JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109, D16206, doi:10.1029/2003JD003833, 2004
Islands as miniature continents:
Another look at the land-ocean lightning contrast
Earle Williams and Twiggy Chan
Parsons Laboratory, Department of Civil and Environmental Engineering, Massachusetts Institute of Technology,
Cambridge, Massachusetts, USA
Dennis Boccippio
NASA Marshall Space Flight Center, Huntsville, Alabama, USA
Received 3 June 2003; revised 15 April 2004; accepted 17 May 2004; published 20 August 2004.
[1] Numerous observations substantiate a pronounced contrast in lightning activity
between continents and oceans. The traditional explanation for continental dominance is
based on a contrast in thermal properties of land and sea. A more recent idea is based on
the contrast in boundary layer aerosol concentration between land and sea. This study
makes use of islands as miniature continents of varying area to distinguish between these
two hypotheses. Scaling law analysis is used to predict transitional island areas for the two
hypotheses. NASA Tropical Rainfall Measuring Mission satellite observations provide a
uniform data set on island activity. The island area dependences of lightning activity
are more consistent with the thermal hypothesis than the aerosol hypothesis, but this
conclusion must be tempered with the extreme simplification of the theoretical
INDEX TERMS: 0305 Atmospheric Composition and Structure: Aerosols and particles
predictions.
(0345, 4801); 3304 Meteorology and Atmospheric Dynamics: Atmospheric electricity; 3307 Meteorology and
Atmospheric Dynamics: Boundary layer processes; 3324 Meteorology and Atmospheric Dynamics: Lightning;
KEYWORDS: aerosols, lightning, islands
Citation: Williams, E., T. Chan, and D. Boccippio (2004), Islands as miniature continents: Another look at the land-ocean lightning
contrast, J. Geophys. Res., 109, D16206, doi:10.1029/2003JD003833.
1. Introduction
[ 2 ] The pronounced contrast in lightning activity
between land and ocean on a global scale has been firmly
established over the course of the twentieth century. The
first global maps of thunder day observations by Brooks
[1925] already revealed a strong land dominance, but at
that time, skepticism remained about the validity of the
oceanic map because of sampling limitations. Later satellite observations from space with uniform land-ocean
coverage [Orville and Henderson, 1986] dispelled this
skepticism and upheld an order-of-magnitude contrast.
Recent observations with the Lightning Imaging Sensor
(LIS) on the Tropical Rainfall Measuring Mission
(TRMM) satellite, shown in Figure 1 and a key record
for the present study, have shown that the land-ocean
contrast is evident for the entire local diurnal cycle
[Christian et al., 2003]. A similar contrast in activity is
evident in satellite observations of radio frequency emission from lightning [Kotaki et al., 1981].
[3] The physical origin of the land-ocean contrast has
not been firmly established. The traditional hypothesis
rests on the physical contrast in land and ocean surface
properties. Land surfaces become hotter in response to
solar radiation, and this leads to more unstable conditions
Copyright 2004 by the American Geophysical Union.
0148-0227/04/2003JD003833$09.00
over land and stronger updrafts that invigorate the ice
microphysical processes believed vital to electrical charge
separation and lightning. The more recent aerosol hypothesis [Williams et al., 2002] rests on the established
differences in pollution between land and ocean. Over
land the enhanced number of cloud condensation nuclei
leads to cloud droplets that are more numerous and
smaller, thereby suppressing the warm rain coalescence
process and enabling more cloud water to access the
mixed phase region where it can participate in the
electrification process. These two hypotheses and observational tests to distinguish them are discussed in greater
detail by Williams et al. [2002] and Williams and Stanfill
[2002].
[4] This study is concerned with lighting activity on
islands as miniature continents, as a further test of the
origins of the land-ocean contrast. On the coarse global
scale in Figure 1 the land ocean contrast is obvious. The
outstanding question posed in this study is, How small can
an island be and still exhibit a land-ocean contrast? Predictions for critical island size according to the thermal and
the aerosol hypothesis are presented in section 2.
