Frontline Volume 16 - Issue 24, Nov. 13 - 26, 1999
India's National Magazine
from the publishers of THE HINDU

Table of Contents


Scientific failures

The scientific systems whose responsibility it was to predict the contours of the cyclone did a far from perfect job. To be able to do a better job next time around, an integrated approach to cyclone studies is needed.


THE cyclone which crossed the Orissa coast at Paradeep caused unprecedented devastation on October 29 owing to a combination of factors, the most important of them being the nature of the cyclone itself. It was a very-high-intensity cyclone (T-7 in what is called the Dvorak Scale of T-1 to T-8 in meteorological parlance) and is very rare in the northern Indian Ocean region. The cyclone that hit Andhra Pradesh in 1977 was the only other recent case which had associated winds of speeds of over 250 kmph and huge storm surges churning sea water to heights of up to 10 metres.

A unique and crucial feature of the cyclone system that hit Orissa was that it remained anchored over land for more than 48 hours and its intensity continued to remain fairly high, with associated winds (of over 100 kmph) in its vortex. Indeed, the well-formed cyclone eye (the central calm region, which in this case was 49 km across) and the eye wall are clearly visible in satellite images received until October 31. This is, in fact, an unusual feature because normally, as they move over land, cyclonic systems weaken and die out in the absence of energy inflow through moisture-laden surface winds. As the cyclone system was anchored near the coast, surface winds continued to feed moisture into it, resulting in heavy rainfall all across the sweep of the cyclonic circulation extending hundreds of kilometres.

The real meteorological cause of the cyclone remaining stationary for long over land, as if a large-scale atmospheric system was blocking its northwestward motion, is not yet known and will become a subject of a major scientific investigation. In fact, the cyclone track looped back sharply from the anchored spot on October 31 (another rare phenomenon) and a considerably weakened cyclone was seen moving in the south-southeastern direction. This behaviour is indicative of an important characteristic of cyclones. Even though, with the advent of satellite imagery, the genesis of a cyclone can be seen and tracked well in advance, the exact behaviour of the cyclone, including its intensity build-up and landfall, is a product of its interaction with the environmental meteorological forcings. This makes predictions of the track, landfall and life of a cyclone quite difficult. In fact, when the system is weak, the prevalent atmospheric factors around the cyclone influence its behaviour and movement. But once the system gets intensified, it becomes more energetic than the surrounding systems and the cyclone begins to influence the environment strongly.

The history and evolution of the cyclone was as follows:

On October 25 afternoon a "depression" had formed, centred about 550 km east of Port Blair in the Andamans. It moved northwestwards and concentrated into a "deep depression" by midnight. It intensified into a "cyclonic storm" on October 26 morning when it was about 350 km from Port Blair. At this point, the India Meteorological Department (IMD) warned of a potential threat to the north Andhra-Orissa-West Bengal coast arising from a "severe cyclonic storm". The system moved northwestward and intensified into a severe cyclonic storm on October 27 morning when it was about 750 km from Paradeep. Predicting that it was likely to intensify further, the IMD issued a "cyclone alert" to Cyclone Warning Centres at Visakhapatnam, Bhubaneswar and Calcutta. Although the IMD predicted heavy winds and rainfall for October 28 itself, the cyclone remained fixed for six hours at about 600 km from Paradeep and intensified into a "very severe cyclonic storm" (Table on Page 28 gives the IMD's classification of low-pressure systems).

On October 28, the cyclone continued its northwestward movement and intensified into a "super cyclonic storm" - a new term, hitherto not used in meteorological parlance - as the wind speed increased to over 260 kmph as it approached the Orissa coast. The IMD had predicted landfall between Puri and Balasore in the early hours of October 29. Although the cyclone hung over the sea near the coast a little longer than expected, it crossed Paradeep on October 29 between 10 a.m. and 12 noon. After crossing the coast the system hardly moved, and it weakened gradually over the next two days. Somewhat unexpectedly, it curved back and finally moved southwards over the sea along the Andhra Pradesh coast. As the system remained stationary, the IMD predicted both on October 30 and 31 its movement in the northeastern direction over land, which did not occur.

INSAT-1D image of the "super cyclonic storm" (T-7) as it crosses the Orissa coast at Paradeep on October 29.

ALL this might suggest - as indeed the IMD has been claiming - that the forecast of the track and the final landfall location, and the corresponding warnings issued by the IMD, were accurate and that it was only the government's disaster management machinery that failed the people of Orissa. While in the main this seems to be true, the scientific machinery and the predictions were far from perfect and these need to be highlighted. Whether or not these did contribute to the improper communication to the officials in Orissa, and the consequent ill-preparedness to mitigate the disaster, is not clear though news reports from Orissa suggested that even meteorological officials in Bhubaneswar did not have proper information.

