Document Sample
MET6480-Anderson Powered By Docstoc
					        Relating Lightning Frequency to SST Gradients over the Gulf Stream

                                HOLLY A. ANDERSON
                    The Florida State University, Tallahassee, Florida


         Lericos et al. (2002) linked increased nocturnal lightning frequency over the
Atlantic Ocean to the warmer oceanic temperatures of the Atlantic’s Gulf Stream
Current. Warmer sea surface temperatures (SSTs) presumably would increase
convection over the ocean and cause lightning associated with the convective storms.
Lindzen et al. documented the connection between SST gradients and low-level wind
convergence in the tropics. Despite many observations and assumptions, no previous
study has sought to quantitatively determine the connection between the Gulf Stream and
lightning frequency. Following the study of Lericos et al., the patterns of nocturnal
lightning during the warm season are analyzed. For this study, a single warm season,
specifically May through September 2003, will be analyzed. The objective of this study
is to research the relationship between the gradient of the SST over the Gulf Stream and
lightning frequency.


           a. Data Sources

        Sea surface temperature data on a 0.25 by 0.25 degree grid from the Advanced
Microwave Scanning Radiometer, AMSR-E, is used for this study. The AMSR-E is
aboard NASA’s sun-synchronous spacecraft Aqua. Due to its sun-synchronous nature,
two daily passes are available; one ascending at 10:30 am local time and one descending
pass at 1:30 am local time. Because this study concentrates on nocturnal lightning, the
descending pass will be used in order to provide the best look at nocturnal oceanic
temperatures before solar heating affects the SSTs. The microwave sensor is affected by
rain, sun glint, high wind speeds (>20 ms-1), and sea ice. For this study, data limitation
due to rain will be the greatest concern, as most lightning strikes are associated with areas
of heavy rain. Side-lobe contamination near the coastlines also affects the retrieval of
SST information near the eastern coast of the United States, so the Gulf Stream Current
near the coastline is not resolved well. In addition, the satellite’s swath coverage makes
daily passes of the Gulf Stream inconsistent.
        In addition to AMSR-E SST data, cloud-to-ground (CG) lightning data from the
U.S. National Lightning Detection Network (NLDN) is used. The NLDN consists of
over 100 ground sensors that can detect a ground flash over 400 km from the sensor.
The NLDN can detect lightning as far east in the Atlantic Ocean as 60W, but the
detection efficiency of the sensor decreases as the distance from the sensor increases.

           b. Methodology

       A single warm season, defined as May through September 2003, has been
analyzed for this study. Nocturnal lightning, defined as lightning occurring between
09:00:00 pm and 05:59:59 am, was subset into daily files. The daily SST gradient in
°Cm-1 was calculated from the AMSR-E SST data, by:

                                             T T
                                    SST      
                                             x y

where SST is the horizontal SST gradient in the x and y directions. Calculations were
made using centered finite differencing. Any missing values due to rain, sun glint, or
other data limitations were not included in the gradient calculations because they would
adversely affect the gradient values.
        If a nocturnal CG lightning strike occurred in the Atlantic Ocean basin domain of
20N to 50N latitude and 80W to 60W longitude, and SST gradient information was
available, it was added to the dataset. This led to a dataset of over 74092 values for the
entire northwest Atlantic region. An example of a daily plot of the SST gradient with
CG lightning overlaid is shown in Figure 1. In order to prevent large areas of non-Gulf
Stream SST gradients from affecting the sample, a smaller subset domain, centered over
the Gulf Stream region, henceforth referred to as the Gulf Stream domain, from 35N to
45N and 75W to 60W was taken and a second dataset was formed. This dataset included
27139 values. Values were binned in 100 equal bins according to SST gradient value
and plotted to analyze the relationships between the SST gradient and lightning frequency
distribution. Data was statistically analyzed and displayed using the statistics package
SPSS and Microsoft Excel.

3.     RESULTS

         Since the AMSR-E did not gather SST data from areas of heavy rain, any
statistical analysis using the datasets generated for this study must acknowledge that the
datasets created lack an all inclusive look of all oceanic lightning occurring in this
domain. Because of the inherent limitations of microwave sensing, interpretations must
be carefully analyzed.
         Figure 2 and Table 1 show the histogram of SST gradient and the related statistics
for the entire northwest Atlantic Ocean domain. From the histogram it is apparent that
the majority of lightning strikes occurred over low SST gradient areas. According to the
results from the larger dataset, lightning most frequently occurs over areas of SST
gradients of 0 °Cm-1. The distribution is skewed towards negative values with a slightly
negative mean value of -2.861 x10-5 °Cm-1. It is noted that lightning is more prevalent
over areas of lower SST gradients, not higher SST gradients as expected.
         With this result, the decision to analyze the isolated area of the Gulf Stream was
made in order to exclude large areas of low gradient in non-Gulf Stream areas of the
Atlantic. These areas of low SSTs could be skewing the sample to low gradient values,
since the majority of the Atlantic basin is associated with low temperature gradients.
The new domain yielded a dataset of roughly 26000 values. Figure 3 shows the SST
gradient for the Gulf Stream domain. Even with the domain restrictions to the Gulf
Stream region, a similar frequency distribution is noted. Once again, lightning

frequencies are elevated near lower SST gradients, as the most frequent value is 8.53373
x10-5 °Cm-1, as noted in Table 2.


