Forecasting travails

The coming together of ISRO and the IMD, which has so far given more credence to outsourcing, to solve problems jointly augurs well for meteorology in the country.

R. RAMACHANDRAN in New Delhi

"FOR precise estimation of heavy rainfall from convective clouds, high-resolution visible channel imagery provides very important clues. In particular, the overshooting cloud tops, which play a vital role in heavy rainfall leading to flash floods, can be very well observed in high resolution imagery. Such observations are useful inputs to the heavy rainfall estimation models and are helpful in improving the accuracy of intense precipitation advisories being issued operationally by the department. This has useful applications in flash flood forecasting."

The above is a remark made by scientists of the India Meteorological Department (IMD) in a paper that appeared in the journal Current Science on the utility of high-resolution (1 kilometre) Charge Coupled Device (CCD) camera on-board the satellite INSAT-2E launched in April 1999 for intense rainfall studies. This information in near real-time, coupled with the images of water vapour plumes using the water vapour channel of the Very High Resolution Radiometer (VHRR), the other onboard camera with an 8-km resolution, should have provided critical information for timely alerts during the record precipitation over Mumbai on July 29 (Frontline, August 26).

In any case, at least for a post-event diagnosis, these should enable insights into the genesis of the unprecedented rainfall. According to J. Srinivasan, an atmospheric scientist at the Indian Institute of Science (IISc), Bangalore, the wind vector data from the high-resolution CCD camera - which has imaging capability in the visible, near infra-red and short-wave infra-red bands - can give information about the small-scale (10-20 km) convective structures and the atmospheric system that caused the heavy precipitation.

But with the software for the INSAT Meteorological Data Processing System (IMDPS) installed at the IMD in disarray, no useful products are being derived. In fact, even the basic cloud imagery from the CCD camera is not being obtained. At present there are three satellites with meteorological imaging payloads: INSAT-2E and 3A with VHRR and CCD camera and METSAT/Kalpana-1 with only VHRR. Of these the VHRR on INSAT-2E was rendered non-functional soon after launch owing to some anomaly in the instrument scanning mechanism. At present, apparently only the routine VHRR images (taken every three hours) from INSAT-3A and Kalpana-1 are being disseminated by the IMD. Most researchers, therefore, prefer to work with images and products derived from other satellites such as Meteosat, which are available online and in quick time. According to many, the IMD has never released CCD images since the camera was launched in 1999.

In fact, the IMD's unpreparedness in this respect had its telling impact when the super cyclone hit Orissa in October 1999. The failure of the VHRR had been compounded by the lack of software for the CCD camera. As a result, the cyclone could not be tracked properly even though tropical cyclone analysis was one of the prime objectives of the CCD camera. Data from other sources were relied upon, resulting in wrong landfall prediction and inappropriate warnings (Frontline, November 26, 1999). The sad part is that even today proper IMDPS software is not in place and CCD images are not being issued or are used by the IMD on an operational mode.

"There have been problems with the software," admitted V.S. Ramamurthy, Secretary, Department of Science and Technology (DST), which is the IMD's parent body, at a press conference in early August . "An offer from a scientist at Wisconsin University, Madison, United States, who has expertise in this field, has come. We are looking into that," he added.

The contract for the implementation of the INSAT-2E IMDPS, according to the IMD, was given (at a high price of over $5 million) to the Montreal-based company Info-Electronics Systems (IES) based on a global tender. (Although the IMD maintains that the contract was only for 2E, there is mention of a similar job for 3A as well on the company's web site but without any details.) However, CCD data from INSAT-2E were apparently contaminated and so could not be processed to give proper images and products therefrom, say IMD officials. As for INSAT-3A, the IMD says that its software group had tried to tune the 2E software for 3A but, since there was some bug in the IES software, it has not succeeded.

The software vendor too has maintained that there was data contamination, particularly in the infra-red, across all the pixels of the CCD. The company also has contended that bugs could have arisen as the Indian Space Research Organisation (ISRO), the satellite fabricator, did not provide the company with the complete camera sensor details for proper software development. But ISRO disagrees on both counts. ISRO scientists point out that this was the first-ever CCD camera in a geo-stationary orbit and there may have been some problems with probably a maximum of 10 per cent of the pixels, which was not something good software could not have handled. As for camera details, ISRO maintains that no satellite manufacturer gives out the proprietary sensor model.

