National Weather Service United States Department of Commerce


Christopher N. Jones1*, J.D. Colton1, M.P. Meyers1, and J.E. Nachamkin2
National Weather Service, Grand Junction, Colorado
2Colorado State University, Fort Collins, Colorado


During the period of 19-21 December 1998 an intense arctic front moved across Colorado and Utah resulting in a drastic change of weather throughout the region. By the afternoon of 19 December, heavy snow was falling across sections of northern Colorado and Utah as the front sagged slowly south. Temperatures remained relatively warm (10�C) ahead of the front; however, temperatures fell nearly 30�C behind the front. During the morning hours of 20 December, the front slowly accelerated south spreading much colder, snowy conditions over most of western Colorado and eastern Utah by 1200 UTC (times hereafter UTC) 21 December. Snowfall amounts exceeded 30cm in many locations in western Colorado and eastern Utah. Heavy snowfall even occurred across the normally dry valley regions of Colorado, such as Grand Junction, where the 15th highest 24-hour snowfall total was measured on 20 December. As conditions worsened on 20 December, travel became difficult and impassable in many locations; major roads such as Interstate 70 were closed due to white-out conditions.

To investigate this case a combination of radar, satellite data, surface observations, and NCEP guidance was evaluated. Special emphasis will be placed on the Colorado State University's Regional Atmospheric Modeling System (RAMS), and its performance in handling the progression of the arctic front and the location and amount of precipitation.


An arctic airmass pushed across the western half of the United States 18-25 December 1998. The arctic surge ended several weeks of seasonably warm weather throughout much of the region. Typical of arctic airmasses pushing across the central United States, the cold air plunged rapidly south in the lee of the Rocky Mountains 18-19 December accompanied by light to moderate snowfall and temperatures below -18�C. Over the next several days the airmass deepened significantly and the cold air spilled west over lower mountain passes. Eventually the cold air edged slowly southwest into the Intermountain West, including Colorado and Utah. By 0000 19 December, the arctic air had moved through northwest Colorado and northern Utah; however, the complex terrain of northern Colorado and Utah temporarily impeded the southward progression of the arctic air. In many locations, moderate to heavy snowfall accompanied the cold air due to strong isentropic lift associated with the front, coupled with strong jet dynamics and strong west-southwest flow aloft. South of the front, warm and windy conditions prevailed over southwest Colorado and southern Utah. Strong southwest flow resulted in wind speeds up to 40 ms-1 over the higher terrain of southwest Colorado.
fig1.jpg (94326 bytes)
Fig.1. AVN model positioning of 500 hPa height (m) and vorticity valid 1200 UTC 20 December 1998.

By 1200 20 December, the mean 500 hPa trough over the western United States had sagged slightly south (Fig. 1). The prevailing synoptic flow for this event coincided well with that seen by Wesley et al. (1990) and Weiland (1994). The 1200 20 December surface depiction (not shown) also showed that the surface frontal position had edged slightly south from the previous evening. Widespread snow developed along and just ahead of the front. During the next 12 hours, the surface front moved more rapidly south. A combination of strong isentropic lift associated with the arctic boundary, strong jet dynamics (70 ms-1 jet maximum), and a favorable moist thermodynamic profile for dendritic crystal growth resulted in moderate to heavy snowfall over a broad area of southwest Colorado and southern Utah on 20 December 1998 (Table 1). Unofficial reports indicated that upwards of 91cm of snow fell at several mountain locations in western Colorado. Temperatures following frontal passage fell dramatically, with readings below -18�C encompassing most of the area. Temperatures in the vicinity of Craig, Colorado dropped to between -36�C and -38�C.

Elev (m) Location Snow (cm) Liquid (cm)
1230 Moab 11.4 0.76
1925 4sw Craig 16.3 1.09
1484 Grand Jct. 16.5 1.10
1761 Glenwood 17.8 1.19
1513 Delta 20.3 1.35
1939 Hayden 22.9 1.53
1636 Olathe 25.4 1.69
2658 Columbine 30.5 2.03
2413 Aspen 33.0 2.20
1633 Rifle 34.3 2.29
1711 New Castle 35.6 2.37
1774 Montrose 28.1 2.54
2047 4s Mesa 43.2 2.88
2526 Douglas Ps. 45.7 3.05
2447 Pyramid 48.3 3.22
2688 Telluride 53.3 3.55
3116 Grand Mesa 55.9 3.73

Table 1. Elevation (MSL), location, snowfall, and liquid water equivalent of event 19-21 December 1998. Liquid water equivalent assumes snow-to-water ratio of 15:1 based on observations.


