Improving Prediction of Weather, Climate, and Other Atmospheric Phenomena
Highlight: Pentagon and Urban Shield
The Pentagon, and its 25,000+ occupants, represents a potential target for a terrorist attack using chemical, biological, or radiological material released into the atmosphere. In response to this concern, the Department of Defense has engaged RAL to develop a building-protection system called Pentagon Shield (PS). The PS system assimilates meteorological and contaminant observations from remote and in-situ sensors into a complex linked system of models which operate together to represent processes from the mesoscale to the building scale. In the event of a hazardous-material release, the system calculates the properties of the contaminant source (e.g., location), the current characteristics of the contaminant plume, and the future path of the plume.
FY07 Accomplishments
Ground level winds in Crystal City VA, just west of Reagan National Airport, from the building aware model. A variety of building flow effects including wake flow reversal and channeling are shown.
A prototype building protection system was installed at the Pentagon in the beginning of the fiscal year. The system incorporates observational data feeds from Doppler radar and Doppler lidar, analyses of 3-dimensional wind fields from Doppler radial winds using VDRAS and VLAS, and numerical weather prediction capabilities from regional to metro scale using the Real-Time Four-Dimensional Data Assimilation System (RTFDDA). It blends observational and model data to provide redundant, continuous spatial and temporal coverage of non-building flow at a horizontal resolution on the order of 100 meters. Inclusion of building-flow effects through the use of two computational fluid dynamics (CFD) models provides characteristic flow fields at a scale of several meters within urban areas. By linking multiple, complex data-assimilation and forecast models to operate synchronously as a single system, to depict the urban boundary layer from the mesoscale to the street-canyon/building scale, the building protection system provides a new ability to detect chemical and aerosol/particulate releases and predict the transport and dispersion of those releases.
The prototype building protection system has been in operation throughout the year and has undergone extensive testing by both the development team and Pentagon staff. In August the domain area was expanded from 4km2 to 8km2 to support additional DoD facilities.
The main task during FY07 has been the development of Urban Shield, a further expansion of the modeling domain to cover a 168km2 area. This has required the development of a distributed computing environment for the building flow diagnostic model, which consists of 21 8km2 tiles computing the flow effects of thousands of buildings. The Urban Shield domain will be used to support emergency response efforts for both the DoD and Arlington County.
Plans for 2008
The current emphasis is on deployment and operation and maintenance of the Urban Shield system. Anticipated enhancements during FY08 include upgrading the transport and dispersion model to account for dense-gas effects and include a variety of release source models; inclusion of TDWR data in the VDRAS system; and porting the transport model to run on high-speed GPU hardware.
Highlight: Flash Flood Forecasting in Bangladesh
In 2000, the U.S. Agency for International Development (USAID) funded the Climate Forecasting Applications for Bangladesh (CFAB) project to provide advanced warning of severe flooding within the country of Bangladesh. CFAB is an innovative program that has made significant scientific and technological progress in flood forecasting and in communicating those forecasts to the Bangladeshi public. This effort is based at the Georgia Institute of Technology, and involves an NCAR/RAL investigator, and colleagues at the Asian Disaster Preparedness Center in Bangkok, Thailand.
While Bangladesh has its own well-developed river forecasting center, it has long been handicapped by very-limited and inconsistent river-discharge data-sharing between India and Bangladesh. As a result, two of its primary rivers, the Brahmaputra and the Ganges, were effectively ungauged basins above their entry point into Bangladesh from India, and advance warning of severe flooding events could only begin once observations were taken of the floodwaters crossing the India-Bangladesh border. To address this problem, CFAB issues operational forecasts designed to provide extended-lead-time information about the upper-catchment discharges of the Brahmaputra and Ganges Rivers before they enter Bangladesh. This is accomplished by utilizing ECMWF forecast products and NASA and NOAA precipitation estimates to produce short-range (1- to 10-day) and long-range (1- to 6-month) forecasts. Medium-range forecasts (20- to 25-day) use a statistical model to bridge these two time scales.
Disseminating forecasts has been a significant logistical challenge. To address this challenge, the Asian Disaster Preparedness Centre staff teamed with Disaster Management Committee chairmen in Bangladesh, with local non-governmental organizations, CARE, and the country’s Flood Forecasting and Warning Centre to establish a pilot dissemination network in 2006. A series of training workshops was conducted for the people within the pilot regions so they could effectively utilize the CFAB probabilistic discharge forecast information. Such training was crucial to building acceptance and trust of these citizens whose lives and livelihoods are so dependent on accurate and accurately interpreted forecast information.
