Through collaborations partially funded by the Alaska Space Grant Program, SPoRT has been closely working with Dr. Don Morton at the Arctic Region Supercomputing Center at the University of Alaska Fairbanks to bring retrieved profiles from the Atmospheric Infrared Sounder (AIRS) into an operational forecast model. Don works very closely with Alaska NWS WFOs by providing an operational version of the High Resolution Rapid Refresh configured for an Alaska domain (HRRRAK) to help improve short-term forecasts. SPoRT has provided near-real-time AIRS retrieved profiles and guidance on configuring and running the Gridpoint Statistical Interpolation (GSI) data assimilation system to most effectively bring the observations into the operational forecasting system. Testing of the impact of the AIRS profiles within the system will continue for a few weeks, after which forecasts including AIRS profiles will be provided each day to Alaska region WFOs. An example of the initial analysis differences for 850 hPa temperatures from last Friday can be seen below. SPoRT will continue to work with Dr. Morton to evaluate and validate the impact of AIRS on HRRRAK short-term forecasts.
Archive for the ‘Data Assimilation’ Category
Starting Tuesday, February 28, and running through March 1, SPoRT will be hosting its 6th Science Advisory Committee (SAC) meeting at the NSSTC in Huntsville, AL.
One of the topics to be shared with the SAC is the SPoRT-WRF, a real-time, experimental high-resolution model that combines SPoRT research projects (Land Information System, SPoRT SST composite, MODIS Greenness Vegetation Fraction, and assimilation of hyperspectral profiles from AIRS and IASI) into a single cohesive model to address the WFO and SPC forecast challenge of convection in numerical weather prediction. Today’s 00Z initialization of the SPoRT-WRF shows clearing conditions over the Huntsville area and only isolated afternoon and evening showers over the Atlanta area (one location some SAC members may be flying through this afternoon/evening) with clearing after around noon local time.
SPoRT looks forward to a great meeting!
The dearth of upper air observations over the Pacific Ocean can lead to model initializations that contain very little observational data and revert to the background field. While this may be adequate, improvements in the initial conditions can yield improved forecasts. One particular forecast challenge for WFOs in Alaska, the Pacific and the West Coast of the United States is obtaining a correct moisture analysis over the Pacific Ocean. Streams of moisture called atmospheric rivers can stream ashore the North American Coastline and produce intense rainfall due to orographic lifting and may not be well-represented in models and analyses.
Currently, some offices are using a total precipitable water (TPW) product from CIRA that SPoRT has helped transition to the operational community. This product blends various satellite and GPS observations of TPW into a blended product that can be used to track atmospheric rivers and other moisture features. However, this particular product is not available to be used in model initialization. An additional source of satellite atmospheric moisture observations comes from NASA’s AIRS hyperspectral sounder. AIRS is able to see temperature and moisture in clear and partly cloudy conditions and produce vertical profiles of temperature and moisture. SPoRT is currently blending a GDAS analysis with AIRS temperature and moisture observations over the Pacific using the Gridpoint Statistical Interpolation data assimilation system to produce a three-dimensional moisture analysis product.
An example of the SPoRT AIRS 3D Moisture Analysis Product is given below for an atmospheric river case on 14 October 2009 that impacted the Western United States.
In the above example, the CIRA TPW (top image) is taken as the observed state of the atmosphere. The GDAS analysis (middle image) used to initialize the 00Z GFS shows the TPW pattern consistent with the atmospheric river; however, the atmospheric river is too wide and is slightly more intense than indicated in the CIRA TPW product. Additionally, the areas surrounding the atmospheric river are too moist compared to the CIRA product. Upon assimilation of the AIRS temperature and moisture profiles (bottom image), the width of the atmospheric river tightens and the surrounding areas are dried. The resulting 3D analysis compares more favorably to the CIRA TPW product.
Future iterations of this same product will assimilate profile data from the Infrared Atmospheric Sounding Interferometer (IASI) and Cross-track Infrared Sounder (CrIS) using lessons learned from AIRS to produce an analysis valid at multiple times during the day.
SPoRT has been collaborating with Don Morton and Kayla Harrison at the University of Alaska-Fairbanks to bring AIRS profile data into the High-Resolution Rapid Refresh for Alaska (HRRRAK), which is being developed for operational use at Alaska Region National Weather Service Offices. In this research activity, SPoRT provided AIRS profile observations in PREPBUFR format for assimilation into the Gridpoint Statistical Interpolation (GSI) used to initialize the HRRRAK. The following figures show the impact of the AIRS profile data on the GSI analysis.
It is clear that the assimilation of both conventional observations and AIRS profiles has a larger impact on the analysis than assimilation of conventional observations alone at this time for this pressure level. More work will be done to determine the forecast impact when the HRRRAK is initialized with each of these analyses and the forecasts are compared to in situ observations of sensible weather parameters.