2. Predictions for Critical Island Size
[5] This section is concerned with estimates for the
critical island areas needed for a manifestation of large
continent behavior evident in Figure 1. The predictions for
D16206
1 of 5
D16206
WILLIAMS ET AL.: ISLANDS AS MINIATURE CONTINENTS
D16206
Taking representative values [Ludlam, 1980; Cotton and
Anthes, 1989] for the parameters r = 1000 m, W = 5 m s1,
t = 3600 s, and h = 500 m, the corresponding critical island
area is 100 km2. Note that if the island is smaller than this
critical value, then cooler oceanic air is ingested over the
cloud lifetime and the cloud begins to take on a maritime
character. If the island possesses a supercritical area, then
the supply of continental boundary layer air is not
exhausted, and full continental behavior is expected.
Figure 1. Global map of lightning flash density based on
observations with the Lightning Imaging Sensor on the
NASA Tropical Rainfall Measuring Mission satellite for the
year 2000 [from Goodman and Cecil, 2001]. See color
version of this figure in the HTML.
the thermal and the aerosol hypothesis, which turn out to be
distinctly different, are discussed in turn.
2.1. Thermal Hypothesis
[6] According to this hypothesis, islands are characterized by a Bowen ratio that is invariably greater than that of
the surrounding ocean surface. In response to solar shortwave radiation, stronger sensible heating occurs over the
island, and so the island’s boundary layer air will possess
greater cloud buoyancy than the air drawn from the adjacent
ocean. Boundary layer thermals from the island surface are
set up which eventually initiate moist convection over the
island. An isolated cumulonimbus cloud with lifetime t,
cloud base height h (assumed matched with boundary layer
height), and updraft velocity w with updraft radius r over a
circular island of radius R, is illustrated in Figure 2.
[7] Simple kinematics is used to solve for the island area
whose anomalous boundary layer volume is just sufficient
to supply the cloud with ingested air over its lifetime t.
Equating volumes,
2.2. Aerosol Hypothesis
[8] Global maps of condensation nuclei concentrations
[Hogan, 1977; see also Williams et al., 2002] show order-ofmagnitude enhancements over continents. We are unaware
of other studies focused on the contrast in the land-ocean
aerosol concentration for islands, which give detailed traverses of boundary layer aerosol concentration across an
island and the adjacent sea. To estimate a critical island area
for continental aerosol behavior, we consider the intrusion
of the sea breeze in replacing the island’s continental
boundary layer air with cleaner maritime air, as illustrated
in Figure 3. The sea breeze intrusion over the Florida
peninsula is widely studied [Byers and Rodebush, 1948;
Pielke, 1974] and can be used to provide parameters for the
estimates here. Indeed, the Florida peninsula has a critical
width in the sense that approximately half a day is required
for ‘‘cleaning’’ by sea breeze intrusion from the east and
west coasts of Florida in a meteorological regime characterized by weak synoptic-scale flow. Taking the peninsula
width (200 km) as a measure of critical island diameter, we
have
pR2c ¼ pð100 kmÞ2 30; 000 km2 :
pR2c h ¼ pr2 wt:
Solving for the critical island area,
pR2c ¼ pr2 wt=h:
Figure 2. Illustration of convective enhancement by an
island, following the thermal hypothesis for the land-ocean
lightning contrast [from Williams and Stanfill, 2002]. See
color version of this figure in the HTML.
Figure 3. Illustration of the sea breeze intrusion over an
island, serving to replace polluted island air with cleaner
maritime air. See color version of this figure in the HTML.
2 of 5
D16206
WILLIAMS ET AL.: ISLANDS AS MINIATURE CONTINENTS
D16206
This estimate is more than 2 orders of magnitude greater
than the critical island size according to the thermal
hypothesis.
of regional island groups (i.e., the Caribbean, the Philippines, the Indonesian archipelago) with common seasonality
and sea surface temperature, thereby providing more control
on the results in this single-parameter (island area) study.
Sampling time is one notable limitation with the LIS data
set for this application. Despite the 3-year compilation the
total view time for some of the smaller islands was of the
order of only 30 min.
