The northern Indian Ocean is not considered to be one of the world's most active cyclonic basins. On average, every year, five or six systems attain at least the stage of tropical storm with sustained winds over a 1-min period of 35 knots or more (Singh et al., 2001). And, the current Joint Typhoon Warning Center (JTWC, 2009) database shows that on average one intense cyclone forms every 2 years in the northern Indian Ocean. In the current debate on global warming and the change in the number of intense cyclones, initial studies carried out have shown very different results for the northern Indian Ocean. Using the Saffir–Simpson scale (Simpson, 1974), Webster et al. (2005) found that there had been a considerable increase in the number of categories 4 and 5 cyclones with a maximum sustained wind reaching at least 115 knots. Landsea et al. (2006) demonstrated that databases were not sufficiently reliable as cyclones archived as being categories 2 or 3 had been re-analysed and assigned as categories 4 or 5 in the northern Indian Ocean. This has the effect of questioning the trend proposed by Webster et al. (2005). Kossin et al. (2007) did not note any trend towards an increase in the number of categories 4 and 5 cyclones in the northern Indian Ocean for their period of analysis, which covered from 1983 to 2005. To provide other considerations, this article treats the original elements of intense cyclones activity in the northern Indian Ocean from 1980 to 2009 on the basis of a homogenous re-analysis of satellite imagery. Intense cyclones are those that generate sustained winds over a 1-min period of 100 knots or more, being categories 3–5 cyclones. This matches with a current intensity reaching at least 5.5 in the 1984 Dvorak's technique.
2. The indispensable re-analysis of databases
INSAT series geostationary satellites were launched by India above the northern Indian Ocean as from April 1982 (Foley, 1995). However, other countries did not have access to this satellite data (DT). In addition, the thermal infrared images of the Indian satellites only had a spatial resolution of 11 km before 1990 and of 8 km as from that year with INSAT 1D. Before the arrival of the European Meteosat geostationary satellite (1998), the northern Indian Ocean was at the western edge of the sector covered by Japanese GMS satellites on the 140th East meridian and on the eastern edge of the sector covered by European satellites stationed above the Gulf of Guinea in western Africa. Despite a 4-km infrared resolution, GMS and Meteosat (1–4) could not restore the ‘real’ temperature of the warmest pixel in the eye of tropical cyclones. This parameter is very important for estimating the intensity of cyclones from infrared images in cyclonic basins where there is no aircraft reconnaissance. Consequently, it was necessary to use the 4-km infrared imagery resolution of orbiting satellites belonging to National Oceanic and Atmospheric Administration (NOAA). The intensity stage of cyclones is in fact obtained using the Dvorak (1984) analysis that needs to know the highest temperature of the eye and the temperature of cloud tops within a radius of 55 km around the centre. The technique is based on cyclone operational procedures that allow the intensity to be estimated through the intermediary of the maximum sustained wind over a 1-min period.
Kossin et al. (2007) carried out a re-analysis of cyclones across the world for a period from 1983 to 2005 that included the northern Indian Ocean. These authors warned that the aim of their research was to determine whether there was a particular trend and that their work could not determine the real intensity of cyclones. In fact, only images with an 8-km resolution from geostationary satellites were used. In concrete terms, this meant that, for the northern Indian Ocean, the analysis by Kossin et al. (2007) was made without resolving the angle of view problem faced by GMS until now and Meteosat until 1998. The intensity of the cyclones was systematically underestimated as GMS and Meteosat (1–4) gave an eye temperature that was colder than in reality.
Consequently, this research is the first re-analysis concerning the intensity of intense cyclones using satellite data provided by a 4-km resolution. All the intense cyclones in the northern Indian Ocean were analysed using the Dvorak (1984) technique for the period from 1980 to 2009. Prior studies had used the archives of the Hawaii Joint Typhoon Warning Center (JTWC, 2009).
Tropical cyclone 03B provides an example of the need to re-analyse the intensity of cyclones and the method used (Figure 1).
This cyclone, formed in November 1984 in the Bay of Bengal, had been estimated at 85 knots (category 2) by JTWC. An examination of the satellite images reveals that 03B attained its maximum intensity on 13 November 1984 in the middle of the day as it skirted the south-east coast of India (Figure 2). NOAA 7 revealed a system with a small central structure and a perfectly circular eye. The latter corresponds to a warm point of + 18.5 °C surrounded by cloud tops at − 70 °C (white belt).