"I cannot get the Ikonos sensor model if I want it," an earth observation scientist at ISRO said. "What is supplied to software vendors are what are known as Rational Polynomial Coefficients (RPCs) that approximate the camera model and are sufficient for the purpose of image processing software development. These are used in a mathematical construct in the form of an equation that includes the necessary radiometric and geometric corrections as well," the scientist added. All the relevant information, according to the IMD, was supplied to the software vendor as part of the contract in the form of a `System Interface Document'.

"VHRR processing is simpler than CCD processing. When a CCD array in a geo-stationary orbit is subject to fast east-west and slow north-south scan, the geometrical corrections are complex," pointed out a scientist associated with INSAT operations. "It is not just giving a few coefficients. Camera model is just one thing. What about the mathematics of the complicated scanning geometry and mosaicking of the different strips in the north-south direction? There are also requirements for inputting on-orbit information like radiometric behaviour of the CCD elements and platform behaviour when the Earth sensors get into a problem. I wonder whether the IES system had provisions for some of these," he added.

"If IES were to provide images on an operational basis, it was absolutely necessary to have involved the Space Applications Centre (SAC) right from the design phase. The SAC should have been the natural choice for implementing the IMDPS as it had full knowledge of the payload as well as expertise in setting up ground systems. In fact, the SAC had built the system for testing the payload at Kourou (French Guiana) and later provided the `quick look system' to the Master Control Facility (MCF)," said a scientist at the ISRO headquarters.

In effect, therefore, CCD data do not seem to have been utilised for meteorological applications at all. In fact, some scientists even doubt if the IMD has even archived past data, which could be studied for analysing past extreme events. But, interestingly enough, despite the problems in the data processing system, the IMD has maintained the following at the international forum of the Consultative Group for Meteorological Satellites (CGMS): "The imaging mission is working satisfactorily with METSAT (Kalpana-I) satellite and INSAT-3A, and they continue to be used operationally. High-resolution (1 km) images in 3 channels are also available operationally from INSAT-3A and INSAT-2E CCD cameras. Activities like image processing, derivation of meteorological products, data archival and dissemination of products to field stations for operational use are being done on a routine basis."

Why did the IMD go for outsourcing in the first place? "We do not have the expertise within the country to process high bit rate data and produce images in near-real time," said an IMD official who did not want to be named. "In fact, no Indian company responded to our tender," he added. This argument is, of course, rejected by ISRO scientists. ISRO itself has been receiving the data and processing them for its own application needs, including meteorology. While ISRO's chief use has been in using the CCD data for agricultural remote sensing applications such as vegetation index, it has also been using the data for meteorological applications, including model-based forecasting.

"The CCD was ISRO's auxiliary payload to provide localised higher resolution imagery as a partial response to the IMD's insistence on a larger VHRR providing high resolution over the globe. As a bigger VHRR would not fit on to the multi-mission INSAT bus and would have required a major new and costly four-year development cycle, the SAC chose the CCD route to partial high-resolution imagery. This also went in tune with the SAC's objective to obtain vegetation index data since CCD data from the French satellite SPOT were no longer available," said a former scientist of the ISRO Satellite Centre (ISAC) in Bangalore.

"In my opinion, the IMD never took this payload seriously. As ISRO had a lot of CCD experience from the polar orbiting Indian Remote Sensing (IRS) satellites, they decided to make an instrument for INSAT and the IMD treated this like an R&D (research and development) effort. The CCD camera on INSAT is an example of how technology was thrust on the end-users; even within the IMD there was no widespread discussion on its usage," the scientist added.

While agreeing that it was an R&D effort, SAC scientists say that the IMD had expressed great interest in the new imaging instrument. "In fact, IMD scientists had written papers on the potential uses of the CCD camera. Without the consensus of the user agency, no launch operation can take place," they point out.

Inappropriate software is also perhaps the reason for errors in the Cloud Motion Vectors (CMVs), the simplest of products derived from satellite images, derived from INSAT data. This is not just with regard to INSAT-2E or METSAT or INSAT-3A. It apparently has been so for long. As a result, the international research community, it is learnt, has never really relied upon the IMD's wind vector data even though they are routinely put on the Global Telecommunication System (GTS) of the World Meteorological Organisation (WMO).

SO why is the IMDPS in such a mess? Sources in ISRO point to the outsourcing of software by the IMD. The truth of the matter, however, seems to be that, for reasons best known to the two organisations, the disconnect between ISRO and IMD runs deep and has been so for long, perhaps right from INSAT-1 days. The INSAT Coordination Committee (ICC) - which comprises basically the concerned department secretaries - and the INSAT Technical Advisory Group (TAG) do not seem to have addressed this important issue - which ultimately has implications for the public service that the IMD is meant to provide - seriously enough and evolved solutions.

