among meteorological observations, weather forecasts, and society. Morss, in collaboration with Heidi Cullen (NCAR/ESIG/CGD), Olga Wilhelmi and Mary Downton (both NCAR/ESIG), also explored how uncertainty in weather and climate information affects floodplain management and flood policy. Currently, they are speaking to floodplain managers in communities along the Colorado Front Range to learn how weather and climate information is currently used and what type of additional information could help reduce society's vulnerability to flooding. Although improving quantitative precipitation forecasts (QPFs) has been identified by a number of meteorological organizations (including NCAR/MMM, the U.S. National Weather Service and the U.S. Weather Research Program) as a priority during the next few years, QPFs can be improved in a number of ways, and different types of improvements are likely to lead to different benefits. Furthermore, achieving different QPF improvements can require different research and operational efforts. Thus, Morss has recently begun an in-depth assessment of the needs of users of warm season QPFs along the Colorado Front Range. The assessment includes gathering and synthesizing existing information about current and potential forecast use, interviews with representative stakeholders in different user sectors, and, in some sectors, a user survey. The methodology was developed and the interviews and surveys will be conducted in collaboration with Eve Gruntfest (University of Colorado, Colorado Springs).

The skill of precipitation forecasts is limited by both fundamental and practical constraints. The practical constraints include the accuracy of the forecast model and the accuracy of its initial conditions, which is, in turn, determined by the available observations and their quality, and by the scheme used to assimilate those observations. The fundamental constraint is the finite limit of predictability, which arises even for an accurate model and initial conditions via the influence of unresolved scales.

Fuqing Zhang (Texas A&M University), Chris Snyder, and Richard Rotunno explored the limits of predictability for precipitation, within the context of the "surprise" snowstorm that paralyzed Washington, D.C. on 25 January 2000 (Zhang et al. 2001, 2002). In their simulations, rapid growth of forecast differences is associated with regions of moist ascent and moist convective instability. Using an embedded 3-km grid, they have shown that initial differences with scales of less than 100 km and amplitudes of less than 1 K grow rapidly by altering the position and timing of individual convective elements (in this case, in a region of negative lifted index over Louisiana). The differences then contaminate larger scales, altering the mature cyclone and the precipitation over the East coast 36 hours later. It is clear that this growth from small to large scales places an upper bound of a few tens of hours on skillful, deterministic precipitation forecasts.

Zhe-Min Tan (Nanjing University, China), together with Zhang, Snyder, and Rotunno, generalized these results beyond a single case study by considering the growth of small perturbations to an idealized, moist baroclinic wave, developing in a channel. They also find rapid growth of forecast differences in regions of (parameterized) moist convection.

While the work of Zhang et al. has focused on the intrinsic limits of predictability, questions of practical predictability are also of interest. Specifically: How skillful can forecasts be for a given observing network and forecast model; and, what are the characteristics of the forecast errors? Thomas Hamill (University of Colorado/ Cooperative Institute for Research in Environmental Sciences), Snyder, and Morss used a quasi-geostrophic model, along with a three-dimensional variational assimilation (3DVAR) scheme, to explore the characteristics of forecast and analysis errors at synoptic scales (Hamill et al. 2002, Snyder et al. 2002). They find that both forecast and analysis errors reflect the influence of the model dynamics: The errors have significant projection on the subspace of leading Lyapunov vectors; their time-mean vertical distribution, in both energy and potential enstrophy, is similar to that in the "true" state; and error variance in potential vorticity is typically confined to regions in which the true state has large gradients of potential vorticity. These results suggest a strong potential for improved data assimilation, on the synoptic scale, by schemes such as the ensemble Kalman filter (EnKF) or 4DVar that utilize dynamical information.

Presently, Hamill, Snyder and Jeff Whitaker (National Oceanic and Atmospheric Administration/Climate Diagnostics Center) are extending this work to the simplified, primitive-equation general circulation model of Held and Suarez, and are examining the analysis and forecast errors produced by the EnKF.

Current grid-point Numerical Weather Prediction (NWP) models have filtering properties that slow error growth, as measured quadratically. These filtering properties arise from both finite-difference methods and explicit diffusion designed to control spurious features. The effects must be taken into account for quantitative prediction and predictability studies. A collaborative effort between Joshua Hacker and David Baumhefner (NCAR/CGD) compared error growth characteristics of a current spectral, the Community Climate Model 3 (CCM3), and a future grid-point model, the Weather Research and Forecasting (WRF) model. This work involves calibration of the WRF model so that quantitative results can be trusted. Because CCM3 has been extensively tuned to reproduce observed atmospheric statistics and spectra, the researchers are using it as a benchmark for calibration. With both models, they generated 6-day ensemble forecasts for five cases from the 2001-2002 cool season in the northern hemisphere. The experiments identified strong sensitivity of ensemble spread, and the associated error growth, to numerical diffusion. Spectral analysis is one approach that shows diffusion-dependent error growth, relative to a forecast as shown in Movie 1. The results demonstrated the need to characterize error growth in the WRF model, and more generally, in all grid-point models. They contribute to the WRF model development by finding classes of problems for which its statistical behavior is unlike the atmosphere.