However, using Dvorak's basic enhancement, it is the ‘black’ belt with cloud tops at − 64/− 69 °C that measured a minimum of 55 km across. The warmer the eye and the colder the cloud tops suggest the greater the cyclone intensity. The application of the Dvorak (1984) technique to 03B reveals a substantial difference between the maximum intensity estimated at 85 knots (category 2) by JTWC and that re-analysed at 120 knots, which ranks this as a category 4 cyclone (Figure 3). On the chart, the satellite DT shows the convection intensity of the system, characterized by the temperature of the eye and the cloud tops. Despite the DT number is displayed every 3 h, the intensity has been estimated every 6 h by JTWC and for the re-analysis.
A plausible explanation for the underestimation of 03B's intensity lies in the fact that only Japanese GMS satellite images with a resolution of 4 km were used. This satellite gave the eye temperature as being − 38 °C on 13 November at 08:30 UTC and − 39 °C at 11:30 UTC. As it was not positioned over the northern Indian Ocean, this geostationary satellite could not reconstruct the highest eye temperature. The NOAA7 orbiting satellite which was at the 03B nadir at 10:14 UTC indicated a pixel at + 18.5 °C in this eye. This difference of 57 °C between the two satellites is sufficient to explain why 03B had an underestimated intensity in the JTWC (2009) archives.
3. An atypical but far from innocuous intense cyclones activity
Apart from the possible approximations in terms of estimating the intensity, it is also necessary to take into consideration those cyclones that moved from one basin to another. For this research, it was decided to assign all tropical systems to the basin in which the maximum intensity had been observed.
The comparison between JTWC DT and those of the re-analysis reveals a considerable difference for the decade from 1980 to 1989 (Figure 4). The Gay (1989) and Forrest (1992) cyclones, formed in the western North Pacific should have been included in the northern Indian Ocean archives as they subsequently became intense cyclones in the Bay of Bengal (JTWC, 2009). Unlike the JTWC best track, the re-analysis does not show a trend towards an increase in the number of categories 3–5 cyclones over the last three decades and in fact the 1990s were twice as active as the other two periods.
Furthermore, it is clear that 30 years of data are insufficient to update a natural activity variation cycle. Despite the peak noted in the decade from 1990 to 1999, the information given in Figure 4 is insufficient to judge whether there is a systematic alternation between an active decade and another that is less so in terms of intense cyclones. A further 30–50 years of reliable data would be needed to better appreciate the changes. In any case, the results, valid here for categories 3–5 cyclones, do not confirm the conclusions of the study carried out by Webster et al. (2005), which state that there was a 600% increase in the number of categories 4 and 5 cyclones in the northern Indian Ocean over the period from 1975 to 2004.
There is another element that worth mentioning, being that the northern Indian Ocean is the basin in the world with the least intense cyclones (Mc Bride, 1995) with, on average, a single system every 2 years over the decades from 1980 to 1989 and from 2000 to 2009 and a phenomenon every year over the period from 1990 to 1999. Despite a lower activity level, extreme intensities are quite comparable with those of other basins. During the ‘record’ 1999 season, of the three intense systems formed (Figure 5), 05B generated sustained winds over a 1-min period estimated at 155 knots, being the peak of the Saffir–Simpson (Simpson, 1974) rating system.
This intensity is similar to that of Katrina or Rita in the North Atlantic in 2005 or that of Monica (2006) and Geralda (1994), respectively in the South Pacific and the southern Indian Ocean.
With only 30 years of reliable data for the northern Indian Ocean, it is very difficult to observe any significant trend towards an increase in extreme intensity: Gay (1989) 140 knots, 05B (1999) 155 knots, and Gonu (2007) 145 knots were respectively the most intense cyclones of the last three decades. However, there is no reason not to believe that comparable or even more powerful cyclones may have taken place in the past.
The inter-annual distribution reveals a high level of irregularity (Figure 5). Over a period of three decades, 13 years did not have categories 3–5 cyclones. In the 1990s, only 1993 did not have any intense cyclone. The maximum of three intense cyclones in 1999 appears low when compared with other basins across the world. The fact is that the North Atlantic with 7 major hurricanes in 2005, the western North Pacific with 12 in 1997, as well as the southern Indian Ocean with 6 intense cyclones in 1980, the South Pacific with 6 intense cyclones in 2003 and the northern East Pacific with 10 major hurricanes in 1992, all show a level of activity at least twice as great as that of the northern Indian Ocean. The statistics mentioned here are official for the different basins.
