ACD ASR 2003 ASP ASR 2003 ATD ASR 2003 CGD ASR 2003 ESIG ASR 2003 HAO ASR 2003 RAP ASR 2003 MMM ASR 2003 SCD ASR 2003 Go to NCAR Go to UCAR Go to NSF Go to NCAR's ASR 2002


Snowfall / Freezing Precipitation


B. Snowfall and Freezing Precipitation

[Background] [Freezing Drizzle Damage to 737 engines at DIA]
[WSDDM system at DIA]
[Addition of Freezing Drizzle Algorithm to Denver WSDDM]
[Short-term Forecasting of Snow in the Northeast Corridor]

1. Background

Progress of Hot Plate Development

For several years NCAR (R. Rasmussen, J. Cole and M. Tryhane), in conjunction with the Desert Research Institute (J. Hallet and R. Purcell), have developed a hotplate precipitation gauge. The principle behind the hotplate gauge is that precipitation striking a heated surface will cool the surface as the moisture evaporates. If the heated surface is maintained at a constant temperature, then the amount of power required to maintain this temperature should be proportional to the precipitation rate. An identical heated plate is mounted on the underside of the hot plate so that it is exposed to the same environmental conditions as the upper plate, minus the precipitation. The precipitation rate is then proportional to the difference in power that is applied to the two plates.

The hotplate concept has several important advantages over conventional precipitation gauges. It has a small profile so that wind effects during snowfall are minimal. It eliminates the need for antifreeze in winter, which means that it is more environmentally friendly. It is low maintenance and doesn't require a wind shield.

The major achievement this year was signing a contract with Yankee Environmental Systems for commercialization of the hotplate. The other achievement was prototyping the system at Denver International Airport.

2. Freezing Drizzle Damage to 737 engines at DIA

On the evening of 31 October 2002, several 737 aircraft departing Denver International Airport (DIA) experienced engine damage resulting from an accumulation of ice on their engine fan blades. At issue is the type of weather occurring at DIA during this event. The official NWS METAR observations at DIA show periods of very light snow, temperatures in the range of -8 to -9oC, winds generally northeasterly at ~10 kt, low ceilings and reduced visibilities. No mention was made of freezing drizzle, although the ASOS icing sensor data, shown in Figure 2.1, indicated freezing precipitation. The observation of freezing drizzle was confirmed by ground personnel responsible for deicing the aircraft. Meteorological observations at major airports in the U.S. are generated by Automated Surface Observing System (ASOS) stations, but may be augmented by observers if they feel that the automated report does not represent the true weather conditions.

Normally when freezing drizzle is reported, pilots follow procedures to eliminate the build up of ice on the engine fan blades. Since freezing drizzle was not reported in the official NWS observation, these procedures were not implemented. When the engines were powered up for takeoff it is believed that the ice was shed, damaging some of the fan blades. Two of the engines on United Airlines 737 aircraft had to be completely replaced due to the extent of the damage. This event points out an important safety hazard to 737 aircraft: the occurrence of unreported freezing drizzle at airports resulting in damage to the fan blades. Additional work involving a similar case in Norway indicates that there is also an issue regarding the accuracy of drizzle intensity reporting using visibility. R. Rasmussen and C. Wade are currently writing a paper describing these cases.

NCAR/RAP has been working with the NWS, the FAA and United Airlines to develop a solution to the problem of unreported freezing drizzle. During 2003 an algorithm was added to the DIA WSDDM system (see Sections 3 and 4 below) to monitor the frequency drop on the ASOS icing sensor and issue advisories or alerts every 5 min when icing conditions are detected. These advisories are intended to supplement the official weather observation from the NWS and are experimental at present. If successful they may be useful in the development of a more system-wide icing alert system for the FAA and NWS.

Figure 2.1

3. WSDDM system at Denver International Airport

During 2003, NCAR/RAP signed an agreement with the City of Denver to install a WSDDM (Weather Support to Deicing Decision Making) system at Denver International Airport (DIA). This system was installed in April 2003 and will provide real-time weather information for the City of Denver Operations at DIA consisting of NEXRAD radar data, storm tracks, current precipitation rates from a network of heated gauges, 30-min forecasts of precipitation rate, and alert information when icing conditions (freezing drizzle, freezing rain and freezing fog) are being observed on the local weather sensors. Computers located at DIA ingest data from weather stations located within 100-km of the airport and provide displays of current and forecast weather to various operational facilities at the airport (Figure 3.1). This information is used to improve decision-making regarding aircraft deicing, runway plowing, and airport operations during convective weather. This weather information is also made available at NCAR/RAP so that the performance of the system can be monitored and improvements made to system software.

Figure 3.1. Example display of current and forecast weather to various operational facilities at DIA.

4. Addition of Freezing Drizzle Algorithm to Denver WSDDM

a. Detecting Drizzle by ASOS

The detection of drizzle, and in particular freezing drizzle, has been an issue of some concern to the National Weather Service (NWS) ever since it began using the Automated Surface Observing System (ASOS) in the early to mid 1990's. Currently drizzle is not reported on ASOS unless an observer is present to augment the observation. This is due to several factors. First it was believed that the ASOS present weather sensor, LEDWI (Light-Emitting Diode Weather Identifier), could not detect drops smaller than ~1 mm diameter (drizzle has diameters of 0.2 - 0.5 mm). Second, there was no requirement in the ASOS specification for the present weather sensor to detect precipitation at rates lower that 0.01 in/hr (0.25 mm/hr). Drizzle generally occurs at rates below these thresholds. Thus, no effort was made to develop a drizzle detection algorithm for this instrument.

