Re-analysis of the Gridded Model Output to Forecast the South Dakota Heavy Snow Event of March 4, 1995


Scott M. Reiter and Troy D. Kleffman
National Weather Service Office
Aberdeen, South Dakota



During the early morning hours of March 4, 1995, heavy snow fell over southwest South Dakota and spread rapidly northeast. As a result, at 850 AM CST, a Winter Storm Warning was issued for northeast South Dakota. Much of the day, moderate to heavy snow occurred over northeast South Dakota, it began to taper off by late afternoon. This winter storm produced 8-12 inches of snow (Figure 1) over northeast South Dakota. Twenty-four hours prior to the event, the forecast indicated that 1-4 inches of snow was anticipated. However, re-analysis of the gridded model output, from the 0000 UTC initial conditions suggested that heavy snow was possible (Heavy snow meaning six or more inches in a 12-hours) This was more than 36 hours before the event!

Figure 1. Total snowfall (in) isopleths from 1200Z 4 March to 1200Z 5 March for eastern South Dakota.

Since the models (AVN, ETA, and NGM) were in close observational agreement, 36-hours in advance, the NGM Gridded model output from the 0000 UTC run on March 3 was used in our re-analysis. Diagnosis of Q-vectors, isentropic lift, moisture advection, and jet dynamics, all suggested a significant snow event was possible over northeast South Dakota. This study focused on using these gridded model output fields to evaluate the numerically simulated dynamics which may have translated to improved predictability of this snow event.


A. Surface and Moisture Fields

The first clue in determining the heavy snow potential came from the location of the surface front. At 1200Z on March 3, an Arctic front propaged southeast over North Dakota and Montana (Note: the surface temperatures over northeast South Dakota were in the single digits above zero Fahrenheit). Gridded model output indicated the front would stall along the South Dakota-Nebraska border by 1200Z on March 4 (Figure 2). Also by 1200Z, cyclogenesis was forecast along the stationary front over southeast Wyoming and the Nebraska panhandle, in advance of an approaching trough. The developing surface low was progged to produce a 25-to-30 knot low level jet, thus advecting gulf moisture into and over the stationary front. This moisture advection shows up readily in the precipitable water forecasts which called for an increase from about 0.15 inches to around 0.40 inches over northeast South Dakota during the period from 1200Z on March 3 to 1200Z on March 4 (Figure 3). Therefore, it suggested there would be a sufficient supply of moisture available.

Figure 2. 36-hour NGM surface prog. (valid 1200Z 3 March). Solid lines (MSL Pressure), wind barbs (Sigma layer S982). Surface front has been subjectively added.

Figure 3. 36-hour NGM precipitable water prog. (Valid 1200Z 4 March). Contours are labeled in tenths of an inch.

B. Shortwave Trough

At the 700 mb level, 12-hour predicted heights indicated an ill-defined trough moving out of the western U.S. and by 36 hours (Figure 4), moving into the western Dakotas and Nebraska. Also at 36 hours, height field values indicated the trough becoming better defined. After examining the predicted thermal fields at 700 mb, areas of baroclinicity were easier to find which helped better define the existence of the trough in question 36 hours prior to the event. Remember, 700 mb thermal fields are only available at 00 hr via AFOS. Therefore, the 700 mb predicted thermal fields proved quite useful for this event.

Figure 4. 36-hour NGM (valid 1200Z 4 March) 700 mb heights in 10s of meters (solid lines), and isotherms, degree C (dashed lines). A trough axis is located from northeast Montana to the north Texas panhandle.

At 24-hrs. (0000Z 4 March ) the models indicated that northeast South Dakota would be under NVA (negative vorticity advection) at 500 mb. By 36-hrs. (Figure 5), a relatively weak wave is depicted at 500 mb from western North Dakota to the Texas panhandle with a stronger vorticity maximum over eastern Utah.

Figure 5. 36-hour NGM (valid 1200Z 4 March) 500 mb heights in 10s of meters (solid lines) and absolute vorticity in 10-5 Rad sec-1 (dashed lines).


A. Vertical Motion

According to Figure 4 the significant warm air advection was expected to take place over northeast South Dakota at 36-hrs. Also, the gridded model output suggested vorticity would be decreasing with height from 850-500mb (not shown). The Omega equation states that upward vertical motion is defined by: (1.) differential vorticity advection in a vertical layer, (2.) warm air advection in a layer (Holton 1992). This suggests these two terms could potentially cancel each other. Therefore, other techniques may be useful to ascertain the nature and forcing of the upward motion.

B. Q-Vectors

Another advantage of PCGRIDDS is the ability to look at Q-Vectors. "Q-Vectors graphically show the synoptic-scale vertical motion resulting from both forcing terms of the Omega equation: namely differential vorticity advection and the Laplacian of the thickness advection. Convergence of Q implies synoptic-scale upward vertical motion, and the divergence of Q implies synoptic scale downward vertical motion" (Chaston 1994). The gridded model output indicated converging Q (dashed lines) along the leading edge of the 700 mb trough (Figures 6-7). Therefore, the model forecasted Q-vectors helped narrow the location and timing of the strongest (implied) upward motion and potentially the most significant precipitation. Indeed, in evaluating a low level warm air advection, over the stationary boundary (Figure 4) The diagnosed amount of this vertical motion (not shown) was the more significant contributor to the upward vertical motion than was contributed by differential vorticity advection.

Figure 6. 24-hour NGM (valid 0000Z 4 March) 700 mb Q-vector convergence. Dashed lines indicate converging Q.

