CRH SSD
OCTOBER 1997

CENTRAL REGION TECHNICAL ATTACHMENT 97-06

 

Effects of Snow Cover on Errors in NGM MOS Minimum Temperature Guidance for Bismarck, North Dakota

 

Viggo E. Jensen III
National Weather Service Forecast Office
Bismarck, North Dakota

 

1. INTRODUCTION

Forecasting minimum temperatures has always been a challenge for forecasters. In Bismarck, North Dakota, forecasters have always tested their skills in trying to predict nighttime low temperatures using the Nested Grid Model (NGM) Model Output Statistics (MOS) for guidance. In North Dakota during the winter, frequent outbreaks of cold arctic air stream southward from Canada into the northern Plains. The shallow layer of very cold, arctic air near the surface tends to collect in the "valley" areas of Bismarck, that leads too much lower temperatures in the morning. With the influence of snow cover, these temperatures can go much lower. It has been documented that the NGM model guidance temperature forecast has warm bias especially when snow cover is present. In this paper, observed data from a winter season is compared to the NGM MOS forecasts of minimum temperatures in Bismarck versus snow depth.

2. DATA

Data for the period from November 1995 through March 1996 was collected (134 days). Observed and forecast low temperatures, observed wind speed, sky conditions, and snow cover at the time of the minimum temperature were assembled. Also, collected was NGM MOS guidance of forecast low temperature, surface wind speed and forecast sky conditions for 1200 UTC.

3. NGM MOS EQUATIONS

The NGM MOS equation is typically derived from a variety of meteorological factors that vary by location. For Bismarck, the wintertime MOS equation is weighted mostly on 2 m surface temperatures, 950 mb dew point temperature, and 10 m surface wind speed forecasts. Maglaras and Carter (1986) show that snow cover is typically considered in the 6-15 hour temperature/dew point predictors in most winter MOS equations. This appears to be the case in Bismarck's wintertime MOS equation. Most MOS equations are usually weighted heavily on thickness forecasts. In the case of shallow surface-based cold air masses in the winter, thickness becomes a poor representation of surface temperatures, as stated by Wiesmueller and Kaplan (1989). For a given MOS temperature forecast, often it is not until the second day of a cold air mass intrusion that the NGM model gets a handle on Bismarck's low temperature. This will be shown later in this paper. McCann (1977) suggests that the radiational cooling typical of the second night has more of an effect on the MOS forecast than the first night.

Other factors in the MOS equation are temperature advection, precipitation amount, precipitable water, relative humidity, vertical velocity, wind speed at the surface and aloft, stability and sky condition. The NGM MOS has often had a warm bias towards minimum temperature forecasting in the winter season in most parts of the country. This seems to be the case in Bismarck, North Dakota. Snow depth appears to introduce error when using MOS guidance, as the minimum temperature often goes much lower than forecast.

4. ANALYSIS

Table 1 shows a compilation of the data collected for this winter study versus snow cover and other factors. As shown, the snow-cover ranges from no snow on the ground up to 9 inches. At first glance, it is apparent that MOS did fairly well with bare ground. However, as the snow cover increased to 5 inches or more, MOS exhibited a warm bias. According to this table, the forecasters made the necessary adjustments to MOS to predict the minimum temperature when snow was on the ground. In Bismarck, snow cover can accumulate as early as late October and, in most cases, last through most of March. The average error between MOS and the observed minimum temperature was -4.1°F (i.e., the observed minimum was lower than MOS forecast), when greater than 6 inches of snow was on the ground. Jacks and Rao (1985) state that this is a typical error especially when extreme temperatures occur. It appears that MOS tended to predict mostly an overcast condition when half the time a mix of conditions was observed. MOS and the observed wind were fairly close most of the time.

TABLE 1

Comparison of Wintertime data versus MOS guidance with snow on the ground ranging from no snow up to 9 inches. Data compiled for the Winter from November 1995 through March 1996 for WSFO Bismarck, North Dakota. (Temperatures are in degrees F and wind speed is in miles per hour).

 

 
 
Snow
Dpths
 
12Z
Obs
Low
 
12Z
Fcst
Low
 
12Z
NGM
MOS
MOS
Wind
Fcst
Spd
 
Obs
Wind
Spd
MOS
Fcst
(Days
Clr
 
 
 
Sct
 
Sky
of)
Bkn
 
 
 
Ovc
 
Obs
Days
Clr
 
 
 
Sct
 
Sky
of
Bkn
 
 
 
Ovc
No Snow 20.8 18.0 17.7 7.36 7.58 6 6 8 18 8 4 10 16

 

1"
17.8 14.7 15.5 11.7 11.3 6 3 1 5 3 2 3 7

 

2"
13.3 15.5 13.2 7.09 7.64 2 0 3 6 2 1 1 7

 

3"
28.3 22.3 18.3 12.0 7.00 0 1 0 2 0 1 1 1

 

4"
10.1 9.64 10.5 5.43 5.86 2 3 3 6 2 2 2 8

 

5"
1.73 -0.07 -0.67 6.73 5.93 2 2 0 11 2 4 1 8

 

6"
-0.25 -2.75 0.25 8.17 7.75 1 1 1 9 2 2 3 5

 

7"
-12.5 -13.0 -8.33 5.50 5.33 0 0 3 3 1 3 2 0

 

8"
17.00 10.0 14.0 14.3 7.67 1 0 1 1 0 1 1 1

 

9"
-19.7 -20.3 -15.1 8.41 7.18 3 6 4 4 8 3 4 2

 

Table 2 shows averages for various ranges of snow cover. Overall, MOS does a good job of showing a trend in the forecast. However, averages show a marked warm bias in MOS when snow depths of 5 inches or greater were observed. An average error of -2.8°F can be seen in this case. When looking at the observed sky and surface wind conditions, a partly cloudy sky with relatively light winds (under 10 mph) occurred with this average error. This is typical in a radiational cooling event anywhere, though it seems to happen quite often in Bismarck in the winter season. Official observations for Bismarck are reported at the airport which lies in the Missouri River Valley. Other parts of town, especially in the north, have rising terrain which can certainly affect the temperatures (McMullen 1991).

