Growing disconnect

G. Madhavan Nair, ISRO Chairman (right), taking a look at the automatic weather station after inaugurating it at the National Atmospheric Research Laboratory at Gadanki near Tirupati. A 2005 picture.-G. Madhavan Nair, ISRO Chairman (right), taking a look at the automatic weather station after inaugurating it at the National Atmospheric Research Laboratory at Gadanki near Tirupati. A 2005 picture.

India Meteorological Department prefers imported software and subsystems when domestic capability exists, particularly with ISRO.

Budget allocations for 2007-08 and 2008-09 indicate that the massive modernisation plan of the India Meteorological Department (IMD), now under the Ministry of Earth Sciences (MoES) created in 2006, is slowly getting under way. Announced in 2005 by Union Minister Kapil Sibal, when the IMD was under the Department of Science and Technology (DST), as a project costing about Rs.500 crore spread over two to three years, the plan was later revised upwards to Rs.700 crore and now it is valued at about Rs.900 crore. The project, which also interestingly includes airport modernisation involving a large investment of around Rs.250 crore towards IMD services for aviation purposes, is to be executed in two phases.

In phase-I, besides internal restructuring of departments and personnel, the proposed modernisation appears to involve only huge purchases of equipment and instruments, such as high performance computing (HPC) systems (generally called supercomputers), Doppler weather radars (DWRs), automatic rain gauge network and automatic weather stations (AWSs). The Budget document itself says, The necessary approvals for launching of the Phase-I programme have been obtained and the Ministry is in advanced stage of placing the order for acquisition of the various components towards modernisation of IMD.

It was perhaps because of the delay in obtaining the necessary approvals of the Cabinet and the Department of Expenditure that last years allocation of nearly Rs.240 crore was revised downwards to Rs.25.5 crore during the midterm revision. This year the allocation for the purpose has, in fact, been increased to Rs.364 crore, which effectively implies a total investment thus far of Rs.400 crore for phase-I modernisation. Correspondingly, the tenders for items of high value included in the first phase four HPC systems, 12 DWRs and 550 AWSs were floated roughly around the same time, mid-2007.

In principle, the IMD modernisation plan affords a great opportunity for the developers of indigenous technologies in the area of meteorology. Unfortunately, evidence on the ground is to the contrary. One organisation that is particularly disappointed by this emerging scenario in the procurement of systems by the IMD is the Indian Space Research Organisation (ISRO), which in recent years has made forays into various aspects of weather and climate research and related technology development in a major way, including numerical weather forecasts, short of actually being a meteorological service organisation. ISROs entry into meteorology is clearly a fallout of its experience with using data from INSAT and the Indian remote sensing (IRS) systems of satellites. Many INSAT payloads have included a meteorological component specifically for the use of the IMD. METSAT or Kalpana-1, in fact, is a dedicated met satellite.

As noted in an earlier article (Frontline, October 7, 2005), despite ISROs regular involvement as a launcher, and payload and systems integrator, for the IMDs requirement of met data, there has been a growing disconnect between the two organisations over the years, with the IMDs apparent preference to import associated software and subsystems even where domestic capability exists, particularly with ISRO. Given that the current Secretary of the MoES, the parent Ministry of the IMD, is a former ISRO satellite specialist, this irony becomes particularly glaring.

A case in point in the present context is the procurement of 12 S-band (2.7-2.9 giga Hertz) DWRs, the component involving the highest expenditure in the plan. According to the tender document issued by the IMD on July 30, 2007, the tender value for this which is an estimate is placed at Rs.410 crore, implying the cost for one DWR system (complete with computer, networking and communication systems, tools and testing equipment, spares and consumables for a period of one year as well as training three IMD personnel for a period of 12 weeks at the suppliers premises) to be around Rs.35 crore. According to the document, six radars are to be delivered within six months of the date of opening of line of credit (LC) and the remaining within nine months of LC.