FY2007 Accomplishments
The dissemination network was activated in 2007 in time to test the benefits of the short-range forecasting system for two severe flooding events. The late July event was one of the largest flooding events on record, inundating much of the northern regions of Bangladesh (and India). A report of the effectiveness of CFAB in mitigating the severity of the impacts on the people throughout Bangladesh, preceding and during the event, is given here by Selvaraju Ramasamy with the Asian Disaster Preparedness Centre: “The forecast was communicated to all our partners and local communities through Disaster Management Committee (DMC) chairmen in the pilot unions. The local partners, non-government organization (NGO) networks and DMC members were advised to inform the poorest of the poor, especially those people living in river islands (“chars”). The Flood Forecasting and Warning Centre (FFWC) incorporated the CFAB forecasts into their model and produced water level forecasts for 18 locations in Bangladesh. On the 28th and 29th, meetings were organized in villages near Rangpur (northern Bangladesh), where the Teesta River was flowing just a few inches below the rim. Local communities informed us that the river level at that time was even above the 1998 level. However, they perceived that the river water level would fall, but our forecasts showed a rising trend. We engaged the local partner NGOs to prepare an evacuation plan urgently, which we further discussed with them. The forecast helped our local partners and DMC chairman to prepare evacuation plans and mobilize the resources for response activities in advance.” As a result of the advance forecasts, areas that would be hardest hit were evacuated in advance; other areas not forecast to be inundated by flood waters had the warning time they needed to mobilize food and safe drinking water for a week to 10 days, protect their rice seedlings and fishing nets, and raise and protect their fish pods. For the first time 10-day advance official forecasts of significant chances of exceeding danger level in all the gauge stations along the Brahmaputra River were successfully communicated to the public. The dissemination network ultimately reached approximately 110,000 persons in five vulnerable regions within Bangladesh that have little access to advanced communication technology, let alone electricity.
FY2008 Plans:
The CFAB team will work to assimilate satellite remotely-sensed river discharge higher up in the Ganges and Brahmaputra watersheds to improve forecast skill and accuracy. The existing forecast warning dissemination network will also be expanded to include additional regions of Bangladesh.
Experience gained through the Bangladesh program will be used in a new effort in Africa. Hopson and David Yates of NCAR plan to work with Ben Lamptey (former RAL post-doc now with the Ghana Met Service), Prof. Robert Brakenridge at Dartmouth, and Prof. MeKonnen Gebremichael at the Univ. of Connecticut to extend the river discharge forecasting scheme to a number of river basins in Africa, with particular emphasis on the Volta River of West Africa and the Awash River in Ethiopia.
Climate Downscaling with MM5/WRF: The Climatological Four Dimensional Data Assimilation (C-FDDA) System
Figure a shows the average January precipitation amount based on MM5 simulations, for the inner computational domain at 15 km horizontal grid spacing. The rain gauge data in Fig. b are consistent with the MM5 estimates, and the coastal amounts of precipitation from the satellite/gauge merged data set show a similar pattern (Fig. c). Visual comparison of the amount and geographic distribution of monthly rainfall between model and the observations reveals considerable skill in the model simulation.
The Real Time Four-Dimensional Data Assimilation System developed at RAL generates a mesoscale re-analysis that is consistent with both observations and model dynamics. Even though it was originally developed for dynamic initialization of mesoscale forecasts with gridded data sets that contained fully developed mesoscale processes, the continuous data-assimilation process in RTFDDA is also ideal for generation of mesoscale climatographies. The RTFDDA technology, when applied in this way, is called the Climate-FDDA (C-FDDA) system.