Figure 7. 36-hour NGM (valid 1200Z 4 March) 700 mb Q-vectors. Dashed lines indicate converging Q.

C. Isentropic Analysis

Isentropic analysis provided even more clues to the possibility of heavy snow. Looking at the 288 Kelvin surface in Figure 8, good isentropic lift was indicated over eastern South Dakota during the day on March 4. The winds at the 288K surface over the "area of concern" were from the southwest at 15 to 25 knots (Figure 8). The cross section depicted in Figure 9 shows theta-E surfaces and mixing ratios. The isentropes are sloping upward, from the warmer air over northern Nebraska, to the much colder air over northeastern South Dakota. Mixing ratio values between 2.5 and 3.0 gKg-1 were expected over northeast South Dakota from 1200Z 4 March to 0000Z 5 March. Garcia (1994), correlated isentropic mixing ratios to the amount of snowfall produced. Using his technique, from 4-6 inches of snow might have been forecast over eastern South Dakota during that 12-hour period.

Figure 8. 36-hour NGM (valid 1200Z 4 March) 288K wind barbs (kts.), pressure (hundreds of mb), and pressure advection. Maximum isentropic lift is advection. Maximum isentropic lift is located near Huron, South Dakota.

Figure 9. 36-hour NGM (valid 1200Z 4 March) cross section from O'Neill, NE to Jamestown, ND of e (solid lines) and mixing ratio (dashed lines). Aberdeen is at 45N 98W.

D. Jet Stream Analysis

At the 300 mb level, data at 36 hours showed a 70-knot jet max over east central Colorado with another stronger jet max of 90 knots over Southern Ontario (Figure 10). This appeared to place northeast South Dakota in the left front quadrant of the Colorado jet and right rear quadrant of the 90 Knot Ontario jet max. This puts the region in a favorable area of enhanced upward motion related to coupled jet streaks (Uccellini et al. 1979). The gridded model output indicated that ageostrophic divergence was forecast to occur over northeast South Dakota. The forecast ageostrophic wind vectors and ageostrophic divergence contours in Figure 10 possibly indicated that an ageostrophic secondary circulation would develop. The cross section in Figure 11 further adds credence to this with a forecast maximum upward vertical motion over the northeast quarter of the state at 1200Z 4 March (In addition to the strong cold front and good low level warm air advection over and north of the boundary as shown in Figure 2.).

Figure 10. 36-hour NGM (valid 1200Z 4 March) 300 mb isotachs in 10s of kts (solid lines), ageostrophic wind vectors, and divergence of the ageostrophic wind (dashed lines).

Figure 11. 36-hour NGM (valid 1200Z 4 March) ageostrophic secondary circulation in the plane of the cross section. Dashed lines indicate upward motion. Cross section is from Valentine, NE to Detroit Lakes, MN. Aberdeen is at 45N 98W.


Meteorological analysis of the gridded model data could have assisted with predicting this significant snowfall event through identifying (1) a strong moisture advection uplift of this moist air over a very shallow layer of Arctic air; (2) converging Q; (3) strong low level warm air advection; (4) isentropic up glide and moisture transport; and, (5) enhanced large scale upward motion due to an ageostrophic secondary circulation.

Twenty-four hours prior to the event, the official forecast indicated that 1-4 inches of snow was anticipated. Re-analyzing the gridded output with PCGRIDDS suggested a strong potential for a heavy snow forecast.

Since the gridded fields are output from the NCEP numerical models, they are only as reliable as the models themselves. Therefore, it is up to the forecaster to determine his/her confidence in the model output through the forecast process. Nevertheless, the gridded model output provides much more information than has been previously available to the National Weather Service forecaster. The information provided and analyzed using PCGRIDDS portrays more and different views of atmospheric variables and various meteorological concepts than ever before. Ultimately, it is up to the forecaster to be familiar with the assumptions and limitations of these data and when they can be applied. While it is often difficult to predict where heavy snow will occur 36 hours in advance, we have shown that a careful, proper analysis of the gridded output can better identify and evaluate key meteorological precesses. This can lead to an improved understanding and application of the forecast process and show any areas with high snowfall potential.


We would like to thank Hector Guerrero, William Tallman, Bill Nichols, and Jim Johnson for their technical assistance and guidance.


Chaston, P., 1994: Omega Diagnostics, Including Q-Vectors. Graphical Guidance 1994. DOC, NOAA, NWS, Training Center, Kansas City, Missouri, 176pp.

Garcia, C., 1994: Forecasting Snowfall Using Mixing Ratios on An Isentropic Surface. NOAA Technical Memorandum NWS CR-105, DOC, NOAA, NWS Central Region, Kansas City, MO, 31pp.

Holton, J., 1992: An Introduction to Dynamic Meteorology. Third Edition. Academic Press. San Diego, 511pp.

Keyser D., and M.A. Shapiro, 1985: A Review of the Structure and Dynamics of Upper-Level Frontal Zones. Mon. Wea. Rev., 114, 452-499.

Moore, J., 1993: Isentropic Analysis and Interpretation, Operational Applications to Synoptic and Mesoscale Forecast Problems. DOC, NOAA, NWS, Training Center, Kansas City, Missouri, 99pp.

Uccellini, L., and D. Johnson, 1979: The Coupling of Upper and Lower Tropospheric Jet Streaks and Implications for the Development of Severe Convective Storms. Mon. Wea. Rev., 107, 682-703. is the U.S. government's official web portal to all federal, state and local government web resources and services.