DWRs use the Doppler principle of frequency shift in signals (transmitted in pulses from the radar system) as they are reflected by moving weather systems and received back by the radar antenna to give information about weather systems up to distances of 300 km to 500 km. Essentially, they give two primary products, cloud reflectivity and system radial velocity. From these, other derived products such as precipitation intensity, rainfall accumulation and vertically integrated liquid water content are generated. (Unlike Doppler radars, conventional weather radars cannot give velocities.) Based on variations in amplitude (or power received), phase or frequency shift, change in polarisation state of the reflected electromagnetic waves, the radar is able to quantify the different characteristics of the precipitating systems. DWRs are particularly useful as early warning systems for severe weather systems such as storms and cyclones.

The IMD plans to establish a network of these across the country, in particular along the coasts, and replace the existing conventional S-band radars (Frontline, August 16, 2002). At present, the coastal network of radars (see map) includes five DWRs at Chennai (2002), Kolkata (2003), Sriharikota/SHAR (2003), Machilipatnam (2004) and Visakhapatnam (2006) all of which are being used by the IMD. Except for the one at Sriharikota, which has been designed by ISRO and fabricated by Bharat Electronics Ltd. (BEL), the rest are imported systems, made by a German firm Gematronik GmbH. The 12 DWRs that are being procured as part of the modernisation plan will be located at Delhi, Goa, Patiala, Agartala, Mohanbari, Patna, Lucknow, Mumbai, Paradip, Karaikal, Bhopal and Nagpur.

As of date, the technical evaluation of the bids there were several bidders, including the government establishments BEL and Electronics Corporation of India Ltd. (ECIL) has been completed by the Technical Evaluation Committee (TEC) and two suppliers have been short-listed. The bids included only one indigenous DWR, that of ISRO-BEL, called MEGHA 2700. (The system offered by ECIL, a unit under the Department of Atomic Energy, is based on technology acquired by it recently from the U.S. firm Radtec Engineering Inc.) The ISRO-BEL system was developed as a multi-institutional R&D effort involving many R&D labs, academic institutions and industry under a project funded entirely by the IMD to the tune of Rs.9 crore in 1996 and completed in 2002.

Here lies the unfortunate rub that hurts ISRO. According to IMD sources, the indigenous system MEGHA 2700 stands rejected by the TEC in preference to METEOR 1500S, offered by Gematronik (four of which are already installed in the country), and WSR-98D offered by Beijing Metstar Radar Co. The final award of contract depends on financial evaluation (for the lower bid) between the two, which remains to be done. Interestingly, the original TEC included a member from ISRO. But just a day before the meeting that decided on the short-listed suppliers, the TEC was apparently reconstituted and the only change was the removal of the ISRO member!

The WSR-98D radar is basically American technology and is an improved version of the original WSR-88D radar developed by the U.S. company Lockheed Martin, 165 of which have been installed as part of the NEXRAD weather radar network in the U.S. However, now Beijing Metstar, the actual bidder for the IMD contract, manufactures it. The company was formed in 1996 as a joint venture between Lockheed Martin and the China National Huayan Technology Development Corporation, a wholly owned subsidiary of the China Meteorological Administration. The first WSR-98D was installed in China in 1999 and currently 31 such systems have been installed in China, Romania and South Korea, according to the Lockheed Martin website. For the purpose of entering the Indian market, Metstar has also forged a partnership with an Indian company SGS Weather Systems, New Delhi, which essentially is a vendor for several meteorological systems, Indian and imported.

From the perspective of ISRO and BEL, the specifications in the tender document seemed ominous right from the start. One of the qualification criteria stated in the document was that the bidder should be manufacturer who must have manufactured and tested and supplied two DWRs similar to the type specified during the last five years. Given that only one ISRO-BEL radar is operational at SHAR, and BEL is yet to supply any other, this would itself seem to rule out BEL, though the document also provided a caveat for assessing the bidders capability notwithstanding the stated qualification criteria.