The County Warning Area (CWA) of the Grand Junction Weather Forecast Office (WFO GJT) varies from desert areas of around 1220m MSL to mountaintop locations surpassing 4270m MSL. The complex terrain of the WFO GJT CWA played a prominent role in the evolution of this event. Foremost, several mountain barriers shown in Fig. 2 served to stall the arctic boundary over the Uinta Basin and northwest Colorado.

The main inhibitors to the southward progression of the arctic airmass were the East Tavaputs Plateau (E. Tavaputs, Fig. 2), the Roan Plateau, and the Flat Tops Range. The East Tavaputs and Roan Plateaus stand at 2400m-2750m MSL, approximately 1000m-1200m higher than the adjacent regions immediately to the north and south. And, contrary to the north-south ranges typical in the Rockies, these two plateaus are oriented west-east. The Flat Tops and Grand Mesa stand between 3000m and 3700m MSL. These higher terrain regions of the CWA all reported widespread snow amounts topping 46cm. Playing a lesser role in slowing the progression of the front were the La Sal, Abajo, and San Juan Mountains (all between 3300m and 4300m MSL) along with the Uncompahgre Plateau (2400m-3000m MSL). It appears that the arctic air, once into the lower valleys surrounding Grand Junction (KGJT), Montrose (KMTJ), and Moab, continued at a steady pace through the entire CWA by 1200 21 December. This forward movement of the arctic airmass was relatively quick in comparison to the time needed for the airmass to filter over the East Tavaputs Plateau (ETP), Roan Plateau, and Flat Tops during the preceding days.
fig2.jpg (43583 bytes)
Fig. 2. Major topographical features and cities in the County Warning Area (outlined with bold line) of the Weather Forecast Office at Grand Junction, Colorado (KGJT). Other cities included are Montrose (KMTJ), Craig (KCAG), Vernal (KVEL), Eagle (KEGE), Aspen (KASE), Cortez (KCEZ), and Moab.

Perhaps the most unique aspect of the event was the heavy snowfall of 19-21 December, and bitter cold temperatures that gripped the high valleys and lower river valleys of the WFO GJT CWA for several days following frontal passage. Most locations in the high valleys of northwest Colorado and the Uinta Basin of Utah range in elevation from 1550m to 2000m MSL. It is typical for the valleys of northwest Colorado to record temperatures below -18�C during the winter months, and indeed the coldest temperatures observed in this study were in this area. In contrast, rarely do heavy snowfall and arctic temperatures affect the lower valley regions, which encompass the cities of Grand Junction, Montrose, and Moab. Elevations of the lower valleys range from around 1220m to 1650m.


Various numerical models were available to the operational forecaster at WFO GJT during the event. The performance of these models is discussed here, with special emphasis placed on the accuracy of RAMS.

4.1 NCEP Models

Both the NGM and ETA models did a poor job of forecasting temperatures within the arctic airmass. The NGM hinted at bringing the colder air west of the Continental Divide, but predicted that the arctic front would lift back to the north before traversing the CWA from north-to-south. Although temperatures were not forecast well by the NGM, the timing of frontal passage was only slightly slower than what actually occurred. Conversely, the ETA model did poorly on both the timing and position of the front, never bringing the front south of the Roan Plateau. The AVN model, although similarly slow, did not bring the front into the lower valleys of eastern Utah and western Colorado.  However, it too, predicted temperatures warmer than those actually observed.  The only NCEP model that showed promise in accurately forecasting the timing and strength of the arctic airmass was the meso-ETA.  The timing of the front through Grand Junction just before 0000 21 December was captured well by the meso-ETA.  Forecast temperatures of this model were still too warm, but were a definite improvement over the NGM, ETA, and AVN.
fig3.jpg (34212 bytes)

Fig. 3. AVN storm total precipitation valid 1200 UTC 21 December 1998.

