B.
Snowfall and Freezing Precipitation 1. Background RAP has a successful history of involvement with airport and aircraft operations dealing with the impact of snow and freezing precipitation. The operation of aircraft during snow and freezing rain or freezing drizzle conditions is a significant safety issue due to the rapid loss of lift and increase in drag produced by ice on an aircraft. For example, a rough ice coating of only 0.8 mm on a plane's wing can result in a 25% loss of lift and increase in drag. Snow and freezing rain accumulations on taxiways and runways also impact the safety and efficiency of ground operations. The main goal of this research is to improve nowcasts and forecasts of snowfall and freezing precipitation. The research emphasis during FY 2002 has been on: 1) developing better nowcasts and forecasts for snow and freezing precipitation using numerical modeling and Doppler radar, 2) finalizing development of a hotplate snowgauge in collaboration with scientists from Desert Research Institute in Nevada, 3) publication of a new algorithm to detect freezing drizzle using raw data from the ASOS sensors, 4) improving the snow machine used to test deicing fluids for endurance time, 5) starting a Northeastern United States ceiling and visibility study, 6) evaluating snow forecasts from the RUC operational forecast model, and 7) obtaining verification data for MM5 snow forecasts during the 2002 Winter Olympics. 2. Northeast Corridor Ceiling and Visibility Initiative The
Northeast Corridor Ceiling and Visibility initiative was recently
established, with the objective of developing improved short-term
(0-12 hrs) forecasts of low ceiling and reduced visibility (C&V)
conditions adapted to the environment of the northeastern United States.
The main thrust of this past year was to characterize the reduced
visibility problem for the New York area by performing a regional
climatological analysis of fog events, and by establishing a field
program aimed at gathering detailed observations of the lower atmosphere
during low ceiling and reduced visibility conditions. An instrumented site was set up on the campus of the Rutgers University in New Brunswick, New Jersey. The site already included some instrumentation, as part of a Photochemical Assessment and Monitoring network (under the responsibility of New Jersey Department of Environmental Protection). NCAR deployed dedicated sensors complementing those already present. From mid-April to September, about fifteen reduced visibility events were identified. A preliminary analysis showed general agreement of the temporal evolution of visibilities at the various locations (Figure 2). Some reduction in visibility during the night was observed for all sites, as would be expected due to higher relative humidities. Nevertheless, differences in amplitude and timing of variations in visibility at the various locations suggest significant variability at the regional scale. The analysis also pointed out the roles of precipitation, radiational cooling/heating and regional advection patterns on determining the life cycle of fog layers.
Figure 1. Frequency of hourly fog reports, over the 1977-1990 period, for 5 stations in the New York region.
Figure
2. Observed horizontal visibilities at the Rutgers C&V site and
at the three major terminals in the New York region, [TOP] 3. Evaluation of Snow Forecasts from the RUC Operational Model Analysis Improvements During FY02, substantial improvements were made to the RUC model verification process, and the resulting evaluation was considerably more comprehensive and accurate than previous assessments. Software improvements included expanded capability, increased reliability, and performance enhancements. Results now address numerical fields, such as temperature and dew point, which may offer additional constraints concerning the occurrence of icing. The elimination of intermediate Netcdf files and the use of the Matlab analysis software from the verification process has resulted in a dramatic performance enhancement. Verification results that previously required a week to derive can now be produced within a few hours. Verification
Results Three- and six-hour forecasts were evaluated for the 2000-2001 and 2001-2002 winter seasons at six locations: KDEN, KORD, KMSP, KLGA, KJFK and KEWR. Forecast parameters evaluated were temperature, dew point, and categorical precipitation type. The temperature analysis revealed a significant, although not profound, model bias. In general, RUC forecasts tended to be about 1C too cold (Fig. 3a). Absolute errors were approximately twice this value (Fig. 3b). The dew point analysis indicated much smaller model bias of 0.2 C to 0.4 C (Fig. 3a). Absolute errors were comparable to those associated with temperature (Fig. 3b).