Besides possible disqualification on the above count, the diameter of the antenna was specified as 8.6 m, same as that of METEOR 1500S of Gematronik, whereas that of MEGHA 2700 is 9 m. Interestingly, at a pre-bid meeting of vendors with the IMDs S.K. Banerjee, the chairman of the TEC, to a vendors specific query on whether a 9 m antenna would be acceptable, it was stated that the radar had to meet the tender specification. However, it is not clear whether the diameter was a determining factor in the final evaluation because the diameter of WSR-98D, too, is 9m. Technically speaking, what is important for radar performance is the beam spread of a one-degree pencil beam from the source at a typical distance at which weather systems are to be observed. At distances of 200-300 km, where cyclone-like formations have to be studied, this difference in antenna diameter should not make much of a difference in the measurements, argue radar experts.

If it was none of the above, what was the reason for ruling out the indigenous system? Considering that the indigenous radar was actually developed by ISRO for the IMD, this would seem to defy logic, unless, of course, the radar failed to perform satisfactorily. But that is far from the case. In fact, based on the performance of the technology demonstrator at SHAR, the IMD had ordered two more BEL radars in 2006 for installation at Kochi and Bhuj. According to BEL sources, the radars are in the final stages of integration and should be ready in a month or two. Only the IMD is yet to identify actual sites and build structures to install these.

The IMD is not happy with the products, said P.S. Goel, Secretary, MoES. The radar is not tested fully and when it is ready it will be considered during the second acquisition as part of modernisation. His remarks would seem to echo some observations made by the National Committee on Meteorological Data Products from DWR, headed by S. Raghavan, former Deputy-Director General of Meteorology, Regional Meteorological Centre (RMC), Chennai.

They have done a commendable job in the indigenous development of the radar, said Raghavan when asked about his comments on the recent developments. I had said that some of the derived products from the SHAR radar needed to be improved with better algorithms and I had recommended how to go about it. After that I do not know what has been done about it and I know nothing about the matter, he said.

That is true, said G. Viswanathan, former Director of the ISRO Radar Development Unit (ISRAD) and the principal developer of the radar. But I am not a meteorologist and I do not know the algorithm. It is the job of IMD scientists to develop it as they use the radar, based on specific local conditions. After all, it was a joint project. For example, the vertical variation in the precipitation intensity has to be statistically modelled and incorporated into an algorithm. But they have shown least interest in doing such things. The imported radars come with some proprietary algorithm developed for some other conditions and the products are assumed to be correct. What is the basis for that? he asks. Indeed, many ISRO scientists point out that the SHAR radar has not been fully utilised by the IMD to gather enough data and develop suitable algorithms for products. The situation recalls the non-utilisation of the pixel data from the CCD cameras aboard INSAT because there was some problem with the proprietary imported software that the IMD was using and it could not be corrected because of lack of source-code that the supplier did not part with while ISRO could use the same data with its in-house software (Frontline, October 7, 2005).

The Raghavan Committee had also recommended constituting a task force for calibration, inter-comparison and validation of DWR products from Chennai and Sriharikota, which are about 66 km apart, through simultaneous observations. Such a task force, headed by G.S. Bhat of the Indian Institute of Science (IISc), Bangalore, was set up by ISRO in August 2006. The committee submitted its report in November 2007. According to it, the calibration and validation of the systems showed that the performances of both the radars compared to the manufacturers specifications were on a par.

The evaluation of products could, however, be carried out only in a limited way because of the prevailing conditions. But the limited data gathered showed that there was good agreement between the products from the two radars.

However, here again, the report suggested minor improvements in the SHAR radar, like smoothening of visual displays, automation of switching between different pulse widths and use of improved slip ring (the electrical interface with the rotator assembly of the radar). According to ISRO scientists, these modifications are trivial, if required. Some of these, like automatic switching, are unnecessary, one radar expert pointed out. Once you fix a pulse width you work with it. You dont switch back and forth. Also how does it matter what slip ring I use as long as it delivers the required scan rate? he added. There is nothing in the report to recommend one over the other, he said, after studying the report.