The other major challenge facing the operational forecaster during the event was that of snowfall amounts. The NGM, ETA, and AVN failed to accurately depict the total precipitation expected. In addition, the areal structure of the precipitation fields were too coarse to give reliable snowfall estimates on the mesoscale (Fig. 3). The one model that did an adequate job of predicting the location of precipitation was the meso-ETA (Fig. 4). This model showed the heaviest precipitation oriented in a northeast-to-southwest band structure that was observed on the KGJX WSR-88D (image not shown). The heavy snow band was associated with the progression of the arctic front through the CWA. The meso-ETA painted the heaviest precipitation over the higher terrain, likely a result of the enhanced grid resolution as compared to the NGM, ETA, and AVN. However, the magnitude of precipitation forecast by the meso-ETA was vastly under-predicted at most locations.
fig4.jpg (50223 bytes)

Fig. 4. Meso-ETA storm total precipitation valid 1200 UTC 21 December 1998.

4.2 RAMS Experimental Design

Through funding provided by the Cooperative Program for Operational Meteorology, Education, and Training (COMET), operated jointly in cooperation with the National Weather Service (NWS), Colorado State University (CSU), and the NOAA Forecast Systems Laboratory (FSL), a local mesoscale model, RAMS, is run daily at both CSU and FSL. We have employed a nested-grid version of RAMS (Pielke et al. 1992) provided by CSU for this project. The RAMS model output is available on the Internet at The CSU model is run daily at 0000 utilizing parallel processing for the interactive nested grid version of the model. The model includes two interactive, nested-grids. The coarse grid covers the western United States with 48km horizontal grid spacing; and the fine grid encompasses all of Colorado and portions of Wyoming, Utah, South Dakota, Nebraska, and Kansas at 12km grid spacing. Both grids utilize 26 vertical levels beginning at 250m grid spacing near the surface, gradually stretching to a maximum of 1000m.

One distinct advantage held by RAMS over the typical NCEP models is the more advanced physics employed. RAMS includes a mixed-phased microphysical scheme described by Walko et al. (1995). The model prognostic fields consist of cloud droplets, raindrops, pristine ice crystals, snow, aggregates, graupel, and hail. Each of these water categories are distributed according to a generalized gamma distribution. The bulk microphysical scheme is a major improvement over the "dump bucket" scheme similar to that found in Rhea (1978), which is typical to those used in other operational precipitation parameterizations. Another distinct advantage of this high-resolution mesoscale model is the improved representation of complex terrain, which is invaluable especially over mountainous regions.

4.3 RAMS Results

RAMS was initialized at 0000 20 December 1998 and simulations out to 36 hours were produced. The 6-hour surface wind forecast (not shown) valid for 0600 20 December positioned the arctic boundary along the north side of the Roan Plateau in Colorado, then northwest through the Uinta Basin. Northerly surface winds of 10 ms-1 were noted behind the front, in sharp contrast to the southwest flow of 10 to 15 ms-1 preceding the front. The southwest flow helped to boost observed temperatures into the 10�C to 15�C range on 19 December. RAMS estimations of temperatures (Fig. 5) behind the front were much colder, with readings confined between -25�C and -10�C. Precipitation forecasts of the initial 6-hour period produced by RAMS did well pinpointing observed snowfall, both quantitatively and spatially (Fig. 6). Of particular interest was the lack of forecast precipitation across the ETP. Throughout the event, the heaviest snow occurred along and just ahead of the arctic boundary. It appears that RAMS's limited precipitation rate forecast over the ETP was affected by the lagging position of the arctic boundary through the Uinta Basin.
fig5.jpg (115112 bytes)

Fig. 5. RAMS surface temperature forecast valid 0600 UTC 20 December 1998. Note: Temperatures are in �F.

fig6.jpg (80062 bytes)

Fig. 6. RAMS precipitation rate forecast valid 0600 UTC 20 December 1998. Note: Forecast is given in inches/hour.