The 19 May SPoRT run of the Weather Research and Forecasting (WRF) model captured a band of strong convection that developed in advance of a dryline across Kansas and Oklahoma. The mode and orientation of the convection appeared quite similar to the observed radar reflectivity in the late afternoon and evening hours. On the multi-model comparison page as part of the 2011 Hazardous Weather Testbed’s Spring Experiment, the SPoRT-WRF model is compared to the National Severe Storms Laboratory (NSSL) and the National Center for Atmospheric Research (NCAR) WRF runs. For this particular day, the SPoRT-WRF best captured the intensity and timing of the convection over Oklahoma and parts of Kansas during the late afternoon and evening hours (see Figure below of reflectivity comparisons valid at 0000 UTC 20 May 2011). The SPoRT WRF model configuration is nearly identical to the NSSL configuration, but incorporates real-time MODIS vegetation fraction, high-resolution land surface initialization data from the NASA Land Information System, MODIS/AMSR-E SSTs, and also assimilates AIRS temperature and moisture profiles to improve initial conditions and subsequent forecast parameters. Of course, this is only one case; SPoRT team members plan to continue examining the model comparisons throughout the duration of the Spring Experiment and beyond.
In the past few months, the NASA SPoRT Center has acquired two new “desktop supercomputers” as part of NASA’s broader “Climate in a Box” program. Locally, these machines are being configured to provide both real-time forecasting and research capabilities to the modeling and data assimilation team at SPoRT. In real-time forecasting applications, the WRF model is being run over a CONUS domain with unique NASA/SPoRT Center data sets and contributions: inclusion of our high resolution sea surface temperature product, high resolution initialization of soil moisture and land cover characteristics via the Land Information System and vegetation composites provided by MODIS, and additional information from AIRS temperature and moisture profiles assimilated around 09 UTC. Current plans are to compare this local run against a similar forecast produced as part of the NSSL Spring Program to identify changes between the forecasts, then to relate these changes to the unique initial conditions. The second system will provide this research capability in an off-line mode and support other activities at SPoRT.
Some differences in forecasts were noted earlier this week in the prediction of a severe squall line moving through the Midwest. In the SPoRT-WRF forecast, the line of thunderstorms appeared to bow out into a series of small segments and with faster propagation speed versus the NSSL run. This case, and others, will be examined over the coming months to identify opportunities for further study. Overall, the purpose is to identify the impacts of these data sets and to improve their use within high-resolution, short-term forecast models.
The SPoRT ADAS produces a high-resolution surface analysis of temperature, dew point, relative humidity, and winds and has been very beneficial to some of SPoRT partnering offices. One of the advantages of the SPoRT ADAS system is it’s flexible, user-defined quality control whereby forecasters can identify questionable observations and request that they be removed from the analysis for one or all of the analysis variables.
Brian Carcione at the Huntsville WFO recently identified two stations in DeKalb County (denoted by “X”s in the figures) as having dew point reports that were consistently over 80F when dew points in the rest of their CWA did not exceed 72F. The top figure shows the SPoRT ADAS dew point analysis when no blacklisting is applied. For this analysis, the dew point over DeKalb County is over 80F and approaches 90F in some areas! Not only does the analysis change in the area around the questionable observations, but the larger, unrealistic-looking dew points spread over much of northeastern Alabama (partially due to there being no other observations in that region). When these two observations are blacklisted, the new quality-controlled dew point analysis produces more realistic-looking fields with similar range across most of northern Alabama (bottom figure).
This example highlights the need for forecasters to actively participate in the blacklisting process for the SPoRT ADAS. If you know of a station in your CWA that is consistently producing unrealistic looking temperature, moisture, or wind observations, please contact SPoRT to have your stations added to the blacklist. The process to add stations is simple and should take only a few minutes, so please send in as many of the observations from your region that you think should not be included. Together, we can make this a great surface analysis!
Based on feedback from the 2010 SPoRT/NWS Partnership Workshop, SPoRT has improved the timeliness and yield of the SPoRT ADAS product to improve its utility in the operational environment. No changes have been made to the product itself (i.e. the same observations and background field are used); however, the analysis processing have been augmented with the changes outlined below. Now, the SPoRT ADAS product will be available in partnering offices’ AWIPS systems at around 25 past each hour. Increased timeliness will allow the operational forecasters to better use the product for initializing their gridded forecasts. Increased yield will allow for greater utility for grid verification.
The timeliness is improved by implementation of the following:
- 2-hr RUC background field is now obtained and processed prior to the analysis run time.
- Analyses are sent directly to the SPoRT LDM server rather than being rerouted due to firewall issues.
- New cluster head node is slightly more efficient in computation than the previous cluster head node.
The yield is improved by implementation of the following:
- Dynamic preprocessing of observations and analysis files based on available data.
- Bug fix in METAR preprocessing script that periodically miscounted cloud levels leading to analysis input error.