For instance, television reports on the night of October 28 and early morning on October 29, quoting an IMD official in Bhubaneswar, said: "The latest reports say that the cyclone centred just 180 km south-east of Paradeep has changed its path slightly and is heading for the West Bengal coast. The cyclone is likely to strike between Puri in north Orissa and Sagar in West Bengal." This apparently caused a great deal of confusion in the minds of the people of Orissa and also led to some complacency in the contingency planning, at least in and around Bhubaneswar. Indeed, this report also led to a rapid, large-scale evacuation from areas around Sagar in West Bengal. The IMD headquarters, however, denies that such a warning or announcement was ever issued. But it is on record that the Bhubaneswar meteorological office did state this to the press, apparently based on a communication from Delhi.

Inexplicably in this age of information super highways, while all the major meteorological centres of the world were putting out real-time information on the Orissa cyclone, including downloadable satellite images from the National Oceanic and Atmospheric Association (NOAA) and Meteosat satellites, there was nothing on the IMD Web page ( or about the calamitous event, not even INSAT images which the IMD receives every half hour. Indeed, there is a link indicated for the cyclone and a click on that only takes you to a United States Navy site. Web sites such as this ( or of the NOAA ( or of the European Centre for Medium Range Weather Forecasting (ECMWF) showed the track of the cyclone in almost real time; the ECMWF and the U.S. Navy even predicted the track and landfall of the cyclone fairly accurately. Incidentally, Pakistan's meteorological department, which does not have a satellite, runs a fairly good Web site using images, analyses and predictions of other centres.

Given this and the apparent confusion in some wings of the IMD about the landfall even less than 24 hours in advance, the question arises as to how accurately the track of a cyclone can be predicted and how far ahead. Given the complexity of cyclone dynamics and a cyclone's highly non-linear interaction with the environment (within a matter of 24 hours a cyclonic system can grow into a very severe cyclonic storm), even state-of-the-art techniques of investigating severe cyclones with powerful computers cannot predict their motion very accurately. But less than 24 hours before landfall, the error in prediction comes down to 50 km, about the size of the cyclone eye itself. So what could have been the reason for the reported announcement of a change in course of the cyclone on October 28?

The circulation in a cyclone usually extends to about 1,000 km horizontally and to about 15 km vertically. Therefore, while the cyclone movement is influenced by the upper atmosphere winds, the feeding of energy and moisture to sustain the cyclone is done by sea surface winds. The initial path is dictated by the dominant currents in the upper atmosphere, known as steering currents, and the presence of easterlies from the Pacific during these months causes the northwestward movement of cyclones over the Bay of Bengal until they become energetic and begin to influence the environment severely.

The methods operationally available to the meteorologist for forecasting the movement are: a) "climate persistence" technique, that is, assuming a continuation of the track already established from previous observations; b) "climatology", that is, normal tracks followed by cyclones in corresponding seasons or previous years; and c) numerical models, using environmental data, including wind speeds and cloud motion vectors of the upper atmosphere from satellite data. In this particular case, as can be seen from the final track, the cyclone followed a path that was just an extrapolation of the evolving track over land. In that sense, from the perspective of prediction, it was a simple and "well-behaved" cyclone.

But if a climatological judgment based on the statistics of past data as well as other meteorological parameters like new strong wind or pressure systems over land had been employed, it could have led to an empirical conclusion that the cyclone was likely to move towards West Bengal. For it is reliably learnt that there were two anticyclones over land, when the cyclone was about to cross the Orissa coast, that would, in general, tend to oppose the cyclone's northwestern motion and weaken it.

As regards numerical models, the National Centre for Medium Range Weather Forecasting (NCMRWF) makes use of the grid model called T-80 (which has a coarse resolution of 150 km as compared to models used by the ECMWF and other centres) and meteorological parameters derived from satellite data. Data on wind vectors, cloud motion vectors, temperature and so on derived from satellite images are important inputs to any numerical model analysis of a cyclone, which is necessary if one has to include all the interactions of the cyclone with the environment in the model calculations.

Interestingly, the upper atmosphere winds as derived by the IMD from INSAT data showed north-easterlies over West Bengal as it normally would have been if the north-east monsoon had become active. If this was indeed the case, then it would have weakened the cyclone system since this would be opposing the cyclonic vortex winds. And indeed, the NCMRWF's analysis had indicated a dissipating cyclone even before it hit the coast. Wind vectors derived from other satellites such as Meteosat of NOAA showed dominant south-easterlies, rather than north-easterlies, which would tend to steer the cyclone in the north-western direction. However, the NCMRWF did not use Meteosat data because apparently Meteosat raw digital data are not received by the IMD.

