Hector is a vigorous convective system that develops on the Tiwi Islands, two islands included in the “Maritime Continent” , an area extending across the Indonesian archipelago, north Australia and New Guinea. This is one of the primary regions of global latent heat release contributing to the forcing of planetary-scale circulations (e.g., Hadley and Walker cells). The Tiwi Islands, located in the northern tropical part of Australia, produce regular tropical convection during the pre-monsoon and monsoon “break” seasons (from November to March) in response to the latent heat released during the diurnal cycle .

This storm has been analyzed during observing campaigns like ITEX (Island Thunderstorm Experiment, 1988), MCTEX (Maritime Continent Thunderstorm Experiment, 1995), SCOUT-O3 (Stratospheric-Climate links with emphasis On the Upper Troposphere and lower stratosphere, 2005) and TWP-ICE (Tropical Warm Pool – International Cloud Experiment, 2006) whose aims were to better understand the triggering mechanisms and the meteorological parameters favorable to the convective development. Particularly the MCTEX campaign collected many environmental factors that are known or believed to influence the initiation, organization, propagation and intensity of deep convection. In addition this data set allowed two distinct forcing regimes leading to Hector to be defined :

type A: resulting from the confluence and convergence of the sea breeze fronts;

Following these campaigns, some ideal and real numerical studies have been performed to understand the forcing and the triggering mechanisms of the Hector storm (e.g., ).

used the UK Met Office's mesoscale model at 3 km resolution initialized by a sounding to examine two cases from ITEX. The results suggest that the model qualitatively reproduces the diurnal evolution of Hector, showing a clear relationship between the storm development and the island topography. used the Japanese Meteorological Institute's mesoscale model at 1 km resolution to simulate a case from MCTEX. The study highlights a good agreement between the simulations and observations and focuses on the five stages of the convective life cycle.

simulated the 30 November 2005 Hector event using two models, the Advanced Research Weather Research and Forecasting (ARW) model, and the Met Office Unified Model, with a resolution of 1 km. Both models reproduce the development of Hector fairly well, even though the two simulated surface heat fluxes are very different. This would mean that the intensity of the storm is not only controlled by this factor. The aim of the paper is to investigate the role of deep convection in the vertical transport of tropospheric air into the lower stratosphere. conducted a further simulation with ARW in large eddy simulation (LES) mode, refining the grid spacing to 250 m, and concluded that the characteristics of the Hector storm are basically similar in time and space to those obtained in the 1 km resolution. Therefore a 1 km resolution is fine enough to simulate the timing, the structure and strength of deep convection when compared with the field campaign observations .

In the study by two Hector events (one observed during SCOUT-O3 and one during TWP-ICE) have been investigated, analyzing the dynamics and thermodynamics. Using the Fifth-Generation Mesoscale Model (MM5) at 1 km resolution over the Tiwi Islands, several numerical experiments have been performed with the aim of understanding the forcing and triggering conditions for the development of Hector. The study demonstrates the key role of the sea breeze, water vapor content and soil moisture content in the growth of Hector. Moreover, carried out a study for highlighting both the triggering factors and microphysical structure of a Hector event. The event was analyzed using MM5 model simulations, ground-based radar and TRMM (Tropical Rainfall Measuring Mission ) satellite data, with the aim of understanding the mechanisms leading to the convective development. The analysis of the horizontal and vertical structure at high temporal and spatial resolution produced by MM5 allows the mechanisms for triggering Hector to be established: sea breeze, a gust front from previous convection and the channeling effect by topography.

simulated four cases of the Hector storm by running the ARW model with a maximum horizontal resolution of 1 km, incorporating and not incorporating the observations collected during the ACTIVE campaign. Only one (30 November 2005) of the four cases was well simulated by the run without the inclusion of observations. Three events (16 November, 6 and 10 February 2006) can be simulated only if the model was run incorporating observations. The major deficiency deduced by in the simulations of Hector is the smaller size and the weaker intensity in comparison with the observations.