By 1200 20 December, RAMS moved the western portion of the arctic front through the Uinta Basin to the ETP. Consequently, a significant rise was observed in the 12-hour precipitation rate forecast (Fig. 7) over the ETP. As was previously mentioned, the NCEP models did poorly at recognizing the push of the arctic air through eastern Utah, and in turn the timing and amount of precipitation across the region. The 1200 20 December forecast depicted isotherm packing along the north side of the ETP and the Roan Plateau (Fig. 8), with northerly upslope flow of 10 ms-1 prevailing (not shown). Wesley et al. (1990) showed how a low-level cold pool, characterized by weak to moderate upslope flow, would cause a bulge in the cold pool. This bulge was shown to form on the edge of the pool ascending the higher terrain. It was also demonstrated by Wesley et al. (1990) that over several days the cold pool could triple in depth. It appears that the depth of this arctic airmass may have been undergoing a similar deepening as it was wedged against the two plateaus in moderate upslope flow.
fig7.jpg (77187 bytes)

Fig. 7. RAMS precipitation rate forecast valid 1200 UTC 20 December 1998. Note: Forecast is given in inches/hour.

fig8.jpg (115988 bytes)

Fig. 8. RAMS surface temperature forecast valid 1200 UTC 20 December 1998. Note: Temperatures are in �F.

Little forward progress of the arctic front was observed between the 12 and 18-hour forecast runs of RAMS. Although colder air continued to filter into northwest Colorado, the arctic airmass had yet to infiltrate the lower valleys of western Colorado and eastern Utah. Consistent with the colder air slipping south into northwest Colorado, the RAMS 18-hour forecast of surface temperatures did show a tightening isotherm gradient behind the front (Fig. 9). A deficiency noted in the RAMS forecast by this time was a lack of predicted accumulated precipitation in the lower valleys. RAMS gave little hint at the snowfall that accumulated in Grand Junction between 1200 and 1800. Although it will be shown later that the storm total precipitation forecast by RAMS was accurate for the Grand Junction vicinity, the accumulated snowfall for some lower valley cities was under-predicted. It is suggested here, that strong isentropic lift and the possible presence of Conditional Symmetric Instability (CSI) may have led to enhanced snowfall rates (Dunn 1999). The presence of isentropic lift and CSI may have hindered the ability of RAMS to predict the onset of precipitation in the lower valley areas.

KGJX WSR-88D radar images from 2020 (not shown) indicated an enhanced band of precipitation ahead of the arctic front. Surface observations at WFO GJT showed that frontal passage likely occurred between 2044 and 2058. The RAMS model forecast of surface temperature accurately captured the passage of the arctic boundary between 1800 20 December and 0000 21 December (Fig. 10), bringing temperatures below 0�C in Grand Junction and slackening the isotherm gradient north of the ETP and Roan Plateau. RAMS surface wind forecasts valid at 0000 21 December (not shown) also showed northwest winds in the lower valleys around Grand Junction and Moab, continuing up the Interstate 70 corridor to Eagle. Precipitation rate forecasts at this same time indicated the ending of precipitation in northwest Colorado, while the highest rates of up to 0.25cm per hour were depicted along the frontal boundary (not shown). This verified well with radar images and surface observations across the CWA. Water equivalents of total snowfall behind the front were nearly mimicked by the RAMS forecast.

The arctic front moved swiftly during the next 12 hours, traversing the entire CWA by 1200 21 December. Observations from WFO GJT showed that the snow had tapered by 0600 21 December, and that surface winds had switched to the north. Both of these features were captured well by RAMS. Also, the low temperature on 20 December at WFO GJT of almost -8�C (just prior to 0700 21 December) fell within the RAMS 0600 21 December forecast of -7�C to -9�C. Temperatures continued to plummet overnight, eventually enveloping the entire CWA with readings of -7�C to -29�C by 1200 21 December (Fig. 11). Although RAMS forecast temperatures at WFO GJT were too cold at 1200 21 December, the overall trend of continuing to filter colder air south was on track.

The performance of RAMS in forecasting precipitation during the last 12 hours of the event was more varied. Precipitation rate forecasts during this period (not shown) were generally on-target over the higher terrain of southwest Colorado and southern Utah, and for locations behind the arctic front where on-going snow was minimal. However, the model under-forecast the heavy snow that fell in the lower valleys surrounding Montrose and Delta by about one-quarter to one-half.
fig9.jpg (149398 bytes)
Fig. 9.
RAMS surface temperature forecast valid 1800 UTC 20 December 1998. Note: Temperatures are in �F.

fig10.jpg (154579 bytes)

Fig. 10. RAMS surface temperature forecast valid 0000 UTC 21 December 1998. Note: Temperatures are in �F.

fig11.jpg (132542 bytes)

Fig. 11. RAMS surface temperature forecast valid 1200 UTC 21 December 1998. Note: Temperatures are in �F.