FY2007 Accomplishments:
The current C-FDDA infrastructure, which involves performing model simulations, computation of statistical products, and model validation, is used in a number of applications:
- The Global Climatological Analysis Tool (GCAT) allows the U.S. National Ground Intelligence Center to generate a climatography for a region of interest with typical boundary layer conditions used to define likely directions and speeds of hazardous-material transport for different seasons and times of day
- Climatographic database for the Defense Threat Reduction Agency’s (DTRA) Joint Effects Model (JEM): C-FDDA will be used to create a global mosaic of moderate resolution (~40 km) climatographies for the 1979-2005 period. This database will be used by state, and local emergency managers for predicting the effects of accidental or intentional releases of hazardous material, as well as by military commanders for whom an understanding of “typical” atmospheric conditions in a given place on a given day will be useful in preparing strategic battle plans
- Eastern Mediterranean Studies: C-FDDA is also being used to study the hydro-climatology of the eastern Mediterranean and the adjacent countries of the Middle East, where the balance between water supply and demand could be significantly altered by climate change. Because precipitation data are not assimilated, and because many aspects of model physics typically need to operate properly for precipitation to be correctly simulated, this will be a good test of C-FDDA’s ability to define unobserved or poorly observed fields. We assess the success of the model by comparing the model-simulated precipitation with gage, radar, and satellite-derived estimates of rainfall in terms of the average monthly totals for January. Results are presented below in Figure 1 for six Januaries, from 2001 to 2006.
FY2008 Plans:
The results presented above are a first step toward downscaling global model simulations of future climates for the Eastern Mediterranean sea and surrounding land area. The preliminary simulation will be repeated later without the use of observations but using “grid nudging” towards the driving analysis to keep the model analysis from drifting. This process of model verification and adaptation for the area will then be repeated with WRF for the entire winter season, when most of the precipitation occurs. After this step, the model that better represents the regional and local climate will be selected and run for the same period using lateral-boundary conditions from a simulation of the present global climate by the NCAR Community Atmospheric Model (CAM) driven by observed sea-surface temperatures and land use. Lastly, the regional model will be run with the lateral-boundary conditions provided by future climate simulations conducted for the Fourth IPCC assessment report with the coupled ocean-land-atmosphere Community Climate System Model, CCSM. This regional-model output will be compared with the regional-model simulation of the present climate in order to assess the impact of future climate forcing scenarios on the components of the water cycle in this geographic area.
Microphysical Observations and Modeling
It is widely recognized that uncertainties and approximations in the microphysics parameterizations within numerical models are a significant contributor to forecast error and that microphysics parameterization in models needs to be vastly improved in order to significantly improve the skill of precipitation forecasts. An effort to develop multi-species microphysics schemes with accurate particle size distribution models and multiple moment schemes that are refined and verified with observations is progressing. The work involves coordinated system development with MMM through retrospective studies using operational and special field observations, and developing a new bulk microphysical parameterization for the Weather Research and Forecasting (WRF) model. Key areas for model improvement through upgraded microphysical schemes are: 1) quantitative precipitation forecasts and 2) cold pool and outflow formation and evolution.
FY2007 Accomplishments:
In recent years polarimetric radar and disdrometer measurements have been used to develop procedures for retrieving particle size distributions (PSDs) in storms. Analyses focused on convective storms reveal that drop distributions are well represented by a constrained-gamma model in which the shape and slope parameters of an assumed three-parameter gamma distribution are related. This essentially reduces the gamma drop size model to two parameters. For forecasting applications the parameters of choice are the liquid water content and the drop median volume diameter. The constrained-gamma is being refined by improving the method for estimating poorly sampled large drop concentrations, especially those at the leading edge of convection.
A new bulk microphysical scheme for WRF has been developed at RAL, incorporating a new snow PSD based on aircraft observations that represents snow particles as a sum of exponential and gamma distributions. The scheme is unique in that bulk snow density varies inversely with diameter instead of having a constant density as assumed in nearly all other schemes. Moreover, the scheme allows hydrometeors to have a generalized gamma form providing an opportunity to determine sensitivity to distribution shape parameter values as found in observational studies of convective storms. Squall line simulations, including the 12-13 June 2002 IHOP storm, have been run in WRF using this scheme with a model grid spacing of 1.0 km.
RAL scientists conducted a detailed microphysical study of two winter storms occurring during the IMPROVE II field program. An unexpected finding was that freezing drizzle formed outside convective updrafts and ice crystals formed in convective cores. These observations suggest that ice nuclei depletion and ice formation via supersaturation need to be included in microphysical parameterizations in order to properly simulate these types of storms.
The distribution of snowflakes in winter storms along the Front Range in eastern Colorado were examined using a video disdrometer. The snowflakes were dominated by roughly spherical particles having quasi-exponential or superexponential size distributions. Upon melting, raindrop distributions were more peaked. A nearly inverse linear relation between snowflake bulk density and particle median volume diameter was derived. Disdrometer measurements were also used to derive temperature-dependent aggregate terminal velocity relations. For a particular snowflake size it was determined that terminal velocities increase as temperatures increase. At –10ºC an aggregate with an equivalent volume diameter of 10 mm has a mean fall speed of about 0.9 m s–1. At –1ºC the fall speed is 1.5 m s–1. The increase with temperature is attributed greater riming.