Simulations were performed for a Hector event observed on 30 November 2005 by using the Meso-NH (mesoscale non-hydrostatic) model, performed with a grid spacing of 1600, 800, 400, 200 and 100 m. The updraft generally decreases with reduced resolution due to the reduced entrainment into the base of the updrafts. Indeed, the strong updrafts in the boundary layer obtained by the three finest simulations reinforce the updrafts in the upper troposphere.

performed an ideal study using both a linear and nonlinear flow model for assessing the most important parameters of the Hector convective system. The low-level moisture is found to be an important parameter for differentiating between type A and B. High values of low-level moisture correspond to earlier convection, then the associated evaporational cooling produces cold pools that retard the further inland progress of the sea breezes. Hence, Hector type B develops because of the convergence of the sea breeze and the gust front related to previous convection. Hector type A, which is associated with low values of low-level moisture, develops when the generation of precipitating cold pools is delayed so that the sea breeze fronts have time to converge. Moreover, performed sensitivity tests to surface heating, wind speed and direction. The results show a strong link between convective available potential energy (CAPE), wind speed and direction and total condensate (sum of all hydrometeors) of Hector cells. The relationship between couples of meteorological parameters was investigated using diagrams that allowed it to be assessed that the convective strength, in terms of vertical velocity, increases as the wind speed decreases and as the wind direction turns toward the major axis of the Tiwi Islands .

In this study, 12 Hector events (from November 1995 to November 2008) are used to investigate the transferability of the conclusions of Crook's study to real events. With this aim, the relationship with the same meteorological parameters used by Crook is investigated for each real Hector event, each one characterized by its own boundary and initial conditions. The case studies are simulated using the MM5 as described in and , and the results are investigated to establish the contribution of water vapor, surface wind speed and direction to the convective strength. These previous works () allow the model's ability to reproduce the dynamics and correctly detect the triggering factors leading to the development of Hector to be assessed, by performing a detailed comparison with observations. However, a temporal and spatial shift is found for MM5. This is also a common issue found for the WRF (Weather Research and Forecasting) model by and . The main focus of this study, as already pointed out, is to investigate the role of a few key meteorological parameters for Hector's development by using Crook's diagrams, which are independent of time. Therefore, a possible temporal or spatial shift in the MM5 simulations of Hector does not affect the results.

The study is organized as follows. A meteorological analysis of the events is presented in Sect. 2 as a function of wind speed, wind direction and shear, CAPE and water vapor and convection modes A or B, with a brief description of the model configuration. The comparison, in terms of cloud total condensate and vertical velocity profiles among the 12 events, is shown in Sect. 3. The fourth paragraph describes Crook's test and outlines the main features in terms of percentage involved in Hector's development. Conclusions are drawn in Sect. 5.

The convective strength of the tropical storm Hector is evaluated using the meteorological variables suggested by . In this study, 12 real events (eight single-cell and two double-cell) are selected. Some of them (20 November 1995; 23 November 1995; double-cell 27 November 1995; 1 and 4 December 1995) were observed during MCTEX, a campaign held in late 1995 on the Tiwi Islands with the goal of monitoring the convective life cycle of the mesoscale convective system (). Four of the remaining events (the double-cell 30 November 2005, 6 February 2006 and 29 November 2007) have already been analyzed by and . The events are named with the acronyms as follows:

20 November 1995: N20;

In order to better understand the mechanisms leading to different convective structures for these Hector events, the vertical structures of the storms are analyzed in terms of total condensate . With this aim, MM5 simulations are used to extract the mean vertical profile of the cloud total condensate (sum of all hydrometeors) and the vertical velocity for each event at the maximum stage (Fig. ). The maximum stage is selected based on the time of the storm maximum height. The profiles are spatially averaged for each layer within the volume encapsulating Hector at a specific time. The maximum of the mean vertical velocity profile is approximately 0.9 ms-1 for N11 (Fig. d), and the largest vertical velocity is 35 ms-1 for the same event. D1, D4, N29, N17 and the first and the second cell, for N27 and N30, respectively, have a maximum updraft value that exceeds 20 ms-1. If it is spatially averaged, it does not exceed 0.6 ms-1. The values of the maximum vertical velocity for N30 are very close to those obtained with the LES simulations by . The maximum updraft obtained by MM5 is approximately 22 ms-1, sustained for a height up to 8–16 km (not shown) and the structure is very close to the one simulated by Meso-NH, using horizontal resolutions of 400, 200 and 100 m (Fig. 3a in ).

The absolute maximum of the total condensate matter is approximately 1.5 gm-3 for the first cell of N27 (Fig. e), but the largest vertical velocity is not reached by this event.

