Michael J. Fuhs
National Weather Service Forecast Office
Sioux Falls, South Dakota
The Nested Grid Model (NGM) has been an important component within the short-range operational forecasting numerical model suite of the National Weather Service (NWS) since March 1985. Six, 12 and 18 hour NGM low-level moisture flux convergence (MOCON) prognoses are available (since May 1988) and are produced from the 0000 and 1200 UTC NGM model runs.
The purpose of this study is to correlate severe weather events with the predictive information given by the 12 and 18 hour forecast NGM MOCON centers, in terms of false alarms and accuracy. NGM MOCON centers starting at 10 units were collected. Only the 1200 UTC model run was examined since 12 and 18 hour forecasts coincided with the "prime time" of severe weather occurrences. In this manuscript, severe weather includes severe local storms of a convective nature involving hail at least three-quarters of an inch in diameter, wind speeds of at least 50 knots and tornadoes.
Previous studies have been performed involving surface or low level moisture convergence. Hudson (1971) found that moisture convergence tends to precede convective development by several hours. Hirt (1982) showed that persistent moisture convergence centers have a greater chance to be associated with severe thunderstorms. Nierow (1989) and Beckman (1990, 1993) examined NGM MOCON in particular and how it related to the location of severe thunderstorms. Nierow found that a large maximum area of MOCON does not guarantee severe thunderstorm development, but the increase of values with time is usually more important. Beckman's results support this statement, but also indicate that severe weather tends to occur in the largest gradient of MOCON to the east and south of where this quantity is maximized (analogous to the warm sector). This is due to the moisture advection and convergence ahead of a cold front. In addition, Beckman found that a MOCON value of 30 g/kg hr-1 x 10 should flag the forecaster as a significant value for the greater likelihood of severe weather development. The NGM MOCON units of g/kg hr-1 x 10 will hereafter be called "units", which is consistent with Beckman's previous work.
Moisture flux convergence is the gradient of the product of the horizontal wind vector (V) and the mixing ratio (q) and is defined by the equation:
Examining the right side of (1), the first term is the mixing ratio advected by the horizontal wind (moisture advection). The second term is the divergence of the horizontal wind times the mixing ratio (mass divergence). The mass divergence term is usually the largest, but moisture advection can contribute significantly. Bothwell (1988) pointed out that moisture convergence should be called moisture flux convergence when the equation includes both terms. The NGM MOCON values are an average value of the four lowest NGM sigma layers, which relates to approximately 150 mb (or about 1.5 km) above the model terrain.
Sixty-eight severe weather days were collected from March 7 through August 31, 1994. The severe weather events were plotted using SVRPLOT (Hart 1993). Multiple reports of hail or wind events occurring within ten statue miles and 15 minutes of each other and in the same county are recorded as one event.
Geographical boundary conditions were set for the severe weather collection. The Canadian border was used as the northern boundary and the 31° latitude line for the southern extent. From west to east, data collection was bounded by the front ranges of the Rocky and Appalachian mountains, respectively. All NGM MOCON centers that were outside these boundaries were not considered, to eliminate mountain and coastal effects. Mountain topography and land-sea interactions can greatly increase mesoscale lift needed for convection (Heidman and Fritsch 1988).
This study examined NGM MOCON centers and their relationship to severe weather in terms of false alarm ratio (FAR) and probability of detection (POD) for three different classes of severe weather. The first, severe (from now on called SVR), is defined by hail diameter of ¾" to 1.99" or winds of 50 knots to 64 knots. The second class, violent severe (VSVR), was defined by Schaefer et al. (1985) as hail diameter ≥ 2 inches or winds ≥ 65 knots. The third category is simply the tornado (TOR) events. The purpose of splitting the severe weather into three separate categories was to determine if any relationship existed between higher NGM MOCON centers and more damaging severe weather, in terms of both FAR and POD. Wind damage reports were excluded from consideration due to the inconsistent nature of a tree or structural damage.
Analysis of the NGM MOCON centers started at 10 units because past observations revealed that significant severe weather events were associated with this value. Also, considering that Beckman's research (1990, 1993) started at 20 units, a starting point of 10 units presented an opportunity to examine smaller values.
In Table 1, the number of NGM MOCON centers collected from the 68 severe weather days is examined. The spring and summer of 1994 produced a very small number of ≥ 35 unit centers. In Table 1, and the tables to follow, positive signs for the MOCON value show convergence (negative divergence). This is consistent with the NGM MOCON charts.
Number of NGM MOCON centers collected from 68 severe weather days examined. MOCON centers are listed on the left with units of g/kg hr-1 x 10. All MOCON centers were collected from the 1200 UTC model run.
|12 HR MOCON PROG||18 HR MOCON PROG|
The first phase of research calculated FARs to determine the performance of MOCON centers in predicting the occurrence of severe weather. Put simply, did severe weather occur at all?