It is hypothesized that the quick progression of the front in the RAMS forecast may have accounted for the limited precipitation predicted in this location. Also, it is possible that the north-south orientation of the lower valleys in this region may have played a role in the underestimation of precipitation. In contrast, precipitation over the lower elevations near the Four Corners between 0000 and 1200 21 December was overestimated. In this case, it appears that the extreme terrain gradient south of the San Juan Mountains was slightly smoothed at the 12km horizontal grid spacing by RAMS, possibly causing the model to overestimate precipitation amounts. Storm total precipitation for the entire 36-hour forecast period ending 1200 21 December is shown in Fig. 12. The major topographic features shown in Fig. 2 are easily discernible in the precipitation contours displayed in Fig. 12.
fig12.jpg (143235 bytes)

Fig.12. RAMS 36-hour accumulated total precipitation from 0000 UTC 20 December 1998 to 1200 UTC 21 December 1998. Note: Accumulated precipitation is given in inches.


The performance of the RAMS model that was run at CSU for the 19-21 December 1998 event was vastly superior to that of the NCEP models, specifically in predicting frontal movement, temperatures, and precipitation amounts.

RAMS was the only forecast model available to the operational forecaster that accurately depicted the arctic air surging through the WFO GJT CWA. The timing of the arctic front to WFO GJT was captured well in the surface wind and temperature forecast fields produced by RAMS. Whereas the NCEP models lacked definition, the areal and quantitative distribution of RAMS precipitation forecasts gave the operational forecaster the needed resolution to predict snowfall totals on the mesoscale. Other studies (Papineau et al. 1994) have shown RAMS to depict the areal distribution of precipitation, while exhibiting a tendency to underestimate amounts. For our research, RAMS forecasts of higher terrain precipitation was excellent; however, the model did underestimate snowfall at several lower valley locations. It is noted that the total accumulated precipitation that fell at WFO GJT was nearly exact. Clearly, the RAMS model horizontal fine grid of 12km covering Colorado and surrounding areas, and the high resolution representation of complex terrain, were significant factors in the accuracy of the forecast regionwide.

Local runs of mesoscale models, such as RAMS, are becoming more prevalent in National Weather Service Forecast Offices. The need for accuracy on the mesoscale, particularly in complex terrain, is necessitating that offices develop and implement these local models to better serve the public.


The authors would like to thank the staff at WFO GJT for their help and support of this research. Thanks are also extended to Dr. Doug Wesley of COMET, Dr. Lawrence Dunn of WFO Salt Lake City, and Pat Spoden of WFO Paducah for their insight and reviews. They would also like to express their appreciation to Dr. William Cotton and Dr. Robert Walko, both of Colorado State University, for their comments and suggestions.


Dunn, Lawrence, 1999: Personal communication.

Papineau, John M., R.A. Pielke, and D.A. Wesley, 1994: A Performance evaluation of the NGM and RAMS models for the 29-30 March 1991, Colorado Front Range snowstorm. Natl. Wea. Dig., 19:2, 2-11.

Pielke, R.A., W.R. Cotton, R.L. Walko, C.J. Tremback, W.A. Lyons, L.D. Grasso, M.E. Nicholls, M.D. Moran, D.A. Wesley, T.J. Lee, and J.H. Copeland, 1992: A comprehensive meteorological modeling system RAMS. Meteor. Atmos. Phys., 49, 69-91.

Rhea, J.O., 1978: Orographic precipitation model for hydrometeorological use. Ph.D. dissertation, Atmos. Sci. Paper NO. 287, Colorado State University, Dept. of Atmos. Sci., Fort Collins, CO 199 pp.

Walko, R.L., W.R. Cotton, M.P. Meyers, and J.L. Harrington, 1995: New RAMS cloud microphysics parameterization. Part I: The single-moment scheme. Atmos. Research, 38, 29-62.

Weiland, Michael S., 1994: Heavy snowfall in the southwest corner of Wyoming. Central Region Applied Research Paper NO. 11-3, 27-41.

Wesley, Douglas A., J.F. Weaver, and R.A. Pielke, 1990: Heavy snowfall during an arctic outbreak along the Colorado Front Range. Natl. Wea. Dig., 15:3, 2-19.