The vertical velocity profiles (Fig. b, d, f) present common features for most of the events. At the lower level, a weak downdraft related to the precipitating hydrometeors prevails, and at the upper level, a very strong updraft can be detected, associated with the latent heat release due to the condensation process. In addition, a downdraft peak in the vertical velocity profile is found at the same level of a relative maximum in the total condensate profile for N11 at 2 km (Fig. c, d), the second cell of N30 at 3 km (Fig. e, f) and the first cell of N27 at 3–4 km (Fig. e, f). These features would suggest that the downdraft is related to the sinking due to melting or evaporation cooling, and the total condensate maximum corresponding to the production of rain or melted graupel. Moreover, the total condensate maxima are at higher altitude than the updraft peaks as for N29, N20 and N11 (Fig. b, d); these events have a maximum updraft at approximately 12 km and show a relatively large amount of total condensate up to 14–16 km (Fig. a, c). However, if the maximum vertical velocity is positioned at lower levels, the most part of the hydrometeors' distribution is at lower levels too. A clear example is the first cell of the N27 (Fig. e, f): the maximum updraft is located at approximately 4 km and the largest part of the hydrometeors is below 10–12 km.

The maximum total condensate and its vertical distribution may be related to the strength of convection, which is generally stronger if generated by the convergence of downdraft of previous cells and the sea breeze front (type B) than the one generated by the convergence of the two sea breeze fronts (type A) . Therefore, larger vertical velocities are expected for type B events; indeed, the largest vertical velocity is found for N11, which is type B. The previous analysis suggests that the strength of convection largely contributes to the vertical distribution of the total condensate. Therefore, the structure of these Hector events agrees with the hypothesis of and allows it to be established that the strength of the event is proportional to the total condensate. However, the large variability of the total condensate vertical distribution among type A and B events suggests that other parameters play an important role beside the strength of convection.

In this section the analysis of the real events is carried out, using Crook's diagrams as a benchmark.

Based on ideal studies of Hector, suggested that the amount of total condensate is strongly related to the low-level moisture, in terms of CAPE, as well as to the surface wind velocity and direction. Therefore, a model-aided analysis of the total condensate and of a few meteorological parameters (wind speed and direction, and CAPE), as used by , may help to highlight the most important factors for previous Hector events. With this aim, the same analysis performed by is applied to the real Hector events analyzed in this study, but some differences are obviously present. The possibility of changing the meteorological parameters, for example, keeping one field constant as is done by , is not applied because its disruptive effect on Hector has already been verified. Indeed, for real events, the variation of a parameter causes a lack in the development of Hector. For example, in the work of , the halving and the increasing of the initial water vapor content disabled the development of Hector.

The following MM5 meteorological parameters are used for the analysis: low-level moisture, in terms of CAPE, surface wind speed and direction. Following , the variables are analyzed at the model start time, which is several hours before the development of Hector. For the single events and for the first cell of double events, the analysis is performed at 08:30 LST, whereas for the second cell of double events, the ending time of the first cell is taken as a reference: 15:10 LST for N30 and 15:30 LST for N27.

All the meteorological parameters are analyzed versus total condensate and mean lifting (vertical velocity at 500 m) as in . The vertical velocity is extracted 3 h before the maximum development of Hector and is averaged all over the island surface. The total condensate, conversely, is averaged within the volume encapsulating Hector at the maximum stage. CAPE, surface wind speed and direction are also averaged all over the surface of the Tiwi Islands. With the aim of understanding the convective response to the flow direction either along the major (90∘) or minor axes (0∘) of the islands, the wind direction is projected in the first quarter of the wind rose.

analyzed the vertical velocity wind with respect to the surface wind velocity and direction (Fig. 7b, c in ), related to shear and no-shear conditions. Hence, Fig.  shows the results by Crook and the Hector events, indicated by gray lines, and 12 symbols, representing the Hector events. A few events closely follow the shear (F6, N20 and the second cell of N27 and N30) and no-shear (D1, N29) lines (Fig. a) or they belong to the right zone (i.e., no shear zone for the first cell of N27), confirming for the shear events the correlation between the decrease of the vertical velocity and the surface wind increase. A few cases do no show any specific signal, probably because the real atmosphere is more complex than the ideal one, and more than one parameter contributes to the vertical lifting, as for example the topography or the convergence line. Therefore, the ideal response can be used to sort the events: the closer the position of the event in the diagram is to the “ideal” one, the more the meteorological parameter contributes to the convective strength.