FY2008 Plans:
Expected activities for the coming year include:
- Collecting and analyzing observations that directly address key uncertainties in microphysical parameterization schemes in forecast models
- Improving the representation of processes in microphysical schemes based on observations
- Conducting case studies
- Verifying the storm simulations with observations
The Colorado REFRACTT Demonstration
Figure 1. Refractivity and reflectivity plots from 20 July 2006. Yellow triangles show the location of MGAUS soundings launched at 19:06 UTC (red profile in Fig. 2) and 22:12 UTC (blue profile in Fig. 2). The refractivity fields depict the moisture variability in the near-surface portion of the boundary layer. Higher near-surface moisture values (blue) can been seen in the NW portion of the domain, while lower moisture can be seen in the SE portion of the domain.
Figure 2. Mobile GAUS soundings from 19:06 and 22:12 UTC illustrate the different moisture profiles and magnitudes at the two different locations within the REFRACTT domain. The lowest 50 mb mean moisture measurements are in good agreement with the moisture measurements retrieved using the refractivity (N) technique. In this case the near-surface moisture (refractivity) field is representative of the moisture present through the depth of the boundary layer.
The NSF-sponsored Refractivity Experiment For H20 Research And Collaborative operational Technology Transfer (REFRACTT) was conducted in NE Colorado from June – August 2006. This experiment provided a unique opportunity to collect high resolution, 2-D water vapor fields derived from refractivity data collected by three research Doppler radars and one operational NWS NEXRAD radar. There were two overarching goals of REFRACTT: 1) improve our understanding of near-surface water vapor variability and the role it plays in the initiation of convection and thunderstorms and 2) build operational advocacy for the refractivity moisture retrieval technique for ultimate installation on the U.S. national network of NEXRADs.
FY2007 Accomplishments:
Following the conclusion of real-time REFRACTT operations, data inventory and quality control were begun. The most important challenge was to reprocess the real-time radar refractivity data from the four REFRACTT radars using the best set of calibration datasets and ground target information collected during 2006. Frederic Fabry of McGill University, the author of the radar refractivity technique, assisted Rita Roberts and Eric Nelson in this process. Results from reprocessing the REFRACTT data using these improved calibration files include increased areal coverage of refractivity from the Denver NEXRAD data and elimination of a suspicious, persistent northwest-southeast gradient in N that was observed in realtime operations but is no longer present in the reprocessed data.
The considerable effort spent on data quality control and re-calibration of the refractivity fields was in preparation for case study analysis of convection initiation and assimilation of the refractivity data into a numerical model. Approximately six cases from REFRFACTT-2006 have been identified for detailed analysis on the variability of water vapor in the near-surface boundary layer in the pre-storm environment and its role in the initiation of new convection. One of the cases is being highlighted in a paper on REFRACTT-2006 that we are preparing for submission to the Bulletin of the American Meteorological Society (Figures 1 and 2).
The larger multi-radar domain of moisture information available from REFRACTT enables us to expand upon the refractivity-related research conducted using only the S-Pol radar refractivity fields collected during IHOP. Refractivity observations have the potential to improve numerical prediction of storm initiation and evolution when assimilated into a numerical model. The radar refractivity observations provide a unique dataset that, when assimilated along with the radial velocity and reflectivity, can enhance low-level moisture analysis in numerical models and hence has the potential to improve prediction of convective initiation. Some preliminary studies have been done assimilating IHOP refractivity data into a numerical model with promising results. Current efforts are underway by Jenny Sun and Eunha Lim to assimilate observations collected during the REFRACTT-2006 for a convective event that occurred on August 1, 2006 when a large scale influx of moisture into NE Colorado was observed prior to the onset of an intense line of thunderstorms.
FY2008 Plans:
Specific research planned for the next year includes documentation of moisture variability and transport during the hours leading up to convection to assess the relative impact of these changes on pre-conditioning the environment towards convection, storm initiation and storm evolution. This includes characterizing the moisture variability along surface convergence boundaries and with terrain-induced features. We propose to incorporate the findings from these studies into storm initiation predictor fields that can be included in the RAL heuristic thunderstorm nowcasting system.