SR SSD 99-24
8-1-99

Technical Attachment

Using Climatologically Favored Thicknesses in Forecasting Heavy Precipitation

Mark A. Rose, Timothy W. Troutman, and Susan K. Ware
NWSO Nashville, TN

1. Introduction

A study was conducted at the National Weather Service Office (NWSO) in Nashville, Tennessee in which climatologically favored 1000-500 millibar (mb) thicknesses were computed for a thirty-year period (January 1, 1967 through December 31, 1996), encompassing 305 cases. Specifically, the relationship between heavy precipitation and climatologically favored thicknesses was examined. Here, heavy precipitation is defined as one inch or more of water equivalent falling during a twelve-hour period.

Mean (or "climatologically favored") thicknesses and confidence intervals were calculated for each month. The results were then used during a two-year trial period (January 1, 1997 through December 31, 1998), in which climatologically favored thicknesses were applied to 23 heavy precipitation events. It was found that climatologically favored thicknesses correctly identified 57% of all heavy precipitation events, and 67% of heavy precipitation events occurring between March and September.

Heavy rainfall correlates favorably with a relatively narrow range of 1000-500 mb thickness values (Bohl and Junker 1987). These values vary by season and location. Climatologically favored thicknesses are least helpful during winter, when strong thickness gradients often exist and heavy rain is therefore spread across a large range of values. Conversely, this method is most helpful during summer when much weaker thickness gradients exist and heavy rain is distributed across a much smaller range of values.

Bohl and Junker (1987) state that quantitative precipitation forecasters often rely on empirical techniques to locate areas of potentially heavy precipitation because of the limitations of numerical models, especially on smaller scales. In this study, mean climatologically favored thicknesses are presented for each month, with confidence intervals used to help draw conclusions about population averages. Specific forecast applications are also explained.

2. Methodology and Results

Sample cases were gathered between January 1, 1967 and December 31, 1996. Twelve hour periods (0000-1200 or 1200-2400 CST) in which at least one inch of precipitation (water equivalent) measured at Nashville International Airport (BNA) were used. Corresponding 1200 or 0000 UTC (0600 or 1800 CST) sounding data were gathered for each case (Forecast Systems Laboratory 1994, 1996), and 1000-500 mb thicknesses were calculated.

The cases were subsequently sorted by month so that mean monthly thicknesses and confidence intervals could be calculated. It was decided to use confidence intervals in order to make broad-based conclusions regarding population averages based on the limited sample data of this study. Here, 99% confidence intervals were calculated using the following equation from Newmark (1992):

L = X_mean - t (s/n^1/2)
R = X_mean + t (s/n^1/2),

where "L" is the lower limit of the interval, "R" represents the upper limit of the interval, "X_mean" is the mean thickness (for the sample), "t" represents the t-distribution for a level of significance of 0.01 and n-1 degrees of freedom, "s" is the sample standard deviation, and "n" represents the number of samples.

Please note, however, that most months (nine of twelve) exhibit a sample size containing 30 or fewer observations. As a rule of thumb, for a sample size this small, the confidence interval may be drawn into question since it is not known whether the population distribution in each of these months is normal. This fact should be considered when evaluating conclusions based on data utilized in this paper. However, it is worth noting that the resulting confidence intervals calculated in this paper show consistency with the findings of Bohl and Junker (1987).

The results are shown in Table 1. (All values of mean thickness, standard deviation, confidence interval, and operational range are given in meters.)

Table 1. Mean monthly thicknesses and confidence intervals.

The results shown in Table 1 were then used during a two year trial period (January 1, 1997 through December 31, 1998), in which climatologically favored thicknesses were applied to 23 heavy precipitation events. It was found that climatologically favored thicknesses correctly identified 57% of all heavy precipitation events, and 67% of heavy precipitation events occurring between March and September (Table 2). March, May, and June produced the best verification results, ranging from 66% to 75% accuracy.

Table 2. Climatologically favored thicknesses applied to heavy precipitation events.

These results support the claim by Bohl and Junker (1987) that months which exhibit greater variability are often unreliable, since heavy precipitation events tend to be spread across a relatively large range of thicknesses. Here, the months October through February, when the highest standard deviations occurred, showed poor verification. During these months, climatologically favored thicknesses correctly identified only 38% of heavy precipitation events.

The test cases were not evenly distributed throughout the year. There were no cases for November, and for the months of February, August, and October, there was only one case per month. Thus, there is some question regarding reliability of this verification data as it pertains to the overall population. More cases should be collected as these mean monthly thicknesses are used operationally in the future.

3. Forecast Applications

As shown by this study, climatologically favored thicknesses are best applied when forecasting events during the warm months of March through September. During these months, when thickness gradients are relatively small, climatologically favored thicknesses can most accurately identify environments conducive to heavy precipitation. Conversely, during the cooler months of October through February, when thickness gradients are relatively large, climatologically favored thicknesses are least able to identify heavy precipitation producing environments.

In addition to the conclusions reached by this study, Bohl and Junker (1987) have established applications when using climatologically favored thicknesses in forecasting heavy precipitation.

Unseasonably high amounts of available moisture (precipitable water values 150% of normal or greater) are often associated with heavy rain.

Training often occurs along the thickness channel where the initial convection occurs. Cyclogenesis or deepening of an existing low pressure system usually justifies departure from the climatologically favored thickness. (The strong upward motion and dynamics are sufficient in producing heavy rain regardless of thickness.)

Synoptic scale stratiform overrunning precipitation usually develops at the lower range of climatologically preferred thickness. Progression of a low pressure system into the mature stage often shifts heavy precipitation toward higher thickness values. (This is most common with squall line activity across the Plains and southern states.)

Rainfall associated with mesoscale convective complexes and systems usually occurs at a higher range of thickness values.

Hurricanes and tropical storm rainfall tends to occur at higher thickness values than the climatological mean. Frequency of heavy rainfall decreases sharply at higher thickness values (generally above 5820 m). This is due to warm mid-level temperatures (capping) when 700 mb temperatures exceed 12^oC.

4. Conclusions

It has been shown that 1000-500 mb thicknesses correlate favorably with heavy precipitation producing environments during warm months (March through September). Data taken at Nashville, Tennessee between January 1, 1967 and December 31, 1996 have been used to quantify this relationship for a specific site. Such empirical studies can greatly enhance precipitation forecasting.

This method is most useful when synoptic scale dynamics are relatively weak and precipitation is being driven by mesoscale or thermodynamic processes (Bohl and Junker 1987), and during warm months when heavy rain is spread across a relatively narrow range of thickness values.

It must be noted, as explained by Bohl and Junker (1987), methods described here must be used in accordance with other forecasting methods. It is still important to locate low level boundaries, upper level impulses, and the location of best moisture inflow. Convection and heaviest rain will occur in conjunction with these features.

Acknowledgements

The authors thank Henry Steigerwaldt, Science and Operations Officer, and Darrell Massie, Lead Forecaster, NWSO Nashville, Tennessee for their reviews of this manuscript.

REFERENCES

Bohl, V. and N. Junker, 1987: Using climatologically favored thickness to locate the axis of heaviest rainfall. Nat'l. Wea. Dig., 12(3), 5-10.

Forecast Systems Laboratory and National Climatic Data Center, 1994: Radiosonde Data of North America, 1946-1993. CD-ROM.

Forecast Systems Laboratory and National Climatic Data Center, 1996: Radiosonde Data of North America, 1990-1995. CD-ROM.

Newmark, J., 1992: Statistics and Probability in Modern Life. Saunders College Publishing, Fort Worth, Texas, 739 pp.