The second cells of both the double structure events show a stronger surface wind speed than the corresponding first cell; this is due to the gust front associated with the downdraft of the previous convective cell. This is why the double-cell events are not aligned with the others.

A similar analysis is performed for the vertical velocity as a function of the surface wind direction. The results clearly show (Fig. b) that most of the events are located in the right position except for four events: D1, N17, N29 and N23, suggesting difficulties in the real atmosphere to completely separate the two regimes. However, it roughly confirms the increase of the vertical velocity when the flow is eastward (90∘). This supports the hypothesis of greatest lifting when the flow is along the major axis of the island as assessed by and . It is useful to highlight that the wind direction related to the second cells of the double events does not show a clear signal, because at this time, the sea breeze regime is either well developed or destroyed, and “leftovers” from the first cell affect the environment. Therefore, their positions in the graph have an uncertainty larger than the one for the single-cell events.

The analysis of the total condensate versus wind speed (Fig. a) shows that most of the Hector events are aligned along a line with a slope close to that of Crook's study, but an intercept that is smaller (gray dotted line in Fig. a), suggesting a sort of bias between Crook's ideal and the real atmosphere. This disagreement can be explained by speculating that the real events need weaker surface wind than the ideal ones to produce the same total condensate. Based on this hypothesis, a new reference line can be assumed, then only two events appear outside of the distribution, the second events of the double Hector N27 and N30 (Fig. a, white square and white star, respectively). Both events have larger wind speed than the expected one on equal terms of normalized total condensate. The absolute maximum of total condensate is reached by the first cell of N27, whereas the second cell shows approximately a 65 % of the first cell total condensate. On the contrary, for N30 the second cell is stronger than the first one, in terms of cloud total condensate. This is partly due to the leftovers of the previous cell because of the very short time interval (1 h and 20 min) that occurred between the two maxima; whereas for N27, the second cell develops 3 h later than the first one, making the two cells more independent than the previous event. Hence, the hypothesis of the increase of the total condensate as the surface wind speed decreases (Fig. a) is still confirmed, but below 4 ms-1 for all single-cell events and the first cell of double ones. Based on the results of the dry linear and non-linear models, assessed that the relationship between convective strength and low-level convergence (i.e., surface wind speed) is not strictly monotonic because the convective strength did not continue to increase as the flow decreases below 4 ms-1. The 12 Hector events also reproduce a monotonic relationship also below 4 ms-1 (Fig. a). This discrepancy between the simulations of the real events and the Crook experiments is not surprising and it may be due to the differences between the model's assumptions and the use of an idealized sounding in the Crook study.

All Hector events, except the second cell of N27 and N17, confirm the increase of the total condensate if the low-level flow is along the major axes as shown by the plot of the total condensate versus the wind direction (Fig. b), but a spread along the Crook “theoretical” line is found for the real events. As for the mean lifting versus wind direction (Fig. b), the major difference between Crook ideal behavior and real behavior is found for the surface wind, of which direction is close to the minor axes. However, Crook's hypothesis of maximizing the low-level convergence if the flow is aligned along the major axis of the Tiwi Islands is confirmed.

Finally, similarly to what was done by , the total condensate versus the low-level moisture, expressed in terms of CAPE, is analyzed. The Hector events do not show a clear signal, but they show a slight increase of the total condensate as the CAPE increase is found, except for the first cell of N27. This is completely outside of Crook's line (Fig. c, gray square). Moreover, on the contrary to what was found by Crook (Fig. 13a in ), no maximum is found for the total condensate versus CAPE because of the lack of Hector's values around the theoretical maximum; hence it is not possible to assess its occurrence.

In this study, 12 Hector events are analyzed using MM5 model simulations, with the aim of highlighting the main meteorological parameters and their role in triggering convection. A brief meteorological analysis of the events is performed using CAPE, water vapor, wind speed and direction and typology of convection. Moreover, a comparison in terms of mean total condensate vertical profiles and mean vertical velocity at the maximum development is carried out. The applicability of Crook's hypothesis to real cases is explored, verifying the linear relationship between both the convective strength and the total condensate versus the low-level moisture, expressed in terms of CAPE, surface wind speed and direction. Crook's tests allow the following to be concluded.

The strength of convection, in terms of mean lifting and total condensate, increases if the wind direction tends to be parallel to the major axes of the Tiwi Islands and if the wind speed surface decreases.