The IMD perhaps preferred a proven and standardised system when they were investing in a large number of these, opined Bhat. They may have wanted a fully operational system. For example, ISRO used an indigenous bearing for tracking which had shorter life for 24x7x365 continuous operation. They used liquid lubricants instead of solid lubricants in some subsystems. Quality of some software could also perhaps be improved, he said. These also would seem to be minor issues that can be quickly addressed. Indeed, the two radars being readied for Bhuj and Kochi have already incorporated most of these improvements, according to ISRO scientists. Software improvement comes with use over time and there is no need for 24x7x365 operation. No weather system of interest would require that, they added.

There were such initial problems with the imported radar also. In fact, one display unit at Chennai has not been working, one scientist said. Indeed, the IMD recently floated a tender for the supply of subsystems for one of the radars. Moreover, the scientists point out, issues of maintenance, subsequent upgradation, recalibration, local condition algorithms for data products, etc. cannot be handled easily because of terms and conditions of the foreign supplier, including withholding of source code. And, most importantly, BEL systems cost is about 40 per cent of Rs.400 crore.

But an important aspect of ISRO-BEL development has been completely ignored. That is, incorporating dual-polarimetric measurement capability of the radar. Interestingly, the IMD tender specifications ask for only single horizontal polarisation measurements and do not seek this improved capability as a mandatory requirement but only hardware provision for future upgradation to dual-polarimetric capability, if desired. Ideally, from the perspective of modernisation, the IMD should have gone for dual-polarimetric radars. While the SHAR DWR, being the first ISRO-BEL system, did not have this capability, all subsequent radars, including those meant for Kochi and Bhuj, will be dual-polarimetric. It must be pointed out that neither METEOR 1500S nor WSR-98D has this capability. While minor and consequential issues are being raised to shut the doors on ISRO, this major development of consequence is not even being acknowledged, feel aggrieved ISRO scientists.

I am not bothered about the IMD, says G. Madhavan Nair, Chairman, ISRO. This remark indeed is a clear reflection of that growing disconnect between the two organisations. The radar has proved itself and we feel we should have been given a chance. But I am not unduly concerned. Besides the two meant for Bhuj and Kochi, we are supplying two more to the DRDO [Defence Research and Development Organisation] to be installed in the north and two to be operated by us in the north-east as recommended by the Cabinet Committee, he added.

The same fate might befall the automatic weather stations (AWSs) developed by ISRO, apprehend ISRO scientists. Two years ago, ISRO transferred its AWS technology, including end-to-end system integration, to the Hyderabad-based firm Astra Microwave Products Ltd., which has since been fabricating and supplying these. Today, the IMD network has 125 AWSs, of which 25 are of ISRO-Astra make and the rest are from the U.S. firm Surtron Corporation, through its Korean subsidiary. In Pune, an inter-comparison of Surtron and Astra AWSs has shown that both systems perform equally well. According to a scientist of IMD, Pune, though, just as the Surtron AWS had problems initially to operate in local conditions, Astra AWS too had some problem relating to its pressure sensor. But after using improved sensors, it has been performing exceedingly well, he said.

Under the modernisation plan, the network is proposed to be expanded to 1350. Besides, tens of AWSs are being specifically proposed for venue-specific weather forecast during the Commonwalth Games in 2010, which is another story. Towards this, 550 AWSs, at an estimated value of Rs.54 crore, are being procured in the first phase. Here again the technical evaluation of bids is over and the award is awaited.

The procurement of the other high-value item, namely the four supercomputers for the National Centre for Medium Range Weather Forecasting (NCMRWF), Noida, IMD, the Indian Institute for Tropical Meteorology (IITM), Pune, and the Indian National Centre for Ocean Information Services (INCOIS), Hyderabad, may not attract such controversy because there was no indigenous supercomputer on offer to meet the requirement, though the later development by the Tatas of a 170 teraflop HPC system, which was rated the fourth fastest in the world last November, could easily have qualified for the bid. But the question that many would ask of the MoES is what is the need for four supercomputers, two of which will be in Pune itself. Many, in fact, question the very procurement of a whole lot of systems at high cost without improving the in-house R&D capability of the IMD.