	To view the movie, place mouse over image. Alternately, for slower connections, you may use the links below to download the movie. WRF movie (animated GIF) WRF movie (AVI format) WRF movie (quicktime format)
Movie 1. WRF control forecast hemispheric Kinetic energy (KE) spectra (solid) and perturbation spectra (long dash), valid every 12 h from 00 UTC 24 Nov. 2001 to 00 UTC 30 Nov 2001.

Mesoscale data assimilation is hampered by two facts. First, observations that are plentiful (e.g., Doppler radar measurements of wind and reflectivity) involve only a subset of atmospheric variables, while observing platforms that measure all variables (e.g., radiosondes) are sparse and resolve mesoscale motions poorly. Second, the balances between variables, such as geostrophy, that are relevant at large scales in the atmosphere are questionable at the mesoscale; these balances are an important component of more traditional assimilation schemes such as 3DVar. To overcome these difficulties, there has been substantial effort within MMM and the Data Assimilation Initiative to explore the potential of the ensemble Kalman filter (EnKF) for mesoscale assimilation.

Snyder and Zhang (Texas A&M University) applied the EnKF to the analysis and prediction of convective scale motions, using the simple cloud model developed by Juanzhen Sun. They have shown that a 50-member EnKF is able to estimate tangential and vertical velocity and temperature, given 4-6 scans (or about 20-30 minutes) of simulated observations of radial velocity, extracted from a reference simulation of a supercell thunderstorm. These results are the first for the EnKF outside of global atmospheric models and hold substantial promise for the application of the EnKF to meso- and convective scales.

William Skamarock and Snyder extended such tests of the EnKF to other cloud models (including a prototype implementaion for WRF). Preliminary results for simulated supercells are similar to those with Sun's model, although the other models appear to be somewhat more sensitive to the details of the algorithm. This could be because they have reduced computational diffusion or because they are fully compressible rather than anelastic. Skamarock and Snyder have also successfully used simulated observations taken from an idealized squall-line simulation in the EnKF. These experiments extend the results from supercells to a situation in which individual convective cells are not quasi-steady state and where the assimilation period of 2 hours covers several cell lifetimes.

Four-dimensional variational data assimilation (4DVar), which is often seen as the state of the art for atmospheric assimilation, also has the potential to overcome the difficulties of mesoscale data assimilation. Alain Caya (NCAR/MMM/RAP), Sun, and Snyder compared the EnKF and 4DVar using the same simulated observations of a supercell. The two methods appear to be broadly comparable in both their performance and their computational cost. 4DVar has the advantage of producing a better estimate when observations are available for only a relatively short period of 10-20 minutes. The EnKF, on the other hand, is simpler to implement and provides an estimate of the analysis uncertainty in addition to the analysis.

David Dowell (NCAR/ASP/MMM) and Andrew Crook applied the EnKF to real radar observations of the Arcadia, OK supercell of May 1981. Dual-Doppler coverage is available for much of the lifetime of this storm. While the assimilations are clearly far from perfect at this early stage, Dowell and Crook find evidence that the EnKF is extracting useful information from the observations. One such result is that assimilating observations from one radar results in a steady improvement, over about 30 minutes, of the fit of the analysis to the observations from the other radar. They also find that the EnKF performs comparably to 4DVar for this real-data case.

To this point, all previous experiments with the EnKF have ignored the uncertainty in the environmental sounding, yet it is well known that simulations of convection are sensitive to the details of that sounding. Skamarock and Snyder have completed experiments, again in the context of a simulated squall line, in which the initial ensemble mean used in the EnKF includes error in the environmental sounding. They find that assimilation severely degraded for errors that are comparable to likely uncertainty in soundings (several m s-1), and they have begun exploring the possibility of estimating the environmental sounding based on the radar observations.

Dale Barker and Skamarock, together with Snyder and Jeff Anderson (NOAA/Geophysical Fluid Dynamics Laboratory), have begun work toward implementing the EnKF within the existing data-assimilation infrastructure of the WRF model. This builds on the prototype EnKF developed by Skamarock but would have access, for example, to the observation data structures, observation ingest capabilities, or background covariance models that already exist for WRF 3DVar.

Concerning adaptive observations, it might be concluded that the present observational network (as well as analyses and forecasts) could be improved by reallocating observational resources from regions where, on a given day, there are many observations or the weather is uninteresting, to regions of fewer observations or greater interest. Performing this reallocation, systematically, will require a quantitative estimate, prior to the actual measurement, of the impact of any observation on the analysis uncertainty. Hamill and Snyder (2002) have demonstrated in a quasigeostrophic model that such estimates can be made using the EnKF; this work represents an important initial step toward the eventual routine use of adaptive observing strategies.