Temperature and dew point Probability of Detectin (POD) scores were determined for a range of tolerances between 1 and 5 C. For a 1 C tolerance, the three-hour temperature forecasts were associated with an average POD of approximately 0.4. For a 5 C tolerance, the corresponding POD was approximately 0.97. In comparison, the three-hour dew point PODs slightly exceeded the temperature values at a 1 C tolerance and were slightly lower than the temperature values at a 5 C tolerance. Because error outliers result in a slower POD convergence to unity with relaxing tolerance, comparisons of temperature and dew point POD convergence rates suggest that dew point errors are more prone to outliers. The rain POD average was 0.62 for both three- and six-hour forecasts. Rain Critical Success Index (CSI) was slightly less than 0.4, in general. The results indicated that the skill scores from KDEN were exceptionally poor (POD of approximately 0.2, and CSI of 0.1 for both forecasts). POD and CSI scores for snow were 0.55 and 0.5, on average, respectively. More variability exists in POD and CSI values corresponding to the three-hour forecast than with the six-hour forecast. However, scores from all stations were generally uniform without prominent anomalies. [TOP]
4. Precipitation Ground Truth for the 2002 MM5 Verification Data Assimilation and Short Term Forecasting of Snow An accurate 3-6 h forecast of snow and feezing rain has been identified as one of the most desired products in a user’s need analysis. Our previous studies have shown that a high resolution mesoscale model is needed in order to realistically simulate the various forcings that are important for forecasting winter storms at this range. These previous studies have also shown that MM5 has certain skill in forecasting the occurrence of snowstorms in the 1-12 hour time scale. For airport applications, the key issues are the accuracy in timing, duration, and detail structures of the storm predicted by MM5. For example, correctly predicting a 30-min break between two snowbands can be crucial for airport deicing decision making. To improve the performance of MM5 in the terminal area, it is essential to initialize the forecasts using observations with higher spatial and temporal resolution. Doppler radars are at present the only observing system capable of sampling detailed patterns of snow storms. Techniques that can effectively assimilate radar data into MM5 have been explored. Using simulated as well as real data, we tested the feasibility of using the four-dimensional variational data assimilation (4DVAR) technique to assimilate Level II radar data into MM5. To be more applicable to real-time operations, a less time-intensive method, Newtonian Relaxation or nudging, is now emphasized. The nudging method, which has been shown effective in assimilating synoptic scale observations, is relatively untested for high- resolution data and model grids. As a first step, Observation System Simulation Experiments (OSSE) were conducted for a snowstorm event, using simulated data that emulate analyses from Doppler radar data. Results This case study is based on a snowstorm event that occurred on 10 December 1997 in the New York City area. The storm system developed on the Great Plains in association with a cold front and extratropical cyclone. It then moved northeastward and produced heavy precipitation in the midwestern states on 10 December 1997. The storm entered the New York City area around 16Z on December 10 and moved out and dissipated by 6Z on December 11. Well-defined snowband structures were the dominant features of this storm. The simulated radar data used in the OSSEs were obtained from an MM5 control simulation of the event. The overall intensity and location of the observed (control simulation) and forecast snowbands at 6 h forecast time (t=12 h) are shown in Fig. 5a. In the observation, there is one major snowband and the center of the snowband is situated over New York City. A set of five assimilation experiments were conducted by nudging various fields obtained from a radar analysis. The five experiments are: EXP1 - nudging qr, EXP2 - nudging u, v, qr, EXP3 - nudging qr, qv, T, EXP4 - nudging u, v, qr, qr, qv, EXP5 - nudging u, v, qr, T. The results from the Control experiment, a no-nudging assimilation, EXP2 and EXP4 are shown in Figure 4. As can be seen, EXP4, where u, v, qr and qv are nudged produced the best forecast at six hours.
. Figure 4. Column-average reflectivity at 6 h forecast time. (a) control simulation; (b) forecast from first guess without nudging; (c) forecast after 6 h nudging of u, v and qr; and (d) forecast after 6 h nudging of u, v, qv and qr. 5. A Multi-Sensor Approach to Detecting Drizzle on ASOS National Weather Service ASOS stations do not currently report drizzle because the precipitation identification sensor (LEDWI) is not considered to have the capability to detect particles smaller than about 1 mm in diameter. An analysis of the LEDWI 1-min channel data has revealed, however, that the signal levels in these data are sufficiently strong, when drizzle occurs, that they can be used to detect drizzle and distinguish it from light rain or snow. In particular, it is shown that there is important information in the LEDWI Particle Channel that has not been previously used for precipitation identification. This is illustrated in Figure 5, which shows a time series of LEDWI Particle Channel values from a freezing rain and freezing drizzle event that occurred at Peoria, IL on 18 February 2000. Five hours of freezing rain was followed by approximately seven hours of intermittent freezing drizzle.