During the past several years C. Wade analyzed 1-min ASOS data collected during drizzle events and determined that drizzle is indeed detectable by LEDWI, provided that data from other ASOS sensors are used to help identify when drizzle conditions are likely. These conditions include low cloud bases (<1500 ft), high relative humidity, and decreased visibility. During periods of clear weather the background signal level in LEDWI is elevated due to turbulence created by heat exchange processes near the surface. The detection threshold for precipitation is therefore set to a sufficiently high threshold so that turbulence is not falsely identified as precipitation. When drizzle occurs the LEDWI signal level is generally below this threshold and therefore the drizzle goes undetected. By using data from other ASOS sensors it is possible to determine when drizzle conditions are occurring, which allows the detection threshold to be lowered. This research, reported by C. Wade (2003), shows how drizzle can be identified, but does not attempt to distinguish it from freezing drizzle (see next section).

b. Detecting Freezing Drizzle by ASOS

To identify freezing drizzle it is necessary to use the LEDWI data in conjunction with data from the ASOS icing sensor. The icing sensor was added to the ASOS suite of sensors in the mid 1990's in response to the need for ASOS to detect freezing rain. The icing sensor consists of a small metal rod, approximately 2 cm in length and oriented vertically, that is induced to vibrate at 40,000 Hz. When ice forms on the rod it causes the vibration frequency to drop. This is illustrated in Figure 4.1 for a freezing drizzle event that occurred in Denver on 22 January 2003. Periodically the rod is heated to prevent the buildup of large amounts of ice, and these heating cycles appear as the sudden increase in frequency back to 40,000 Hz. Heating cycles typically last for about 5 min before the temperature of the rod cools to environmental conditions. Experience with the icing sensor has shown that it is responsive to many forms of condensed moisture when temperatures are below freezing, including freezing fog, freezing drizzle and freezing rain. The second trace in Figure 4.1 is from the LEDWI Particle channel and is proportional to the particle size. The small increase in the particle signal just after 0800 UTC is consistent with the onset of freezing drizzle; while the more significant increase at 1600 UTC represents the point at which snow began. Notice that shortly after the snow began, the icing sensor frequency stopped decreasing, indicating that the freezing drizzle had ended.

Figure 4.1. Illustrating the behavior of the ASOS icing sensor frequency on 22 January 2003 during a period of freezing drizzle and snow at Denver International Airport.


c. Demonstration of a Freezing Drizzle Algorithm on the Denver WSDDM System

Using 1-min ASOS data from freezing drizzle events like the one illustrated in the previous section, it is possible to develop an algorithm that can indicate not only when freezing drizzle is occurring, but the icing rate during the event. Results from previous studies of the relationship between the rate of frequency decrease on the ASOS icing sensor and the accumulation of ice on horizontal surfaces are used in this algorithm as indicators of the instantaneous icing intensity and the total ice accumulation since the start of the event. This information is displayed as a part of the WSDDM system display at Denver International Airport. This experimental product (Figure 4.2) is intended to augment the official NWS weather observation. The greatest difficulty in using this algorithm in an operational setting is obtaining the latest 1-min ASOS data in near real-time. RAP is working with the NWS to develop an efficient way to do this without interfering with the normal operation of the ASOS system.

Figure 4.2.

5. Short-term Forecasting of Snow in the Northeast Corridor

The goal of this effort, conducted by M. Xu, A. Crook and R. Rasmussen is to develop a data ingest and numerical forecasting system for short-term forecasts of snowfall in the terminal area. During FY03, a real-time, mesoscale prediction system, based on the real-time four dimensional data assimilation (RTFDDA) system RAP developed for the Army test ranges, was adapted and tested. A radar data assimilation component was added to incorporate Level II multi-radar observations. Case studies were conducted for snowstorm events that occurred in northeastern US during winter 2002-2003 and performance of RTFDDA and effects of assimilating radar observations in the Northeast Corridor (approximately Cleveland to New York City) were evaluated.

The RTFDDA system is built upon a multi-grid, high-resolution MM5. Surrounding the numerical core are the data ingest, decoding and processing modules for different types of observations. For forecasting in the Northeast Corridor, a three-grid configuration, with grid increments of 3.3 km, 10 km and 30 km, is used (Figure 5.1). In addition to traditional observations, real-time 3D mosaic radar reflectivity datasets are available in an area similar to the model grid 2 (Figure 5.2, for example). An analysis nudging scheme is used for assimilating the mosaic reflectivity. The radar reflectivity is converted to 3D rain or snow mixing ratio (qr) and interpolated to the model grid. The mixing ratio field, together with the corresponding latent heating, are nudged on the two inner meshes of MM5. Case studies reveal that the model is able to simulate the formation and general movement of the storms. However, without radar data assimilation, the modeled storms are usually a few hours behind their observed counterparts. When the snow mixing ratio field (qr) alone is nudged, at the end of nudging, the model qr field is comparable to that observed. When the latent heat is nudged in addition to qr, its effect on the temperature and wind is significant. The changes in temperature and wind, though noisy, are in the direction of reproducing the observed snowbands. The vertical velocity produced by latent heat nudging is in general consistent with the observed qr, but not in exact balance. As a result, there is still a period (~3 h) of spin-down/spin-up after nudging. Overall, the correlation between forecast and observed snowfall is improved when the mosaic reflectivity is nudged (Figure 5.3).


Figure 5.1 The model grid used in RTFDDA for snowfall forecast in New York. The grid increments are 3.3 km, 10 km and 30 km, respectively.

Figure 5.2 Mosaic reflectivity observations of winter storm on Dec. 11, 2002.

Figure 5.3 Mean correlation between the forecast and observed qr fields. The curves are for RTFDDA without any data assimilation (dashed); with traditional observations but no radar data (solid); and with radar and other observations (dotted dashed).