        Given the inherent data limitations, the results appear to indicate that a positive
correlation between the SST gradient and lightning frequency is not observed in the large
Atlantic Ocean basin nor the smaller Gulf Stream domain. Instead of seeing higher
frequencies of lightning over larger SST gradients, it is apparent that lightning occurs
preferentially in areas of just slightly negative SST gradients. This could imply that
lightning occurs over oceanic areas with near constant warm or near constant cool
temperatures, not where SSTs fluctuate the most as expected. This led to the question of
whether lightning occurred most frequently on the warmer or the cooler side of the Gulf
Stream Current.
        Given this unexpected result, the search for a physical explanation led to the
location of a paper by Lindzen et al. (1897), which linked SST gradients to increased
convection. In Lindzen et al., the authors state that differences in SSTs lead to different
pressures over the water due to differences in density. Therefore, warmer SSTs are
linked to lower densities and low pressure, whereas cooler SSTs are linked to higher
densities and higher pressure. Because low-level wind would flow from higher to lower
pressures, low-level convergence and convection would be expected over the warmer
SSTs. Therefore, a logical continuation of this theory would lead to the extension of
increased lighting frequency over the warmer SSTs.
        A visual inspection of the daily SST gradient and lightning files shows the
majority of nocturnal lightning strikes occur south of the Gulf Stream, or in areas of
warmer SSTs. Therefore, the theory proposed by Lindzen et al. (1987) supports this
study’s findings that lightning occurs preferentially in areas of low SST gradient.
        Despite the applicability of Lindzen’s theory to this study’s findings, it would be
incorrect to assume that lightning frequency is related to the SST gradients alone, as the
oceanic and atmospheric interactions are highly complicated. Further research would
benefit from looking at the synoptic patterns over and near the domain of study to see the
influence of frontal systems, troughs, highs, and other weather systems on the frequency
of lightning and the results of their interactions with changes in SSTs. It is possible that
there is a time delay between a weather system moving over areas of increased SST
gradients and the actual occurrence of lightning after the weather system’s modification.
In addition, future research hopes to associate the low-level wind patterns to increases in
vertical convection using AMSR-E wind speed data.


        In conclusion, it appears that the SST gradient does indeed affect the frequency of
cloud-to-ground lightning strikes. It appears that lightning occurs preferentially in areas
of low to no-gradient, where SSTs are at near constant values. These findings provide a
natural extension to a theory by Lindzen et al. (1987) in which increased convection
occurs over warmer SSTs due to density effects. Lightning therefore occurs near SST
gradients, but not directly above those areas, as was expected. Future research hopes to

link low-level wind field patterns to the increased convection, and study the influence of
synoptic systems to the lightning frequency as well.


Lericos, T.P., H.E. Fuelberg, A.I. Watson, and R.L. Holle, 2002: Warm Season Lightning
Distributions over the Florida Peninsula as Related to Synoptic Patterns. Wea.
Forecasting, 17, 83–98.

Lindzen, R.S., and S. Nigam, 1987. On the role of sea-surface temperature gradients in
forcing low-level winds and convergence in the tropics. J. Atmos. Sci., 44, 2418-2436.

Figure 1. Plot of SST gradient for 9 May 2003. Gradients are in units of °C/m. Pink
dots of CG lightning strikes are few over the oceanic area of interest for this
particular day. Increased SST gradients associated with the Gulf Stream are apparent.

Figure 2. Lightning frequency distribution for the entire Atlantic domain.


  N                         V alid                 74092
                            Mis sing                      0
  Mean                                             ********
  Std. Dev iation                                  ********
  V arianc e                                          .000
  Skew nes s                                         -.437
  Std. Error of Skew ness                             .009
  Range                                            ********
  Minimum                                          ********
  Max imum                                         ********

 Starred values above are as follows:

 Mean: -2.861515130053e-5
 Mode: 0
 Std. Dev: 0.0001142452688411
 Range: 0.001688542
 Min: -0.00110681
 Max: 0.000581732

Table 1. SST gradient statistics for the entire Atlantic domain.

Figure 3. Lightning frequency distribution for the smaller Gulf Stream domain.


  N                         V alid               27139
                            Mis sing                    0
  Mean                                           ********
  Std. Dev iation                                ********
  V arianc e                                        .000
  Skew nes s                                       -.641
  Std. Error of Skew ness                           .015
  Range                                          ********
  Minimum                                        ********
  Max imum                                       ********

   Starred values above are as follows:

   Mean: -3.700997284279e-5
   Mode: 8.53373e-5
   Std. Deviation: 0.0001330702191255
   Range: 0.001688542
   Min: -0.00110681
   Max: 0.000581732

Table 2. SST gradient statistics for the smaller Gulf Stream domain.