In the northern Indian Ocean, the connection between the number of intense cyclones and a hydro-climatic phenomenon such as El Nino is not self-evident. The oceanic Nino index (ONI) is the standard that NOAA uses to identify El Nino and La Nina events in the tropical Pacific (CPCM, 2010). It is the running 3-month mean sea surface temperature anomaly for the Nino 3.4 region (5°N–5°S and 120°–170°W). Events are defined as five consecutive months at or above the + 0.5 °C anomaly for warm events (El Nino) and at or below the − 0.5 °C anomaly for cold events (La Nina). Weak, Moderate and Strong events are those with an anomaly of 0.5–0.9, 1.0–1.4 and 1.5 or above, respectively. In the northern Indian Ocean, among 21 intense cyclones, only 4 formed during El Nino events, 8 during La Nina events, and 9 were associated with neutral conditions (ONI between − 0.4 and + 0.4 °C). Also, it is interesting to notice that 13 intense cyclones formed when the ONI was negative, 6 when the ONI was positive and 2 when the ONI was at zero. The intense cyclones formed more frequently during La Nina events or under neutral conditions associated with a negative ONI. The record of three intense cyclones in 1999 took place during a strong and long La Nina event. Chang Seng and Jury (2010a, 2010b), who studied the intense cyclones (90 knots over 10 min or 100 knots over 1 min) in the south-west Indian Ocean, found also that La Nina event was a governing factor.
Another indicator is not in accord with the conclusions found by Webster et al. (2005). In fact, the proportion of categories 3–5 cyclones did not increase in a steady manner in all cyclones having reached an intensity of 65 knots and more over the last three decades (Figure 6).
The proportion of intense cyclones reached a peak of 44% during the 1990s and, a considerable reduction can be seen since 2000 with 31.25%. Consequently, no potential global warming influence can be noted here as Webster et al. (2005) have demonstrated that the northern Indian Ocean had been at its warmest during the 2000s.
Apart from the low annual number of cyclones when compared with other basins across the world (Mc Bride, 1995), the characteristics of the northern Indian Ocean lies in the unique bimodal distribution of its activity (Figure 7).
The first part of the season concerns the quarter from April to June, with a peak of seven intense cyclones in May. The second seasonal cyclone activity occurs in October to November with a maximum of eight intense cyclones in November. No intense cyclones were observed during the July to September quarter between 1980 and 2009. In fact, maps produced by the University of Wisconsin show that this period, corresponding to the middle of the monsoon season, is characterized by a strong vertical wind shear preventing storms from becoming intense cyclones (Krishna, 2009). The south-west flows of the lower and middle troposphere are crowned by winds with an easterly component in the upper troposphere. In addition, from July to September, the monsoon trough responsible for the tropical storms genesis is positioned near or on the landmass. Five of the 21 intense cyclones in the northern Indian Ocean are formed in the Arabian Sea which has its greatest activity level in May and June. However, although it has a smaller ocean surface area, the Bay of Bengal generates three times more categories 3–5 cyclones. To attain the high intensity stage, 17 of the 21 cyclones in the northern Indian Ocean (81%) developed rapidly over a 24-h period (Figure 8). This means that an increase occurred in the speed of the sustained wind over a 1-min period of at least 35 knots, being from 65 to 100 knots. This minimum threshold is considered as representative in the Dvorak (1984) method.
To assess the intensification speed of tropical cyclones, DeMaria and Kaplan (1999) for the North Atlantic and Holliday and Thompson (1979) for the western North Pacific used the drop in the central atmospheric pressure measured or estimated in the eye. However, the dropsondes used since 1997 for aerial reconnaissance in the North Atlantic have permitted better wind measurements. These dropsondes revealed that for an identical central pressure, two cyclones could have very different sustained winds (Hock et al., 1999). This very clearly signifies that the wind is the most significant parameter for evaluating cyclone intensity variation.
By increasing from 70 to 145 knots in 24 h, being a 75 knots wind increase, Gonu (2A in June 2007) intensified in a remarkable manner in the Arabian Sea. This value, which represents a record for the northern Indian Ocean over the period from 1980 to 2009, is comparable with the values estimated in cyclones having taken place in the southern Indian Ocean, South Pacific and northern East Pacific. Only the North Atlantic and the western Pacific had greater values with a 95 knots increase in 24 h for Hurricane Wilma (October 2005) and Typhoon Forrest (September 1983). Fourteen systems developed rapidly in the Bay of Bengal and three in the eastern Arabian Sea, including Gonu (Figure 9).
Nine of the 17 cyclones completed the rapid intensification process at less than 200 km from the coast. An additional intensification but at a more moderate rate can continue up to landfall.