[12] Three LIS quantities were used as measures of
lightning activity. The mean ‘‘area’’ flash rate (flashes
min1), the peak ‘‘area’’ flash rate (flashes min1), and
the flash density (flashes km2 yr1). These quantities were
extracted from the algorithm-derived quantities ‘‘area’’ and
‘‘flash.’’ The area, also discussed by Williams et al. [2000],
is equivalent to a thunderstorm. The flash is the best
representation of an individual lightning flash, based on
the analysis of the space-time sequence of raw pixel
illuminations in the LIS charge-coupled device array. The
mean area flash rate for an island is computed, for example,
by finding all thunderstorm areas on that island in the 3-year
data set, computing the flash rate over the satellite view time
for each area, and then taking the mean flash rate for all the
areas collected. Similar procedures were followed for the
other satellite-derived quantities.
3. Methodology
4. Results
[9] A comprehensive data set of the world’s islands has
been compiled by the United Nations and is generally
available (http://islands.unep.ch/). This information includes
island name, latitude, longitude, maximum elevation, and
surface area. The islands are arranged alphabetically and by
increasing area, an organization well suited to this study. A
second key source of information on island coastlines can
be found at a National Oceanic and Atmospheric Administration web site (http://www.ngdc.noaa.gov/mgg/shorelines/
shorelines.html). This enabled the clean delineation of
island area from adjacent ocean for sorting the satelliteobserved lightning activity.
[10] Tests of the predictions in section 2 for critical island
area require homogeneous data sets on lightning activity.
In an earlier study of the land-ocean lightning contrast
[Williams and Stanfill, 2002] the thunder day data set
[World Meteorological Organization, 1956] was briefly
considered. These comparisons with island area are reproduced in Figure 4. The principal limitation of this analysis
lies in the limited number of islands with surface station
reports of thunder days, about 70 in total. Furthermore,
these islands are located throughout the world in regions of
different sea surface temperature, different synoptic conditions, different seasonality, and different topography.
These uncontrolled variables undoubtedly contribute to
the scatter in the results shown in Figure 4.
[11] Improvements in the documentation of cloud electrification on islands for the present study have come from
observations with the LIS on board the NASA TRMM
satellite. A 3-year compilation of LIS observations (December 1997 to December 2000) was used for this study. The
LIS detects lightning with high efficiency (>80%) in both
daytime and nighttime conditions and with a spatial resolution (4 km 4 km) sufficient for the smallest islands
selected for study. The LIS observations enabled the study
[13] Observations with the LIS were made on approximately 300 islands ranging in size from the satellite resolution (16 km2) to the world’s largest islands (Borneo,
Madagascar, Sumatra, and Java) with areas approaching 1
million km2. The cleanest results were obtained for groups
of islands, most notably the Caribbean region in the Atlantic
Ocean and the Philippines Islands in the Pacific Ocean.
Integrated lightning maps for the largest islands and their
surrounding ocean showed clear concentrations of events
over the island, like the continental-scale result in Figure 1
on a global scale. As the island size diminished, this
concentration of lightning over the island became less
obvious, and for the smallest islands considered, it was
not apparent at all. The local diurnal variation of lightning
was also briefly considered, and a clear daytime predominance was found. As one specific example in the context of
other scientific exchange on distinguishing the thermal and
aerosol hypotheses [Williams and Stanfill, 2004], all 20 LISdetected flashes on the island of Guadeloupe were daytime
events.
[14] Figure 5 shows the flash rate density (flashes km2
yr1) for 27 islands in the Caribbean region, including the
peninsula of Florida for comparison. Florida, Cuba, and
Juventud (an island adjacent to Cuba to the south) all show
values of the order of 20 flashes km2 yr1, consistent with
the general level over large continents shown in Figure 1.
For island areas in the range of 100 km2 to a few thousand
square kilometers the total flash density falls off substantially (with considerable scatter), with the smallest values
(0.4 flashes km2 yr1) (and roughly 2% of the maximum
continental values in Figure 1), comparable to the higher
range of values in the mid-Atlantic and Pacific oceans.
Trinidad is an outlier (large-than-average area for its flash
density) that may be attributable to its extreme southern
latitude in this group.
Figure 4. Annual number of thunder days [World
Meteorological Organization, 1956] for islands versus
island area (adapted from Williams and Stanfill [2002]).
The shading divides the data set into three equal subsets.
3 of 5
D16206
WILLIAMS ET AL.: ISLANDS AS MINIATURE CONTINENTS
D16206
the range of 102 –103 km2, consistent with the results in
Figures 5 and 6.