Chris Davis and Kevin Manning, as part of a group led by Barbara Brown (NCAR/RAP) and including Randy Bullock (NCAR/RAP), Morss (NCAR/MMM/ESIG), and Cynthia Mueller (NCAR/RAP), have been investigating non-traditional verification techniques for Quantitative Precipitation Forecasts (QPF) in mesoscale models. The group initially focused on identifying contiguous rainfall areas as objects. They used the gross geometric properties of idealizations of these objects to assess model errors in rain area location, size, orientation, and PDF of rainfall rates, within areas. Work is continuing on defining an objective means of matching observed and predicted rain areas that accounts for inherent limits of predictability and is applicable across a range of spatial and temporal scales.

Davis and Manning, in collaboration with John Tuttle, David Ahijevych, and Richard Carbone, examined the ability of NWP models, including the National Centers for Environmental Prediction's Eta (Eta) model and the WRF model, to reproduce time-space statistics of warm season rainfall. While forecasts tend to exhibit some propagating rainfall features in time-longitude representations of meridionally averaged rainfall that are similar to observations, they poorly represent the diurnal and systematic zonally propagating signals of convection. In general, the models appear more able to predict the "corridors" along which convective systems propagate than the actual propagation. While the cause of such shortcomings remains unquantified, inadequacies in the parameterization of deep convection, on grids of 10-km spacing or coarser, are believed to contribute substantially.

To support flight forecasting and research activities over Antarctica, Jordan Powers, Manning, and Ying-Hwa (Bill) Kuo have been overseeing and developing the Antarctic Mesoscale Prediction System (AMPS). The team has been collaborating with a group headed by David Bromwich at the Ohio State University's Byrd Polar Research Center. AMPS is an experimental system generating twice-daily MM5 forecasts, at resolutions as high as 3.3 km, in support of the United States Antarctic Program (USAP). Additionally, AMPS serves a broad range of international activities, such as the Global Ocean Ecosystem Dynamics (GLOBEC) Program and emergency operations. In June 2002, the German supply ship Magdalena Oldendorff was servicing Russian stations along the Antarctic coast when she became trapped in thickening sea ice. To rescue the scientists and crew aboard, South Africa dispatched its vessel Agulhas from Cape Town to retrieve the personnel by helicopter. The South African Weather Service relied on the AMPS forecasts during the event. Special products relevant to the Oldendorff were provided, and Manning expanded an AMPS forecast grid to South Africa to cover the Agulhas' full route. Both the poor weather hampering the Agulhas' voyage, as well as favorable conditions allowing the final helicopter airlifts, were accurately forecast by AMPS. Figure 1 presents an example of an AMPS forecast during the transit of the Agulhas (position marked by "X") to the Oldendorff (position marked by dot). As it had in April 2001 with the South Pole rescue of American Dr. Ronald Shemeski, AMPS contributed to the success of an international emergency operation in Antarctica.

Fig. 1. Sea level pressure (contoured, interval= 4 hPa) and 3-hourly precipitation (shaded, mm) for Hr 39 of the 00 UTC 22 June 2002 AMPS forecast. 90-km grid output plotted; window of full grid shown. Forecast valid 15 UTC 23 June 2002. Dot marks position of Magdalena Oldendorff, while "X" marks the position of Agulhas. A deep low of 934 hPa southeast of the Agulhas, which caused her some difficulty in transit, is seen..

Fiscal year 2002 saw the culmination of a 6-year project to develop and implement a state-of-the-art, mesoscale prediction and data assimilation system for the Civil Aeronautics Administration (CAA) of Taiwan. The system, known as the Advanced Operational Aviation Weather System (AOAWS), represented a long-term collaboration between MMM and RAP to design, plan, create, and install the system at CAA and the Central Weather Bureau of Taiwan. In support of CAA's aviation forecasting, the AOAWS uses the MM5 and a 3DVAR data assimilation system to produce high-resolution predictions over Taiwan and East Asia, eight times a day. CAA formally accepted the system in June 2002. Powers, James Bresch, and Barker were the key scientists working as a team to bring the effort to a successful conclusion.

Understanding the relation between meso- and synoptic-scale flows involves, in part, understanding how atmospheric dynamics change as the Rossby number increases. At small Rossby number, virtually all dynamical theories rest upon the foundation of quasigeostrophy (QG), which is the leading-order theory in Rossby number. David Muraki (Simon Fraser University, Canada), Snyder, and Rotunno have introduced a convenient technique for extending QG to an additional order in Rossby number; they call this extended theory "QG+1."

Greg Hakim (University of Washington), Snyder and Muraki applied QG+1 to simulations of quasi-two-dimensional, balanced, decaying turbulence to understand the observed preference for cyclonic vortices on the tropopause, at subsynoptic scales (Hakim et al. 2002). These simulations produce numerous, small cyclonic vortices with sharp edges, while the anticyclones are infrequent, relatively large scale, and diffuse. These asymmetries appear to arise, not from corrections to the basic geostrophic balance (e.g., gradient wind balance), but through the action of the horizontally divergent component of velocity during frontogenesis.