Hector: Thunderstorms over the Tiwi Islands in the Maritime Continent


Hector from Gunn Point, Northern Territory

Three articles involving the thunderstorm phenomenon known as Hector were investigated. These storms take place over the Tiwi Islands, which are located approximately 80 km north of Darwin, Australia, during monsoon transition periods (November to December and February to March). Since the islands (Bathurst to the west and Melville to the east) are only separated by a narrow channel (Apsley Strait), they are treated as a single island in all three studies (see Figure 1, Beringer, et al.) and the total dimensions are approximately 150 km east to west and 60 km north to south.

Figure 1

Two of the articles attempted to describe and classify the thunderstorms themselves. Beringer, et al. used satellite imagery and soundings from the Darwin airport while Crook used linear and nonlinear models. The third article used data from the Maritime Continent Thunderstorm Experiment (MCTEX) to describe the diurnal cycle of the boundary layer.

All three articles recognized the existence of two wind regimes and also noted various modes of development or non-development. Only Beringer, et al. related the importance of Hector to the global climate: “[T]he Tiwi Islands heat the atmosphere from below…result[ing] in convective towers that penetrate upward carrying latent heat into the upper troposphere…resulting in the transport of heat and moisture that provides a meridional energy balance” by driving the Hadley and Walker circulations which are “vital to global atmospheric circulation.” (Beringer, et al., p. 1022) The other two studies focus very narrowly on the Tiwi Islands, never mentioning global implications.


All three studies took advantage of data from MCTEX, which was conducted in November and December of 1995, and an earlier experiment, the Island Thunderstorm Experiment (ITEX), that was conducted in 1988 and 1989. Schafer, et al. made extensive use of the MCTEX data to describe the boundary layer while Crook and Beringer, et al. relied on the same data for initialization and validation. All three studies compared and contrasted their results with ITEX data.

Beringer, et al. used infrared images from the Japanese Geostationary Meteorological Satellite between 0700 and 1800 local time (all times will be reported in local time which is UTC + 9.5 hours). They also took advantage of the 0730 radiosonde soundings at the Darwin airport. From these two sources they determined convective available potential energy (CAPE), precipitable water (PW), wind shear at the 600 hPa level, surface and 700 hPa wind speeds, convective inhibition (CIN) and the bulk Richardson number (BRICH). These values were used to sort 55 unambiguous event days (which were also not complicated by widespread oceanic convection) into four modes: Hector, suppressed Hector, no Hector, and late developing Hector.

While Beringer, et al. relied on soundings from the Darwin airport, Schafer, et al. took advantage of a comparatively dense network of sensors on Bathurst and Melville Islands. This network consisted of 15 automatic weather stations, three boundary layer wind profilers (levels of  every 100 m up to 2000 m and every 300 m up to 20 km), radiosonde launches every two hours at one location, a Doppler weather radar, a remotely controlled aircraft (aerosonde) and two all sky cameras. These observations were used to determine horizontal winds, mean virtual potential temperature, boundary layer depth, air temperature, and convergence. Horizontal and vertical variations of these descriptors were used to characterize the development of the boundary layer.

Rather than collecting or interpreting data, Crook used a linear and a nonlinear model to approximate the environment. The microphysics, topography and island shape were held constant in both models. Five parameters (surface wind speed, surface wind direction, surface flux of moisture, surface flux of heat and low-level moisture) were varied and each case was compared to a control experiment. Typical values for the control experiment (as well as a typical wind profile) were derived from MCTEX data.

Wind Regimes

All three studies recognized the existence of two dominant wind regimes. The east to northeasterly regime resulted in development on the western side while the west to southwesterly regime resulted in development on the eastern side. Beringer, et al. found a majority of the 55 cases they studied were in the easterly group while Schafer, et al. had eight cases in the easterly group and nine cases in the westerly group. Crook used only the westerly regime in his studies and demonstrated that convection was favored when winds were parallel to the major axis of the island. Only Beringer, et al. argued that storms in easterly flow had a larger areal extent (consistent with an earlier study: Down Under Doppler and Electricity Experiment) while storms in westerly flow had greater vertical development (as evidenced by -700 C cloud top temperatures instead of -600 C). This was explained by enhanced convection in the westerly regime due to advection of moist oceanic air over land.


Crook offers a good discussion of the historical understanding of Hector. Prior to MCTEX, it was hypothesized that sea breeze convergence over the island was the main trigger for convection. MCTEX demonstrated that sea breezes seldom collided. This led to a new explanation with two modes of development. The most common (80% of Hector events) is due to the collision of one of the sea breeze fronts with an evaporatively produced cold pool and is termed “type B”. This is similar to the “Hector” mode described by Beringer, et al., but also seems to lump the mode Beringer, et al. terms “suppressed Hector” into the same group. 36 out of 42 storm events studied by Beringer, et al. were either determined to be in the Hector mode or the suppressed Hector mode. This gives a frequency of 70%, very similar to the 80% mentioned by Crook.

The other 20% of events were “type A” according to Crook (who takes his categorization from a 2000 paper by Carbone, et al.). Type A Hectors owe their convergence to the collision of the two sea breeze fronts. This is analogous to the “late developing Hector” mode described by Beringer, et al. 14 of 46 or 30% of developments were determined to be late developing. Although Schafer, et al. were studying boundary layer development rather than the development of the actual Hector events, they also described a difference between early (type B after Crook and Hector/suppressed Hector after Beringer, et al.) and late (type A after Crook and late developing Hector after Beringer, et al.) developing events. They noted specifically that storms occurring after 1330 had a better developed boundary layer (~2 km deep) while storms occurring prior to 1330 had a shallower boundary layer (~1.5 km).

