1. Introduction

Thunderstorms accompanied by cloud-to-ground lightning and heavy rainfalls during the summer period can be significant threat to the human life. In order to analyze and predict severe storms with large hails or strong gusts, there have been a lot of researches by using proximity soundings (Brooks et al., 1994; Craven and Brooks, 2004; Doswell and Evans, 2003; Groenemeijer and Delden, 2007; Rasmussen and Blanchard, 1998). These studies usually covered the tornado events or thunderstorm over the Europe and US. The purpose of this study is to diagnose the possibility for the thunderstorm detection by using proximity soundings, and to adjust the threshold values or develop new indices appropriate for South Korea. We included the 5-year data (2002-2006) and the threshold values will be evaluated by using the recent data (2007).

2. Data and Method

a. Lightning data Lightning data (2002-2006) from Korean Meteorological Administration (KMA) are used. The location of lightning sensors and rawinsonde stations, currently operated in the South Korea are plotted in Fig. 1. In order to remove the lightning with incorrect locations, we only included the lightning observed from more than three sensors. Time-of-arrival (TOA) sensors do not discriminate multiple return strokes in a lightning, so multiple strokes occurred in 10km distance within 500 milliseconds are treated as a single lightning.

Fig. 1. The distribution of lightning detecting sensors and rawinsonde stations over South Korea.

b. Sounding data All of the sounding data are obtained from the department of atmospheric science at Wyoming University in the US (available online at http://weather.uwyo.edu/upperair/sounding. html). Although seven observatories make routine observations, four of them are used (Table 1) due to relative short history of observation and the quality of the data. Simple quality controls are applied. To exclude extreme soundingps, the data greater that mean + 3sigma (or 99^th percentile) or less than mean – 3sigma (or 1^th percentile) were removed based on 10-year (1997-2006) climatology.

Table 1. List of rawinsonde stations used in this study.

Station	Lat. N	Lon. E	Elv. (m)	Frequency of observations (times a day)
Osan	37.10o	127.03 o	52	4
Pohang	36.03 o	129.38 o	4	2
Gwangju	35.11 o	126.81 o	13	4
Jeju	33.28 o	126.16 o	73	2

c. Proximity criteria Four times-daily (Osan and Gwangju) and twice-daily soundings are used. Proximity is defined as being within 100 km of the sounding release location and during the period (-2 hrs to + 1 hr from the release time). Table 2 shows the criteria used in this study. We examined the thunderstorm classified by its mean intensity in four regions. The first category (NO) is strictly defined as none of lightning with no precipitation and no convective potential available energy (CAPE). Other categories are defined by using lightning cases, but they were subjectively classified depending on the mean intensity.

Table 2. Definitions and number of proximity soundings for the categories in four regions.

 

 

d. Parameters

A list of the parameters considered in this study is shown in Table 3. These parameters cover the instability (SSI, LI, and KI), the temperature of LCL, and water vapor contents in the air (mixing ratio and precipitable water). CAPE and convective inhibitions (CIN) are not included because CAPE and CIN during summer period over South Korea have relatively  small values (Eom et. al., 2008)

Table 3. List of parameters used in this study.

* Dimensionless unit

3. Preliminary Results

We analyzed the statistical distributions of each parameter for the different categories using box and whisker plot (Fig. 2) Most striking result is considerable discrimination (box range, no overlap in the middle 50 percent) between no lightning and lightning events as shown in LI, LCLT, and MLMR. Especially LI shows the least overlap. However, there are significant overlaps among lightning categories. These results almost coincide with other regions (Gwangju, Pohang, and Jeju). On the basis of these results, we combined some parameters (LI, MLMR, and LCLT), and analyzed the distributions by using probability density functions (PDFs). As shown in Fig. 3, there are some overlaps between no lightning (black solid line) and light cases (red solid line). We are going to adjust the threshold values (Fig. 4) and evaluate them using scalar attributed contingency table (e.g., POD, FAR, CSI, and so on).

(a)     SSI                      (b) LI                        (c) KI

      

    (d) LCLT                     (e) MLMR                    (f) TPW

      

Fig 2. Box and whisker plots of (a) SSI, (b) LI, (c) KI, (d) LCLT, (e) MLMR, and (f) TPW for each category in Osan. The mean and median values for each category are plotted in the left and right of the box, respectively.  

(a)     LI                        (b) MLMR – LI                  (c) LCLT - LI

Fig. 3. Probability density functions for (a) LI, (b) MLMR – LI, and (c) LCLT – LI for no lightning and lightning cases.

Fig. 4. Schematic diagram for determining threshold values using the PDFs.

4. Summary and future works

We examined the possibility for detecting the lightning events using proximity soundings over South Korea. The results show that LI, LCLT, and MLMR discriminate well between no thunder and thunder soundings. More parameters such as wind shear, lapse rate, and combination of some parameters will be done in the near future. And using PDFs, optimal threshold values of each parameter will be investigated for predicting thunderstorm over South Korea.