According to former ISRO personnel, for the INSAT-1 series, the satellite contractor, namely Ford Aerospace, designed the quick look data processing system. The SAC did the ground station and a collaborative venture of Computer Maintenance Corporation (essentially for system maintenance) and the Canadian image processing software company MacDonald Dettwiler Associates (MDA) implemented the processing module. Here perhaps the IMD preferred a foreign vendor because of inexperience within ISRO itself. However, the ICC stepped in so that a trilateral agreement among ISRO, the IMD and CMC-MDA was in place and Ford/SAC provided all the necessary technical inputs such as RPCs and other on-orbit details for VHRR (which then had visible and infra-red channels only).

Around INSAT-2A's time in 1992, the CMC-MDA collaboration broke up because of key people leaving, but since INSAT-2B was identical to 2A there was no immediate problem for the IMD to continue using the IMDPS developed by CMC-MDA. The following satellites, 2C and 2D, did not have any met payloads. So there was no immediate crisis situation.

Around 1993-94, with the imminent development of an improved three-channel VHRR and a CCD camera for INSAT-2E, which would require modified IMDPS, P.P. Kale, former Director of SAC, put together a team to develop the necessary software at SAC and offered to implement the same for the IMD. But, apparently, the IMD wanted to go on its own and ISRO was told not to interfere in the IMD's operational work.

ISRO too apparently did not press, as its first priority was to get the IMD to accept CCD-based high-resolution data. Also, high on ISRO's agenda for INSAT-2E was the leasing of 11 transponders to INTELSAT. Interestingly, while SAC had established the ground station for the IMDPS right up to INSAT 2B/2C, even this part was given to the foreign vendor. While this could be justified since many experienced vendors were in the market for this job, SAC's involvement for the development of IMDPS software was important, which the IMD failed to realise. It brought in IES, another Canadian company. Since then successive Chairmen of ISRO seem to have taken the attitude that ISRO was a mere satellite builder, leaving the utility of the satellite to the end-user agency.

Now faced with a crisis situation, the IMD has modified, it is reliably learnt, the CMC-MDA software to process INSAT-3A and METSAT/Kalpana-1 data. Today the expertise available within ISRO is vastly different and, besides developing software for the IMDPS, ISRO has evolved expertise in many areas associated with meteorology. ISRO has already developed an automatic weather station: 15 AWSs are currently undergoing prototype evaluation at various locations. The IMD has plans to augment its surface observing instruments with 100 such stations but had planned earlier to import them. The performance of imported AWSs in Andhra Pradesh - with inadequate maintenance and operational support from the vendor - should be a reminder before shopping overseas. Similarly, ISRO has developed a Doppler Weather Radar (DWR) for use in cyclone detection and tracking (Frontline, August 16, 2002). One of them has been installed at Sriharikota (Andhra Pradesh). However, having imported four others, which have been installed along the eastern coast, it has been unable to take a decision on the remaining DWRs along the coast.

However, things seem to be changing, in particular for the IMDPS. "I have told the IMD that there would be no more outsourcing when you have domestic expertise, and after all it is ISRO's satellite," said Ramamurthy. A joint working group, under K.L. Majumder of SAC, has been established to arrive at a mechanism by which ISRO and the IMD will be able to work jointly in implementing the IMDPS, immediately for INSAT-3A and Kalpana-1 and later for the next dedicated meteorological satellite, INSAT-3D - the fifth in the INSAT-3 series - with advanced features (which is due for launch next year). To begin with, IMD personnel will go to SAC to familiarise themselves with ISRO's software and its working, which could later be implemented for INSAT-3A and Kalpana.

INSAT-3D, of course, would require development of IMDPS software de novo because of its several new features. This satellite is designed to launch dedicated meteorological satellites of a new era with much improved capabilities as compared to all earlier satellites of the INSAT series. It will carry a multi-channel high-resolution imager and a 19-channel atmospheric sounder for providing improved atmospheric observations. This satellite will provide capabilities similar to the U.S.' GOES-A satellite.

The coming together of ISRO and the IMD to solve problems jointly augurs well for meteorology in the country. With appropriate software for the processing of INSAT-3D data, coupled with ISRO-built AWSs and DWRs, the IMD's forecasting abilities should greatly improve in the future.