Based on Crook’s nonlinear moist model runs, he determines that low-level moisture (represented in his Figure 2 – originally Crook’s Figure 15 – by CAPE) is the primary factor in determining which mode will occur. From this figure, it can be seen that large CAPE results in type B development while small to moderate CAPE results in type A development. Beringer, et al. did not rely on a single parameter but rather the combination of low-level moisture (CAPE), CIN and shear (whereas Crook used a constant shear based on MCTEX data and does not consider CIN). They conclude (as is summarized in Table I) that large CAPE combined with large shear and large CIN lead to type A or late developing Hector. This contradicts Crook’s conclusion.

Figure 2. Total condensate against time for simulations with upstream CAPEs of 1000, 1500, and 3000 J/kg. All measures are normalized by the maximum value from the control experiment.

Figure 2. Total condensate against time for simulations with upstream CAPEs of 1000, 1500, and 3000 J/kg. All measures are normalized by the maximum value from the control experiment.

What is conspicuous about Table I, however, are the large standard deviations, especially for CAPE. Schaefer, et al. found that variability in boundary layer characteristics was minimized when sea breezes are most active, between 1000 and 1400. This helps to explain the large variability in Table I since 0730 soundings were used.

Table 1

Unique Findings of Each Study

Up to now, the overlapping areas of each study have been examined for similarities and differences. This analysis could only be carried so far since each study had its own unique objective. Exciting findings from each paper that do not necessarily overlap with the other two will now be discussed.

    • Evolution of maritime continent thunderstorms under varying meteorological conditions over the Tiwi Islands (Beringer, et al.)

      This study was the only one concerned with conditions leading to the most rare situation, the non-Hector. Beringer, et al. first discuss the suppressed Hector development mode, which is most likely due to suppression by pre-existing convection from a squall line moving through the area consuming CAPE. Although the non-Hector mode is also characterized by low CAPE, the presence of large shear (as opposed to small shear in the suppressed Hector development mode) indicates a different mechanism is responsible for this mode than for the suppressed Hector mode. Two proposed explanations were (1) mid-level subsidence combined with advection of dry continental air over the islands and (2) existence of residual cloudiness from previous deep convective activity over the ocean.

    • Understanding Hector: The dynamics of island thunderstorms (Crook)

      Upon recognizing that a majority of Hector events are not due to convergence of the two sea breezes, Crook proposes that perhaps a peninsula will produce the same results as an island. For the linear case, rainwater decreases by a factor of approximately two when a peninsula is used. For the nonlinear case, very little convection occurs over the peninsula (which has the northern Tiwi Islands coastline). From this he concludes that although sea breeze convergence is not the primary mechanism, interaction between the sea breezes and cold pools is enhanced by the proximity of the two coastlines and thus that an isolated heat source is a necessary condition for the Hector convective system.

    • Boundary layer development over a tropical island during the Maritime Continent Thunderstorm Experiment (Schafer, et al.)

      As mentioned previously, the first two papers are similar in that they strive to classify the Hector thunderstorm event. This paper practically ignores the thunderstorm events while it focuses on the development of the boundary layer throughout the diurnal cycle. Because of this, it was the only study to investigate the overnight environment and to describe what happens to reset it such that development can occur again the next day.  After sunset, a surface-based temperature inversion forms. A remnant mixed layer persists over the surface inversion until it is replaced by advection of an oceanic mixed layer over the island. Without this oceanic mixed layer, convection for the following day would be suppressed.


The three articles discussed took very different approaches to the same phenomenon. Beringer, et al. used available data (satellite images and airport soundings) to study the different wind regimes and development modes of Hector thunderstorms over the Tiwi Islands. From the reader’s standpoint, this was the best paper by far since it emphasized the importance of understanding Hector to making more accurate global predictions.

Crook approached the problem dynamically by applying linear and nonlinear models. In this manner, he was able to limit variability to one parameter at a time in order to investigate its effect on convection. Although conclusions about the different modes of Hector did not agree exactly with Beringer, et al., the importance of low-level moisture content was demonstrated. One of the main shortcomings of this study was the failure to investigate the easterly wind regime.

The most expensive of the three studies was certainly Schafer, et al. since it relied almost solely on data collected in situ with multiple sensors (rather than taking advantage of previously collected data sets). This paper was also the most difficult to compare and contrast with the other two since it focused on boundary layer development rather than the thunderstorms themselves. An important conclusion was the resetting of the mixed layer by oceanic air.


Beringer, J., N. J. Tapper, and T. D. Keenan, 2001: Evolution of maritime continent thunderstorms under varying meteorological conditions over the Tiwi Islands. Int. J. Climatol., 21, 1021-1036.

Crook, N. A., 2001: Understanding Hector: The dynamics of island thunderstorms. Mon. Wea. Rev., 129, 1550-1563.

Schafer, R., P. T. May, T. D. Keenan, K. McGuffie, W. L. Ecklund, P. E. Johnston, and K. S. Gage, 2001: Boundary layer development over a tropical island during the Maritime Continent Thunderstorm Experiment. J. Atmos. Sci., 58, 2163-2179.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s