SO, why did INSAT data show north-easterlies in the upper atmosphere which Meteosat did not? According to NCMRWF scientists, this is perhaps because INSAT-2E, which was launched in April this year, has not been providing any meteorological data for more than two months. (The satellite image accompanying this article has been sourced from INSAT-1D). This hitherto undisclosed information was confirmed by reliable sources in the Indian Space Research Organisation (ISRO). Apparently, the entire Very High Resolution Radiometer (VHRR) has become non-functional because of the failure of the servo motor mechanism in the "scan mirror" assembly. As a result, no data in the visible, IR or water vapour channel region are available from the VHRR, the key instrument of the meteorological payload of INSAT-2E.

Since Dr. K. Kasturirangan, ISRO Chairman, and Dr. S. Rangarajan, Director of the Master Control Facility (MCF), Hassan, are out of the country, no official comments were available. Dr.S.N. Srivastava, IMD Deputy Director-General, and Dr.H.V. Gupta, Director, Satellite Meteorology, declined to confirm or deny the failure of the instrument. Both used the euphemistic phrase that INSAT-2E meteorological payload is still in the "test mode" and is not yet operational. This would be strange for a satellite launched in April 1999.

However, according to ISRO sources, the CCD (charge-coupled device) camera is working perfectly and images from it are available to the IMD. In fact, CCD images are of a better resolution (of 1 km) as compared to the VHRR's 2 km and are particularly suited for tropical cyclone analysis. Unfortunately, some seven months after the satellite launch, the software for analysing CCD images and deriving wind vector data is still not in place at the IMD. So the data used for deriving meteorological parameters are from INSAT-1D, which is nine years old and well beyond its normal life-time. It is on an inclined orbit and any wind vectors derived from it could lead to errors. In fact, if INSAT-2E data were available, it would have included moisture data around the cyclone as well, which would have been extremely useful in forecasting the intensity of the cyclone because it is the moisture-carrying low-level currents that feed energy to and sustain cyclones.

As part of the cyclone warning system along the coast, the IMD has a network of cyclone detection radars (CDRs), which have a range of 400 km (essentially determined by the curvature of the earth). These high-power radars are basically capable of detecting the cyclone eye, the eye wall cloud and the spiral bands of clouds at a distance which an average cyclone would cover in a day. This network also includes an experimental Japanese radar at Visakhapatnam, which can also measure precipitation in the cyclone. Doppler radars (DRs), on the other hand, could provide details of the cyclone's wind field in addition to the usual data from CDRs. Their ability to monitor continuously the wind maxima associated with the wall cloud and the radial extent of the hurricane strength wind field can be of great value to the forecaster.

The track of the cyclone, from October 27, when it intensified into a "severe cyclonic storm" (T-3.5), until November 1, when it weakened into a "deep depression" and finally a "depression" and moved southwards along the Andhra Pradesh coast. The loop back of the cyclone occurred on October 31 near Cuttack. This is based on Meteosat imagery.

But, unfortunately, even though the decision to set up three DRs was taken nearly seven years ago, and money sanctioned immediately thereafter, not even one has been set up. One is being built by ISRO in association with Bharat Electronics Limied (BEL) and is undergoing tests. The other two were to be imported in order to save time, but a German supplier was identified only recently. Meanwhile, the indigenous effort seems to have already come good. Once tested, BEL should be entrusted with producing these in greater numbers to be installed all along the coast, say experts.

All these may not have had any direct bearing on the warnings issued by the IMD. It was perhaps one of those chance events that the cyclone track was a simple one. That in spite of this the administrative machinery simply collapsed in the face of it is another matter. But future ones could be those more complex recurving ones, which would defy such simple extrapolations. The important question of why the cyclone remained anchored over land and then recurved could throw light on new and unexpected dynamic forcings on cyclone. It is possible that the two anticyclones observed over land shortly before the cyclone struck were responsible for this, and the cyclone being very intense just stayed put and could not be diverted.

From the point of view of an operational prediction of cyclone intensity and movement, the lessons of the Orissa cyclone for the scientific community are clear. Sufficient manpower and resources must be directed towards an integrated approach to cyclone studies on a mission mode that would work towards a large-scale model which can be used for operational forecasts. Today, even though the NCMRWF, with its CRAY Supercomputer, is situated on the same campus as the IMD, there is no regular interaction between the two organisations.

The NCMRWF does not get the IMD wind vector data directly from the IMD; it has to wait for them to be put on the World Meteorological Organisation's global data network.

[ Subscribe | Contact Us | Archives | Table of Contents]
[ Home | The Hindu | Business Line | Sportstar ]
Copyrights © 1999, Frontline.

Republication or redissemination of the contents of this screen are expressly prohibited
without the written consent of Frontline.