India has been dealing with the changing climatic conditions and needs reliable forecasts for extreme weather conditions like droughts, floods, cyclones, lightning and air quality, among others. On November 28, India announced that over the next 2 years it plans to augment the existing Supercomputer to 100PetaFlops for accurate weather forecasting. This announcement made at the 4-day workshop on Prediction skill of extreme Precipitation events and tropical cyclones: Present Status and future prospect (IP4) and Annual Climate Change.
Why use Supercomputers?
With India’s 70% of livelihood still depending on agriculture, increasing the accuracy of weather prediction becomes essential. To understand how supercomputers help in weather prediction, we have to understand a little about how weather forecasting works.
Weather forecasting uses what we call weather forecast models. These models are the closest thing to a time machine for meteorologists. The weather forecasting models are a computer programme which simulates what atmospheric conditions could look like in the foreseeable future. These models solve group mathematical equations that govern the climatic conditions, and these equations approximate the atmospheric changes before they take place. Now, there are two types — Statistical and dynamic models. The statistical models haven’t been providing a reliable result, so a dynamic model has been developed for Indian conditions.
However, to run such dynamic models and provide such forecasting, enhanced supercomputing power is necessary.
Every hour weather satellites, weather balloons, ocean buoys, and surface weather stations around the world record billions of data. This large volume of data is stored, processed and analysed — and that’s where supercomputers come in. The meteorologists could solve the governing equations by themselves, but the equations are so complex that it would take months to solve them by hand. In contrast, supercomputers can solve them in a matter of hours. This process of using models equations to forecast the weather conditions numerically is called Numerical Weather Prediction.
Some of the world’s most famous weather forecast and climate monitoring models include:
- Global Forecast System (GFS)
- North American Model (NAM)
- European Center for Medium-Range Weather Forecasting Model (European or ECMWF)
When the supercomputer gives the output, the forecaster takes this information in consideration along with their knowledge of weather process, personal experience and the familiarity with nature’s unpredictability to issue the forecast.
Meet The Supercomputer
Now, what is this 100 PF Supercomputer that India is planning to use?
First, 100PF stands for 100 petaflops like there are measuring units for speed for a device, the supercomputer has FLOPS.
A fun way to understand what a PF is like — if 1 PetaFlop took 1 second to prove a complex mathematical theorem, then you would take 31,688,765 years to solve it.
These supercomputers are usually hard to manufacture and have an operating span of 5 years because of the enormous temperature conditions they operate in. These supercomputers need a large facility to be kept in because of their size. US’ supercomputer ‘Summit’ which holds the top position when it comes to supercomputer rankings with a capacity of about 148.6PF, spreads over 5,600 sqft, which is about the size of two tennis courts.
Regarding India, in 2018 the French company, Atos won the 4500 crore tender for manufacturing these supercomputers, other competitors being Lenovo, HP, NetWeb Technologies. Atos has a 3-year contract with India to manufacture 70 HPCs under National Supercomputing Mission.
India’s Supercomputers over the years
Over the last ten years, India has successfully upgraded its supercomputing capacity facility. Below is a list of some of the acquisitions of High-Performance Computing (HPC) systems over the years:
- Prithvi- Indian Institute of Tropical Meteorology (IITM) acquired Prithvi in 2009. It was IITM’s first HPC with 7.2 teraflops (TF). It was increased to 70TF in 2010
(A TF speed unit is like, what 1 TF computer system can do in just one second, you’d have to perform one calculation every second for 31,688.77 years.)
- Aaditya- In 2014, IITM acquired Aaditya which had a computing speed of 790TF.
- Pratyush- Pratuyush has a computing capacity of 4.0 PF and is in IITM Pune.
- Mihir- This HPC is in the National Centre for Medium-Range Weather Forecasting (NCMRWF), Noida with a computing capacity of 2.8 PF.
Both Pratyush and Mihir were inaugurated in 2018. With Pratyush and Mihir, India is in the top-30 position (from 368th position) in the top 500 list of HPC facilities in the world. The facility will also place India at the 4th position after Japan, UK and USA, concerning the HPC resources for weather and climate community.
Some notable predictions and uses by supercomputing power in India:
- Predicting monsoon on different temporal and spatial scales, notably the seasonal and extended range predictions.
- Predicting cyclones like ‘Phailin’, ‘Hudhud’, and the ‘Ockhi’ with adequate lead time.
- Predicting extreme rainfall events like the Uttarakhand floods, Kashmir floods, etc.
- Providing tsunami warnings.
Dr M Rajeevan, the Union Secretary for Earth Sciences, this week said that in 2019 the Govt has predicted five cyclones accurately. India is using supercomputers with a combined capacity of 6.8 PF. One can imagine how powerful the services of a 100PF Supercomputer can be because even with accurate predictions of the weather this year, some improvement is needed. Therefore to provide precise prediction at a high resolution, more supercomputing power is required. It will not only benefit India but also the neighbouring countries.
Also, here is a short video by NASA Goddard