4. Intense cyclones considerably influenced by nearby land masses
Intense cyclones considerably influenced by nearby land masses 13 intense cyclones attained their intensity peak at less than 200 km from the coast (Figure 10). This represents a majority of 62% for which it is possible to believe that the proximity of land could have impeded intensification. Chang Seng and Jury (2010a, 2010b) found that Madagascar could also block the flow converging towards the cyclone in the south-west Indian Ocean. The fact is that when the cyclonic circulation interacts with a landmass, there is less humidity and therefore less energy in the central part of a tropical system. However, relatively small cyclones can continue to intensify as they approach coastlines. This was the case for cyclone 03B in November 1984 which intensified by 40 knots over 24 h before attaining category 4 along the coastline near Chennai (Figures 1 and 2). The very cold cloud tops and the excellent organization of the structure in the upper troposphere lead one to suppose that 03B would have attained category 5 had the distance from the coast been any greater. This is why it is always difficult to make significant comparisons, in this case between the extreme cyclones to be found in the northern Indian Ocean. What can be stated is that two category 5 systems were formed in each of the three decades.
Despite a fundamental role in certain cases, the sea surface temperature is not always a determining factor governing intensity. In fact, the correlation coefficient at 0.018 does not indicate any link between the maximum intensity of cyclones and the sea surface temperature in the northern Indian Ocean for the period between 1980 and 2009 (Figure 11). Only 3 of the 21 intense cyclones, being 14.3%, approached their maximum potential intensity defined by the energy available in a given oceanic space (Emanuel, 1988): the three cyclones had an intensity of 125, 140 and 155 knots, over a sea surface temperature, respectively of 27.7, 28.2 and 28.7 °C. This concept assumes that there are no thermodynamic constraints in the atmosphere and that conditions are almost ideal. Chang Seng and Jury (2010a, 2010b) found that the intense cyclones in the south-west Indian Ocean depended mainly on the warm sea surface temperature and the favourable vertical wind shear.
Working on the North Atlantic, DeMaria and Kaplan (1994) found that only 20% of hurricanes (65 knots and more) attained at least 80% of their maximum potential intensity. For the western North Pacific, Baik and Paek (1998) advanced a proportion of 37%. This clearly signifies that factors such as a strong vertical wind shear between the lower and the upper troposphere, the intrusion of dry air or the weakness of the divergence in the upper troposphere have a considerable influence in limiting the intensification of cyclones (Merril, 1988). In the northern Indian Ocean, if one simply considers cyclones having reached a minimum of 65 knots, only 5.66% (3 of 53) approached or attained their maximum potential intensity over the last three decades. In addition to the limiting factors mentioned above, it is clear that the influence of landmasses represents a considerable element for explaining this very small proportion. More than elsewhere, the considerable presence of landmasses largely explains variations in the sea surface temperature over the months (Figure 12). The northern Indian Ocean is closed off from subtropical latitudes by the southern tip of the Asian continent. As a result, the sea surface reaches a very high temperature much earlier than in most of the other basins. Along with the western North Pacific, it is the only basin where a category 5 cyclone, 02B (1991), formed as from the month of April (there is no equivalent in October for the southern hemisphere). On average, the northern Indian Ocean is at its warmest in May when the water temperature exceeds 29 °C over large areas (Figure 12). It is the month of the first peak for intense cyclones. As from June, the monsoon flows begins to considerably cool the ocean and, in August, there exists an upwelling accompanied by masses of cool water in the western part of the Arabian Sea (Benestad, 2009). As soon as the monsoon ‘ends’ around the end of September, the ocean begins to slowly reheat, although it does not reach the sea surface temperatures to be found in the month of May and the vertical wind shear diminishes. Chronologically, November corresponds to the second peak of intense cyclones, although quantitatively to the first peak, with eight intense systems as opposed to seven in May. Globally, the sea surface temperature remains far more favourable in the Bay of Bengal than in the Arabian Sea. In addition, in the latter, the intrusion of dry air limiting convection is more frequent than in the Bay of Bengal which is better ‘protected’ from the continental air mass by the Himalayan barrier. This somewhat explains why there were three times more intense cyclones in the Bay of Bengal over the last three decades. However, this does not exclude the intensification of a cyclone to category 5, as was the case with Gonu on 4 June 2007. This very powerful system was presented by the media as being the result of global warming, but the fact is that on average (1971–2000), the central and eastern parts of the Arabian Sea a sea surface temperature greater than 29 °C in May and June (Figure 12). This high threshold should have resulted in a larger number of intense cyclones than the five that developed between 1980 and 2009. Incidentally, Hoarau (2001) cites the case of the Daniela cyclone which reinforced explosively up to 125 knots in the south-west of the Indian Ocean in December 1996 above oceanic masses having a temperature profile of 27 °C. The case of the Arabian Sea confirms that the sea surface temperature alone is insufficient to comprehend the intensity of cyclones (Lal, 2001).