[TOP] 6. The NCAR Snow Machine Roy Rasmussen, Alan Hills, Scott Landolt, Matt Tryhane, and Charlie Knight developed a machine to produce artificial snow. The snow produced by the machine is used to test aircraft de/anti-icing fluids. Deicing fluids are designed to remove ice and/or snow from critical aircraft surfaces via direct spraying of the heated fluid onto portions of an aircraft contaminated with ice and snow. Anti-icing fluids are designed to protect an aircraft from re-icing by melting any impinging snow when it falls into the fluid. Current methods of determining how long anti-icing fluids last during snowfall (this time period is called holdover time) involve outdoor testing of the fluid under natural snow conditions. While providing data on a particular fluid under actual snow conditions, this approach to testing can only be done outdoors in winter snowstorms, requiring considerable effort and expense. In addition, outdoor conditions are often highly variable, with quantities such as wind speed, direction, temperature, and snowfall rate changing rapidly. Thus, it is often difficult to compare tests conducted during a particular snowstorm with tests from other storms. In contrast to outdoor testing, indoor testing in a cold room provides a well-controlled environment in terms of temperature and wind (calm), and provides the opportunity of conducting testing year round. It also offers the opportunity of repeating tests to establish reliability and error tolerance limits. However, to be valid, a correlation between indoor and outdoor testing is required. . The NCAR snow machine generates artificial snow by shaving an ice core mechanically. Snowfall rate is controlled by feedback from an electronic balance. Previous tests have both shown that the NCAR snowmachine produces fluid holdover times 20 to 50% shorter than observed in outdoor tests. In order to help resolve the cause for this shorter time, a procedure was devised to conduct indoor tests at the same temperature and snowfall rate as the outdoor tests. During the past two years, anti-icing fluid tests were conducted outdoors at the NCAR Marshal Field Site using the same test plate assembly used for the indoor testing (Figure 6). Equivalent indoor tests with the same snowfall rate and temperature were conducted indoors. The comparison of the indoor to the outdoor tests (Figure 7) revealed good correlation, but a significant bias towards shorter times for the NCAR tests. Analysis of the plate temperature of the indoor and outdoor tests revealed that the outdoor plate temperature was significantly warmer than the indoor plate temperature. The plate temperature cools due to the latent heat release from the melting snow in the fluid. The outdoor plate, however, does not cool as much due to wind heating the plate. Since the deicing fluid viscosity is temperature dependent, with colder fluids having less viscosity, the fluids with colder temperatures run off the plate faster, and therefore fail more quickly than the outdoor fluids. Future tests will be conducted with constant plate temperature for the indoor plate in order to more closely replicate the outdoor testing.
Based on these data, drizzle detection algorithm was developed. Since noise in the LEDWI channels can sometimes obscure the drizzle signal, a technique was proposed that uses data from other ASOS sensors to identify non-drizzle periods and eliminate them from consideration in the drizzle algorithm. These sensors include the ASOS ceilometer, temperature and dew point sensors, and the visibility sensor. Data from the Peoria, IL, 18 February 2002 freezing rain and freezing drizzle event were used to illustrate how the algorithm can differentiate between these precipitation types. A comparison was made between the results obtained using the algorithm presented here and those obtained from the Ramsay freezing drizzle algorithm and precipitation type recorded by the ASOS observer. Data from the ASOS icing sensor show when ice was accreting on the sensor probe. By
using data from the LEDWI Particle channel, in combination with data
from other ASOS sensors, the ability exists to detect drizzle with
the current suite of ASOS instrumentation. 7.
The Hotplate Snowgauge
The Hotplate snow gauge was tested at the NCAR Marshall Field Site.
An analysis of the data collected during the 2001-2002 winter season
showed that wind speed was an important factor in the operation and
collection efficiency of the Hotplate snow gauge. By incorporating
the 10-meter wind speed, an improved algorithm was developed that
accounts for wind effects on the offset between the upper and lower
plates, threshold for precipitation onset, and under-collection due
to wind.
Figure 8. Accumulation and winds for the January 10, 2002 snow event. [TOP]
|