In this study, a false alarm was defined as no severe weather events occurring within 300 nm of the MOCON center for specific time frames. The time frames used were divided into four three-hour segments; one before, and one after the valid time of the 12 and 18 hour NGM MOCON prognoses (2100 to 2359 UTC and 0000 to 0259 UTC for the 12-hour forecast, 0300 to 0559 UTC and 0600 to 0859 UTC for the 18-hour forecast). The time frames were devised to determine if a significant difference in false alarms occurred as the evening progressed. FAR was calculated by FAR = a/a+b (Beckman 1990, 1993), where a is the number of false alarm events and b the number of severe events that occurred within 300 nm. FARs were calculated for all three severe weather classes and for all time frames starting at 10 units. If no severe weather events occurred within 300 nm, the FAR = 1. Ideal results would indicate a higher FAR for smaller MOCON centers.
Table 2.1 display's the FARs for the 12 and 18 hour NGM MOCON forecasts, for each severe weather class. The FAR values for ≥ 35 units should be used with caution due to the very small sample size. Usually, the FARs are higher for the 18-hour prognosis when compared to the 12-hour. Moreover, the FARs are higher for VSVR and TOR events than for SVR episodes. Except the 0600 to 0859 UTC time-frame, the FARs showed that most of the time there is a greater than 50 percent chance that at least one SVR event will occur within 300 nm of a MOCON center (even for 10 unit centers).
These findings suggest that NGM MOCON, by itself, is not a good predictor for the intensity of severe weather. However, the FARs strongly suggest that NGM MOCON is a good predictor that some type of severe weather (SVR) will occur within 300 nm of the center, even at 10 units.
Table 2.2 shows the FARs for the 12 and 18 hour NGM MOCON forecasts with all three severe weather classes combined. These values display a trend of decreasing FARs as the MOCON center increases for all time frames. Once again, the FARs are higher for the 18-hour forecast than for the 12-hour. Except most of the 0600 to 0859 UTC time-frame, much of the FARs in this table predict a greater than 50 percent chance that severe weather will occur within 300 nm of the NGM MOCON center. Most important, the 30 unit maximums display very low FARs for all four time frames.
FARs for the 12 and 18 hour NGM MOCON prognoses, for each severe weather class and each selected time period. All times are in UTC. MOCON centers listed on the left have units of g/kg hr-1 x 10. No 40 unit centers were collected from the 18-hour forecast (nc = not collected.)
|12 HR PROG||18 HR PROG|
FARs for all three severe weather classes combined. All times are in UTC. MOCON centers listed on the left have units of g/kg hr-1 x 10. No 40 unit centers were collected from the 18-hour forecast (nc = not collected.)
|12 HR PROG||18 HR PROG|
The second phase of this work examined only the MOCON centers which had at least one severe weather occurrence within 300 nm of the center. This phase determined the detection rate (accuracy) of severe weather for each individual MOCON center by calculating POD ratios. POD ratios were calculated by POD = x/y (Beckman 1990, 1993), where x is the number of severe weather occurrences within a given radius from the MOCON center and y the total number of occurrences within 300 nm of the center. A POD of one show that all severe weather events that occurred in a given radius were "predicted" by MOCON. PODs were analyzed for all three classes of severe weather and for all four time-frames.
The radii chosen for analyses were 60 and 180 nm outward from a MOCON center because these distances may lie within a county warning area (CWA). Any severe weather event beyond 300 nm from a MOCON center is of little value to a forecaster due to the large distance in relation to the size of a CWA. More emphasis should be placed on the 60 and 180 nm radii.
Table 3.1 display's POD ratios for each time-frame and MOCON center, for the 180 nm radius. Many undefined (u) values exist for ≥ 35 units due to the very small sample of MOCON centers collected. POD values for the 180 nm radius show some detection consistency for all four time frames, for each MOCON center. Also, a fair of amount consistency exists among the three classes of severe weather. Except the 10 unit MOCON center, most of the 12 hour maxima display detection greater than 50 percent. This was not true for the 18 hour maxima when the centers were less than 25 units. This means that when examining all severe weather occurrences, most severe weather occurred within the nearest 180 nm of the 300 nm limit imposed in this research, for the 12-hour MOCON centers ≥15 units. That was also true for the 25 and 30 unit 18 hour MOCON centers. In particular, detection is the highest for the 30 and 40 unit MOCON centers within the 2100 to 2359 UTC period. If more 35 unit centers were collected, it is possible they would show the same trend. This collaborates the statement made by Beckman that a value of 30 units should alert the forecaster that a very good probability exists that some type of severe weather could occur fairly close to these maxima.
The 60 nm radius is not presented, but can be summarized by stating that the POD values are far less than the values determined at 180 nm (less than 25 percent). As with the 180 nm radius, there is some consistency in POD ratio values between the three classes of severe weather within each time-frame, for each MOCON center.