The considerable presence of landmass in the northern Indian Ocean is also translated by a considerable number of landfalls by intense cyclones. Among the 21 systems formed, 16 made a landfall at an intensity at least equal to 100 knots (Figure 13).
This decadal distribution, true to that of the number of intense systems (Figure 4), did not increase on a regular basis between 1980 and 2009. A reduction in the number of cyclones landfalls' was obvious in the 2000s.
This was accompanied by a reduction in the number of intense cyclones when compared with the decade from 1990 to 1999. Four countries in the northern Indian Ocean were affected by intense systems over the last three decades: India, Bangladesh, Burma and Pakistan (Figure 14). While India is the country most frequently affected due to the length of its coast, it is remarkable that it has been spared in the 2000s. Despite the same number of five intense cyclones formed in the northern Indian Ocean during the two decades 1980–1989 and 2000–2009, India suffered three intense cyclones ‘landfalls’ in the 1980s. It is also worth mentioning the case of Burma as the Nargis cyclone, which killed over 135 000 people.
The fact is that this catastrophe took place in a highly vulnerable and densely populated region and the archives cannot provide any comparable examples over the last 30 or 40 years. Like Gonu, Nargis formed in a period (2000–2009) characterized by a reduction in the number of categories 3–5 cyclones when compared with the previous decade.
While countries such as the Sultanate of Oman, Somalia or Yemen have not been concerned by intense cyclones since 1980, this possibility cannot be ruled out despite the relatively unfavourable thermodynamic conditions generally to be found along the north and west coasts of the Arabian Sea.
This study highlights the particularities of categories 3–5 intense tropical cyclones in the northern Indian Ocean over the last three decades. The re-analysis of intensity using satellite images indicates that 21 intense cyclones were formed over a 30-year period. The decadal distribution does not reveal a regular trend towards an increase in the number of these cyclones despite a doubling of activity in the 1990s when compared with the 1980s and 2000s which were wholly comparable. Nor does new data reveal a continuous increase in the proportion of intense cyclones among all cyclones (categories 1–5) over the three decades studied. The multi-year analysis shows that 13 intense cyclones formed when the ONI was negative. Eight intense cyclones formed during La Nina events, whereas only four intense cyclones formed during El Nino events. Associated with a weak vertical wind shear, tropospheric conditions of La Nina events could intensify the cyclones further. The 1999 year, a strong and long La Nina event, was the most active season with three cyclones having attained at least the category 3. With 155 knots in October 1999, 05B reached the highest intensity estimated for a cyclone of the northern Indian Ocean in the 1980–2009 period.
Apart from the small annual number of cyclones when compared with the number in other basins across the world, the particular uniqueness of the northern Indian Ocean lies in the bimodal distribution of its activity. The first part of the season concerns the quarter running from April to June with a peak of seven intense cyclones in May and a second occurring in October–November with a maximum of eight intense cyclones observed in November. No cyclones were noted during the July to September quarter between 1980 and 2009. This is because this period, which corresponds to the middle of the monsoon, is characterized by a strong vertical wind shear preventing storms from becoming intense cyclones. Four of the five cyclones formed in the Arabian Sea developed in May and June. The subsequent monsoon winds coming from a south-westerly direction cause a notable cooling in the sea's temperature in the north-west part of this region. Seventeen of the 21 cyclones reached a minimum of category 3 following a rapid intensification phase of at least 35 knots in 24 h. As in other basins, favourable tropospheric conditions can facilitate the accelerated development of storms. The small size of the northern Indian Ocean provides an understanding as to why 13 of the 21 intense cyclones (being 62%) reach their maximum intensity at less than 200 km from the coast. This interaction with the landmass explains why the large majority of cyclones do not attain their maximum potential intensity despite the presence of particularly warm oceanic masses. It is also worth underlining another remarkable aspect, being that 16 of the 21 intense cyclones retained the 100 knots stage at the moment they penetrated the landmass. However, there has not been a regular increase in the number of cyclones ‘landfalls’ over the last three decades (1980–2009). While India and Bangladesh are the countries most affected, relatively spared countries, such as the Sultanate of Oman, Somalia and Yemen, could suffer the assault of an intense yet small cyclone in May or June if the thermodynamic conditions temporarily became favourable in the western part of the Arabian Sea.
The current period of 30 years for which we have reliable data is too short to pick out natural cycles in the decadal variations of intense cyclones activity. It appears essential that a few additional decades of observation are needed for these cycles to be understood.
I would like to thank the reviewers. Their comments and suggestions have been a precious help to improve this article.