Extended Abstract (624K)5.3a
Mesoscale model simulations in Quasi-forecast mode of the great western storm of 16–20 March 2003 (Formerly Paper Number 10.3)
Douglas A. Wesley, UCAR/COMET, Boulder, CO; and G. Poulos, J. Snook, E. Szoke, M. P. Meyers, G. Byrd, R. A. Rozumalski, and H. McIntyre
A massive snowstorm crippled large portions of the central Rockies and adjacent plains during the period 16-20 March 2003. Snowfall accumulation in the foothills and mountains exceeded four feet in relatively large regions, while on the plains amounts above two feet were common. The large impacts of this historic storm are well documented. This paper examines meso-gamma scale model simulations of the event, utilizing larger model-generated boundary conditions, from a forecasting standpoint.
Public forecasts of this event were generally accurate up to several days before the storm hit. NCEP model guidance provided initial alarms (in the form of ensemble forecasts) up to one week prior to the storm. As the potential event got closer, Eta model forecasts were trending towards a large precipitation event, and by about two days before the onset of snowfall along Colorado’s Front Range very large precipitation totals (five or more inches) were output by this model for portions of the region during the period of 17-20 March. Accuracy of these forecasts was perhaps unprecedented in the area, for such a large event, primarily because the orographic forcing was so strong. The Eta forecasts clearly provided a crucial asset towards forecast operations prior to the storm. The model, however, did show some shortcomings regarding the precipitation type distribution, and of course was limited by its large grid spacing, a required feature given the domain size of that model.
The crippling nature of the subsequent storm period, in terms of disrupting transportation and other day-to-day activities, has shown that even if a very large snowfall potential is emphasized in, say, a 2-4 day forecast, society is still vulnerable to this type of storm. Insurance claims and a paralyzed international airport attest to this fact. Importantly, the current challenge is to increase the resolution and details of the forecast to minimize this vulnerability, as much as currently possible.
Close examination of snowfall totals revealed extremely sharp gradients in snowfall, on the order of several feet within a horizontal distance of 15 miles or less. Many of these sharp gradient regions coincided with strong gradients in elevation; however some did not. For example, an area on the plains/foothills interface just north of Denver accumulated only 3-6 inches of wet snowfall, while 15-25 miles to the south, 24-36 inches fell, and areas another 20 miles to the south recorded nearly four feet. Meanwhile, 20-30 miles north of the aforementioned area of snowfall minimum, 24-36 inches fell. All of these locations are at the same approximate elevation. The current configuration of NWS forecast zones along the urban corridor is not designed to handle these types of gradients, nor is the current configuration of the Eta model.
The purpose of this study is to closely examine the causes of extreme snowfall and wind variations in this storm from a mesoscale modeling standpoint in order to better predict them in the future. Both the “workstation” Eta and MM5 were run in forecast mode utilizing non-hydrostatic and multiple-grid configurations, with the smallest grids exhibiting 1-2 km horizontal grid spacings. The primary reason for utilizing such a small grid spacing is the presence of steep and variable topography throughout the foothills and higher terrain of the Front Range
The role of the barrier jet in the storm in producing, first, snow instead of rain in the urban corridor, and, second, uplift strong enough to produce snowfall rates of 1-3 inches per hour for 2-3 days, cannot be overemphasized. Clearly the barrier jet was located on the cold side of a persistent rain/snow boundary that exhibited the classic characteristics of strongly diabatically-forced mesoscale dynamics, a feature documented in previous heavy springtime snowfalls in the urban corridor. Furthermore, the three-dimensional configuration of this barrier jet is examined closely in this study to attempt to explain the astounding snowfall and wind gradients along the urban corridor.
Preliminary indications are that both mesoscale models produce generally accurate precipitation distributions, and both produce cooler (but still above freezing) low-level conditions along the urban corridor for much of the storm evolution when compared to the operational Eta forecasts. The MM5 forecasts appear to capture better detail in the precipitation distributions, as expected, and exhibit low-level temperatures closer to freezing in critical areas near the rain/snow line. Comparisons with operational profiler winds show some problems with the strength of the mid-level upslope, a critical component of the storm, and one perhaps related to the relatively warm low-level conditions along the urban corridor. This component is likely a primary factor in determining precipitation rates, in the sense of the warm conveyor belt running up and over the barrier jet, and thus a critical determinant of surface precipitation type.
It appears that an accurate prediction of the depth of the barrier jet is a crucial requirement to an accurate precipitation forecast. Another important feature of the mid-level easterly flow is its strong variations through the 3-4 day period as synoptic waves passed through the region, and these variations will be compared to the barrier jet depth and distributions of precipitation rates. Additionally, relatively subtle terrain features along the plains/foothills interface interacted with the barrier jet to contribute significantly to low-level vertical motion fields, and likely played a role in the cause of the snow minima discussed above.
These features of the simulations will be discussed and compared to storm observations at the conference.
Session 5, Case Studies (ROOM 605/606)
Monday, 12 January 2004, 1:30 PM-2:30 PM, Room 605/606
Previous paper Next paper
Recorded presentation