POD ratios for severe weather occurring within 180 nm of the 12 and 18 hour NGM MOCON centers, for each selected time-period. All times are in UTC. MOCON centers listed on the left have units of g/kg hr-1 x 10. No 40 unit centers were collected for the 18-hour forecast (nc = not collected.) The letter (u) indicates an undefined value where no severe weather events occurred within 300 nm. This is largely due to the small number of ≥35 unit MOCON centers collected.
|12 HR PROG||18 HR PROG|
Table 3.2 displays the POD results for the 60 and 180 nm radii with the three severe weather classes combined. The 35 and 40 unit MOCON maxima POD values should be used with caution due to the very small sample size. The same trends exist here as when the severe weather classes are kept separate. Severe weather detection was poor within the 60 nm radius. The major item which stands out is the high POD ratios for the 30 unit MOCON centers within the 180 nm radius. These results reaffirm Beckman's findings that 30 units could be a significant value. With more data samples, the 35 and 40 unit centers could show the same trend.
In previous research with NGM MOCON, Beckman found many severe weather events occurring on the east and south sides of the MOCON centers, analogous to the warm sector. Beckman's findings, combined with the results from this study, suggest that forecasters should focus their attention on the warm sector of the 180 nm radius from a NGM MOCON center, especially for values of at least 30 units.
POD ratios for all classes of severe weather combined, occurring within 60 and 180 nm of the 12 and 18 hour NGM MOCON centers, for each selected time-period. All times are in UTC. MOCON centers listed on the left have units of g/kg hr- x 10-1. No 40 unit centers were collected for the 18-hour forecast (nc = not collected.) The letter (u) indicates an undefined value where no severe weather events occurred within 300 nm. This is largely due to the small number of 35 unit MOCON centers collected.
|POD RATIOS FOR ALL SEVERE WEATHER CLASSES COMBINED WITHIN 60 NM|
|12 HR PROG||18 HR PROG|
|POD RATIOS FOR ALL SEVERE WEATHER CLASSES COMBINED WITHIN 180 NM|
|12 HR PROG||18 HR PROG|
NGM MOCON was collected from 68 severe weather days from March 7 through August 31, 1994, starting at 10 units and from the 12 UTC model run. The most significant results from this work are the following:
Future research with NGM MOCON could involve combining this parameter with such fields as NGM synoptic scale lift (vertical velocity) or NGM lifted indices. For example, during this study the 12-hour NGM MOCON forecast was combined with the 12-hour NGM vertical velocity prognosis by simply overlaying (and adding) the two fields. Most often, both fields had a maximum value very near each other. A single 6-hour time-frame was briefly analyzed from 2100 to 0259 UTC. Preliminary results indicated that synoptic scale lift did not aid MOCON in the prediction or detection of any type of severe weather. More in-depth work could verify this.
The author wishes to extend appreciation to Cindy Fay, Met Intern, NWSO, Hastings, Nebraska, for assisting with collection and plotting of the data. He would also like to extend his gratitude to Ron Holmes, SOO, NWSFO, Sioux Falls, South Dakota for the editing of this manuscript. In addition, he would also like to thank Cindy Fuhs, programmer/analyst, Hughes STX Corp., for her expertise and assistance in programming the C software.
Beckman, S.K., 1993: Preliminary results of a study on NGM low-level moisture flux convergence and the location of severe thunderstorms. Preprints, 17th Conf. on Severe Local Storms, St. Louis, MO, AMS (Boston), 138-142. ____________, 1990: A study of 12-h NGM low-level moisture flux convergence centers and the location of severe thunderstorms and heavy rain. Preprints, 16th Conf. on Severe Local Storms, Kananaskis Park, Alta., Canada, AMS (Boston), 78-83. Bothwell, P.D., 1988: Forecasting Convection with the AFOS Data Analysis Programs (ADAP-Version 2.0). NOAA Technical Memorandum NWS SR-122, Scientific Services Division, Southern Region, Fort Worth, TX, 92pp. Hart, J.A., 1993: SVRPLOT: A new method of accessing and manipulating the NSSFC severe weather database. Preprints, 17th Conf. on Severe Local Storms, St. Louis, MO, AMS (Boston), 40-41. Hirt, W.D., 1982: Short-term prediction of convective development using dew-point convergence. Preprints, 9th Conference on Weather Forecasting and Analysis and Aviation Meteorology, Seattle, WA, AMS (Boston), 205-210. Hudson, H.R., 1971: On the relationship between horizontal moisture convergence and convective cloud formation. J. Appl. Meteor., 10, 755-762. Nierow, A., 1989: An evaluation of the NGM's low-level moisture convergence forecasts in predicting heavy precipitation and severe weather. Preprints, 12 Conf. on Weather Analysis and Forecasting, Monterey, CA, AMS (Boston), 626-630. Heideman, K.F., and J.M. Fritsch, 1988: Forcing mechanisms and other characteristics of significant summertime precipitation. Wea. and Forecasting, 3, 115-130. Schaefer, J.T., D.L. Kelly and C.A. Doswell III, 1985: The annual progression of non-tornadic severe thunderstorm events. Preprints, 14th Conf. on Severe Local Storms, Indianapolis, IN, AMS (Boston), 5-8.