TABLE 2

Overall averages of observed low temperatures versus MOS guidance with snow on the ground for WSFO Bismarck, North Dakota. Also, shown is the MOS forecast surface wind, the observed surface wind, and observed sky conditions for the 1995-1996 Winter season.

Snow
on
Grnd
Obs
Lows
(deg F)
Fcst
Lows
(deg F)
MOS
Guid
(deg F)
Fcst
Wind
(mph)
Obs
Wind
(mph)
Obs Sky
(# of days)
clr
sct
bkn
ovc
Ave.
Ttls
6.67
5.41
6.54
8.68
7.31
28
22
228
255
Ave.
0-4"
18.08
16.03
15.05
8.72
7.84
15
10
17
39
Ave.
5-9"
-4.75
-5.22
-1.97
8.63
6.77
13
13
11
16
5. RESULTS

Several graphs were created to better understand how the NGM MOS did during this winter period versus snow depth. Figure 1 displays the 38 cases where no snow was on the ground. As can be seen, MOS and the observed low temperatures were reasonably close. In 15 cases with at least 5 inches of snow depth, MOS had difficulty handling the larger variation of low temperatures. Figure 2 displays this variation. There were two events where low temperatures dropped significantly. Cases #5-#7 illustrate how MOS was able to forecast the minimum well the first day (#6), but the second day (#7) MOS showed a warm bias as minimum temperatures were some degrees colder. The other event showing bitter cold temperatures occurred with cases #8-10. MOS and the observed minimum were fairly close the first night, but the second night MOS had a warm bias (#10) as the low temperature dipped to -26°F. The MOS forecast was about 19 degrees warmer than the observed temperature. Figures 1 and 2 suggest that because there is more snow cover with lower temperatures, and given the usual light winds and partly cloudy sky, that MOS guidance has trouble handling very cold events.

Figure 1. Displays 12Z NGM MOS minimum temperature forecast versus the 12Z observed minimum temperature for 38 cases with no snow on the ground. For the Winter season from November 1995 through March 1996.

Figure 2. Same as Figure 1 but for 16 cases with 5 inches of snow depth.

Figure 3. Same as Figure 1 but for 18 cases with 9 inches of snow depth.

Figure 3 shows MOS versus low temperatures, but with 9 inches of snow on the ground. Cases #1-6 continue to demonstrate a MOS warm bias when very cold temperatures are observed. The remaining cases show smaller discrepancies, but still some warm bias in MOS.

6. CONCLUSIONS

It has been shown in this preliminary study that the effects of snow cover may not be adequately reflected in the NGM MOS minimum temperature guidance for Bismarck, North Dakota. As the snow depth increases with increasingly cold nighttime temperatures, MOS has a strong warm bias. The forecaster is, in most cases, having to adjust the forecast minimum temperatures lower than guidance suggests. When no snow or a minimal snow depth occurred, MOS was fairly close to observed morning minimums. An average error of 4°F or 36 percent, was observed when greater than 6 inches of snow depth occurred between MOS and observed minimums for these situations.

More work needs to be done with studying how to make the NGM MOS temperature guidance better. Snow depth appears to be an important factor, as well as the existence of snow cover. An adjustment factor to apply to MOS to better predict low temperatures in Bismarck, North Dakota appears necessary. In addition, having an awareness of the biases in the MOS minimum temperature forecast when there is snow cover will help the forecasters in Bismarck.

7. ACKNOWLEDGEMENTS

The author would like to thank Mr. Sam Walker for his help with figures and other staff at WSFO Bismarck, North Dakota for reviewing this manuscript and giving helpful comments.

8. REFERENCES

Jacks, E., and S.T. Rao, 1985: An Examination of MOS Objective Temperature Prediction Model. Mon. Wea. Rev, 113, 1, 134-148.

Maglaras, G.J., and G.M. Carter, 1986: How to Use MOS Guidance Effectively. Preprints 11th Conference on Weather Forecasting and Analysis, Kansas City, MO, AMS (Boston), 17-22.

McCann, D.W., 1977: An Analysis of Minimum Temperature After Cold Frontal Passage. Central Region Technical Attachment 77-5. DOC, NOAA, NWS, Central Region Scientific Services Division, Kansas City, MO, 2 pp.

McMullen, K.W., 1991: Comparative Lows for Northern Bismarck and WSFO Bismarck. Central Region Technical Attachment 91-20. DOC, NOAA, NWS, Central Region Scientific Services Division, Kansas City, MO, 8 pp.

Wiesmueller, J.L., and J.A. Kaplan, 1989: Mid-Winter Bias of LFM-Based MOS and NGM Perfect Prog Temperature Forecast. Central Region Technical Attachment 89-6, DOC, NOAA, NWS, Central Region Scientific Services Division, Kansas City, MO, 6 pp.

 


USA.gov is the U.S. government's official web portal to all federal, state and local government web resources and services.