5. Discussion
[15] Figure 6 shows the mean thunderstorm flash rate for
the same collection of Caribbean islands as in Figure 5.
Again, Florida, Cuba, and Juventud dominate the values,
with a rough order-of-magnitude contrast with the smaller
islands. Despite considerable scatter the transitional island
area is clearly in the range 100– 1000 km2. The predictions
for the thermal and the aerosol hypothesis are also shown.
[16] The final Figure 7 shows the mean thunderstorm
flash rate for thunderstorm areas observed over selected
Philippine Islands in the Pacific Ocean. The mean flash rates
for the largest islands are clearly dominant in general,
though Dinagat, Siquihor, and Burras are outliers. Curiously,
these three islands all have relatively low relief. The key
point for this study is the evidence for a transition scale in
[17] It is first appropriate to discuss several caveats in the
interpretation of the observational results. These caveats
pertain to both the thermal and the aerosol hypotheses.
[18] The treatment of the thermal hypothesis illustrated in
Figure 2 is highly simplified in its emphasis on cloud
buoyancy rather than on the issues of forcing at a larger
scale (by elevated terrain or the sea breeze front). Also not
considered is the scaling of updraft width with cloud base
height and the possibly enhanced efficiency with which
convective available potential energy (CAPE) can be transformed to updraft kinetic energy and ultimately to lightning
activity [Williams and Stanfill, 2002; Toumi and Qie, 2004].
The daytime predominance of lightning over islands is more
consistent with the thermal hypothesis than the aerosol
hypothesis to the extent that instability and CAPE are
enhanced by solar heating of land surface.
[19] The treatment illustrated by Figure 3 is also highly
simplified. The intruding sea breeze does not destroy
preexisting continental aerosol but merely lifts it higher in
the atmosphere, mixes with it, and dilutes its concentration.
The aerosol concentration, rather than total storm-ingested
aerosol, is the important parameter for the aerosol hypothesis [Williams et al., 2002], but it is still unknown how large
an aerosol concentration is needed to enhance the lightning
activity following this hypothesis. Thunderstorm convection
initiated on the progressing sea breeze front and ingesting
island continental aerosol prior to the complete island
‘‘sweep’’ by the sea breeze is also ignored in this treatment.
This neglect may exaggerate the contrast in the predictions
for critical island area for the two hypotheses. Also
neglected is a synoptic-scale flow capable of transporting
continental island aerosol over the sea and also capable of
Figure 6. Mean flash rate (flashes min1) for thunderstorms versus island area for islands in the Caribbean
region. The shading divides the data into three equal
subsets.
Figure 7. Mean thunderstorm flash rate (flashes min1)
versus island area for selected Philippine Islands. The
shading divides the data into three equal subsets.
Figure 5. Flash density (flashes km2 yr1) versus island
area for islands in the Caribbean region. The shading
divides the data set into three equal subsets.
4 of 5
D16206
WILLIAMS ET AL.: ISLANDS AS MINIATURE CONTINENTS
disrupting the symmetry of the sea breeze convergence
idealized in Figure 3.
[20] The variable sea breeze penetration over islands and
the mixing between the continental and maritime air will
lead to corresponding variations in the aerosol concentration
from maritime levels off the island to fully continental
levels in the island interior. The sampling of the LIS
observations is inadequate to examine the transition in
lightning flash density across island coastlines, as noted in
section 3. However, in the case of the National Lightning
Detection Network over the continental United States,
where observations have accumulated on a nearly continuous basis for more than a decade, the variation of flash
density across coastlines is well defined and can be
explored. As one specific example, Steiger and Orville
[2003] show maps of lightning flash density along the Gulf
Coast. The pronounced drop in flash density from land to
sea at the coastline, with a gradient region approximately
one thunderstorm diameter (10 km) in width, strongly
suggests that thermal properties, rather than boundary layer
aerosol concentration, are responsible for the lightning
behavior. The transition in thermal properties (heat capacity
and Bowen ratio) is expected to be abrupt at any coastline.