The image below is the first ADAS image–from 1500 UTC on 7 April 2010–to be sent using the new scripts.
Each NWS WFO can provide additional enhancements to the product by pinpointing observations in the MADIS data sets that they know have a consistent bias. The SPoRT ADAS has a dynamic blacklisting system by which observations that are questionable only during one part of the day (e.g. a poorly placed mesonet site that is placed in a shady location that consistently reports lower temperatures) can be removed from the analysis. Thus, forecasters should communicate their subjective assessments of observation data quality within their CWA to SPoRT for incorporation into the SPoRT ADAS objective analysis.
NASA’s Atmospheric Infrared Sounder (AIRS) has the capability of producing near-radiosonde quality vertical soundings with a vertical resolution much greater than previous sounders. SPoRT has successfully assimilated temperature and moisture soundings retrieved from AIRS into the WRF 3DVAR analysis system (WRF-Var) and assessed using 37 case study days from winter 2007. The assimilated data produce a 3-dimensional analysis that is then used to initialize a WRF forecast on a regional, 12-km domain.
One particular case in which AIRS has excelled is a February 13, 2007 extreme rainfall case in Eastern Texas prompted by a cold front pushing eastward across the Great Plains. On February 12, clear skies over Eastern Texas and the Western Gulf of Mexico allowed for the observation and assimilation of many high-quality AIRS profiles at around 0900Z. The figure below shows the location of the AIRS profiles. The colored pixels indicate the pressure above which the highest quality AIRS profiles are assimilated (e.g. black pixels denote profiles that have quality data all the way to the surface). Note that there is a large swath of AIRS profile data located over the Western Gulf of Mexico.
The WRF model was initialized at 0000 UTC on February 12, data were assimilated at 0900 UTC, and these AIRS-enhanced initial conditions were used to finish the forecast out to 24 hours. The image below depicts three model metrics from the 24-hour forecast (valid at 0000 UTC on February 13): 6-hour cumulative precipitation (top row), 1000 hPa dew point temperature (middle row), and convective available potential energy (CAPE). The far right column is the verifying “ground truth” analysis. For 6-hour cumulative precipitation, the truth field is gridded Stage IV precipitation fields, which combine radar and rain gauge data. For the dew point and CAPE fields, the North American Mesoscale (NAM) model analysis valid at 0000 UTC on 2/13 is used. The left column represents the Control (no AIRS) forecast; the middle column the AIRS-enhanced forecast.
With the inclusion of AIRS temperature and moisture profiles (upper middle), the intensity and location of the 6-hour cumulative precipitation maximum in the line of storms crossing Eastern Texas is represented much better than in the Control (upper left) compared to the Stage IV precipitation data (upper right). This convective precipitation is improved with the inclusion of AIRS due to improved forecasts of CAPE and lower-level moisture, which lead to greater instability in the model in that region. These improvements are validated in the comparisons to the NAM analysis. The advance of the 18-20 degree Celsius lower level dew points off the coast of Texas is depicted better in the AIRS-enhanced model run (middle) compared to the Control (middle left). Similarly, the AIRS-enhanced CAPE (lower middle) shows some of the 1800 J/kg CAPE values approaching the Texas coast that appear in the NAM analysis of CAPE (lower right). The Control run (lower left) has no CAPE values in that region above 1400 J/kg.
The LMA Source Density Product has been a very useful tool during severe weather operations, particularly when used in conjunction with radar data. At WFO Morristown TN, an AWIPS bundle has been created that allows for quick evaluation of the LMA Source Density data along side radar data. The Panel Combo Rotate feature in AWIPS is utilized to display multiple radar elevation slices in a four panel display while also displaying other radar products and the LMA data.
The example below is from a supercell that moved across southern Middle Tennessee (in Lincoln County in the WFO HUN area).
The first image above shows the base bundle, and the second image shows it when the image toggle is used.
The following products are displayed in the above bundle:
Upper left panel: All-tilts reflectivity (first image), image toggle with VIL (second image).
Upper right panel: All-tilts SRM Velocity with TVS Rapid Update (first image), image toggle with digital VIL (second image).
Lower left panel: LMA Source Density with 5 minute lightning and Hail Index (first image), image toggle with VIL Density (second image).
Lower right panel: All-tilts Velocity with Digital Mesocyclone Data (first image), image toggle with Enhanced Echo Tops (second image).
Using the Panel Combo Rotate feature in AWIPS allows for the images to be looped in time, as well as looping through radar elevation slices. The four panel display allows for quick and easy comparison of radar features with the LMA data.
One interesting note with this particular storm – the NLDN 5 minute lightning data only showed one lightning strike, while the LMA Source Density values was very high – over 200. The high LMA Source Density value indicated the presence of large hail, confirmed by a large three-body scatter spike.
Doug Schneider, WFO MRX