The proximate cause of the disaster was heavy rains for several days. This started three weeks earlier than the monsoons' normal arrival date. Maybe similarly heavy rains during the monsoons would have caused similar damage. But the loss of life would have been much less simply because rain would have been expected. Given adequate warning, locals would have evacuated danger zones. The tourists and pilgrims would also have gone by then.
The Indian Meteorological Department (IMD) issued warnings about heavy rains and snowfall in the upper reaches around 48 hours before the bad weather. The state chief minister claims the warnings were imprecise. Without getting into blame games, it is also possible that the warnings weren't taken seriously and it is clear that the state's disaster management authority wasn't up to the task.
More From This Section
The terrain caused many problems. In hilly areas, rain or snow at heights causes damage to downhill areas via flash floods. The hardest-hit are often not the places where the heaviest rain occurs. This did happen. Rain at heights fed rivers, which ran downhill in spate, sweeping entire villages away downstream.
The IMD cannot predict flash floods. As it happened, the rainfall was more than heavy; this has been the wettest June in many years for the hill state. Climate scientists believe that unseasonal rain, and unusual or changing weather patterns, will become more common as global warming increases.
This makes the task of weather prediction even harder. Weather forecasting models are complicated and require a lot of number-crunching as well as good data. The IMD has started inducting technology in the form of supercomputers. But, although accuracy and granularity has improved, even the best weather models have big errors.
Weather prediction began with the statistical regression of historical data. This meant assuming the same patterns would hold in the future. The IMD started using statistical models in the 1920s. By the 2000s, it was using a 16-variable statistical model, which divides India up into four major regions.
Unfortunately, this model has large errors. This may be because the climate scientists are right and weather patterns are indeed changing. The assumed error range for monsoon predictions is supposed to be plus or minus five per cent. In the 17 years between 1994 and 2010, monsoon precipitation deviated from the predictions by over 10 per cent as many as 13 times.
What is more, the statistical model is insufficiently granular. It doesn't offer timely, actionable district-level information farmers can use to plant crops. Nor does it give administrations a chance to brace for possible floods or drought.
Beating the statistical approach involves modelling the physics of weather systems. The data has improved a lot, thanks to Indian Space Research Organisation's satellite-based inputs. But the physical models require supercomputers to solve and even so, there are either/or outcomes.
Multiple variables like air pressure, humidity, temperature, wind-speed, cloud cover, and so on, interact in dynamic non-linear fashion to create weather systems. The weather is caused by feedback loops from their interactions. A small change in the initial value of one variable can lead to different weather patterns. Thousands of differential equations must be programmed and solved to simulate weather, inputing a broad range of assumed values for all variables.
Weather models divide areas up into grids - the smaller each grid, the better the granularity. In India, a 38 km x 38 km grid is generally applied. Then, the equations are solved for every grid. Actually "area" is inaccurate, since the grid is three-dimensional, incorporating the atmosphere. Data samples must be updated as often as possible and the equations run again with new data to modify predictions.
The results are probabilistic. The computer model may predict, for instance, a 79 per cent chance of rain, given that temperature stays between 33-35 degree Celsius and humidity stays above 80 per cent. The weather department then translates those possible outcomes into what it thinks most likely, applying human knowledge and judgement.
The IMD has been adapting the climate forecast system of the National Centre for Environmental Prediction, USA, to fit Indian conditions. According to studies by the US National Weather Service, local knowledge and human understanding can increase the accuracy of computer models by at least 25 per cent. So, institutions like the Indian Institute of Tropical Meteorology, Pune and National Centre for Medium Range Weather Forecasting in Noida and the Council of Scientific and Industrial Research's Centre for Mathematical Modelling and Computer Simulation will have to play a major role in training computer scientists and getting them to "train" computers.
The IMD wants to develop dynamic seasonal and extended range predictions (16 days to one season) as well as short to medium range predictions (one to 15 days). The models should also be able to predict extreme events like cyclones. Paradoxically, cyclone trajectory and intensity are somewhat easier to predict than normal weather because cyclones they have defined, powerful patterns.
Repeated floods and droughts across India make the need for accurate weather predictions more obvious. And, as US Secretary of State John Kerry recently said, referencing Uttarakhand, "The science of climate change is screaming at us for action." If the tragedy of Uttarakhand leads to greater political acceptance of the need for environmental sustainability and a climate that is more conducive to weather research, at least some good will have emerged from the calamity.