6. Conclusion
[21] The dependence of lightning activity on island area
has been used to distinguish thermodynamic and aerosol
contributions to the land-ocean lightning contrast. All results
demonstrate a transition range in island area (102 – 103 km2)
that is more consistent with the thermodynamic explanation
than the one based on aerosol to the extent allowed by the
simplified theoretical treatment. The consistency found
between the present results based on LIS observations (with
sampling times as short as 30 min over a 3-year period) and
the earlier thunder day tabulations (with decadal sampling)
supports a general result, independent of data set.
[22] Acknowledgments. This work was supported by the NASA
TRMM project on grant NAG5-9637. We thank Ramesh Kakar and Bob
Adler for their support. Karen Rothkin assisted with earlier organization of
island and lightning data sets, and Marie Dow and Glenn Cook assisted
with figures. We thank D. Rosenfeld for numerous discussions and one
particularly persistent reviewer for identifying a numerical error and for
other important improvements to the manuscript.
D16206
References
Brooks, C. E. P. (1925), The distribution of thunderstorms over the globe,
Geophys. Mem. London, 24, 147 – 164.
Byers, H. R., and H. R. Rodebush (1948), Causes of thunderstorms of the
Florida Peninsula, J. Meteorol., 5, 275 – 280.
Christian, H. J., et al. (2003), Global frequency and distribution of lightning
as observed from space by the Optical Transient Detector, J. Geophys.
Res., 108(D1), 4005, doi:10.1029/2002JD002347.
Cotton, W. R., and R. A. Anthes (1989), Storm and Cloud Dynamics, 883
pp., Academic, San Diego, Calif.
Goodman, S. J., and D. J. Cecil (2001), Structure and characteristics of
precipitation systems observed by TRMM, paper presented at 11th Conference on Satellite Meteorology and Oceanography, Am. Meteorol. Soc.,
Madison, Wis., 15 – 18 Oct.
Hogan, A. (1977), Meteorological variation of maritime aerosols, in Atmospheric Aerosols and Nuclei, edited by A. F. Roddy and P. C. O’Connor,
pp. 503 – 507, Galway Univ. Press, Galway, Ireland.
Kotaki, M., I. Kuriki, C. Katoh, and H. Sugiuchi (1981), Global distribution
of thunderstorm activity observed with ISS-b, J. Radio Res. Lab. Jpn., 28,
49 – 71.
Ludlam, F. H. (1980), Clouds and Storms: The Behavior and Effect of
Water in the Atmosphere, 405 pp., Penn. State Univ. Press, University
Park.
Orville, R. E., and R. W. Henderson (1986), The global distribution of
midnight lightning: December 1977 to August 1978, Mon. Weather
Rev., 114, 2640 – 2653.
Pielke, R. A. (1974), A three-dimensional numerical model of the sea
breezes over south Florida, Mon. Weather Rev., 102, 115 – 139.
Steiger, S. M., and R. E. Orville (2003), Cloud-to-ground lightning
enhancement over southern Louisiana, Geophys. Res. Lett., 30(19),
1975, doi:10.1029/2003GL017923.
Toumi, R., and X. Qie (2004), Seasonal variation of lightning on the Tibetan Plateau: A spring anomaly?, Geophys. Res. Lett., 31, L04115,
doi:10.1029/2003GL018930.
Williams, E., and S. Stanfill (2002), The physical origin of the land-ocean
contrast in lightning activity, C. R. Phys., 3, 1277 – 1292.
Williams, E. R., and S. Stanfill (2004), Reply to the comment on ‘‘The
physical origin of the land-ocean contrast in lightning activity,’’ C. R.
Phys., 5, 157 – 158.
Williams, E., K. Rothkin, D. Stevenson, and D. Boccippio (2000), Global
lightning variations caused by changes in thunderstorm flash rate and by
changes in the number of thunderstorms, J. Appl. Meteorol., 39, 2223 –
2230.
Williams, E. R., et al. (2002), Contrasting convective regimes over the
Amazon: Implications for cloud electrification, J. Geophys. Res.,
107(20), 8082, doi:10.1029/2001JD000380.
World Meteorological Organization (1956), World distribution of thunderstorm days, WMO/OMM 21, 21 pp., Geneva, Switzerland.
D. Boccippio, NASA Marshall Space Flight Center, Huntsville, AL
35805, USA. ([email protected])
T. Chan and E. Williams, Parsons Laboratory, Department of Civil and
Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA. ([email protected]; [email protected])
5 of 5