Why is it so difficult to forecast rain?

Posted on June 1, 2017  /  0 Comments

As Sri Lanka is drying itself out after yet another disaster, people are beginning to ask what went wrong and what could be done better in the future. Some of the comments are not fair, for example the comparison of the Bangladesh and Sri Lanka responses, but most are useful. Every disaster must be treated as a learning opportunity.

First, let’s get the Bangladesh comparison out of the way. Once a cyclone forms, its track can be seen from satellites. Even if Bangladesh did not have access to satellite images, the very fact that 200 people died in Sri Lanka would be warning enough.

Sri Lanka was not harmed by a cyclone. Cyclone Mora formed after the storm passed Sri Lanka. There is no reliable way to estimate the quantity of rainfall that will be dumped from a storm that approaches Sri Lanka from the sea.

He told IRIN that, on 25 May, the DMC received forecasts for the normal amount of monsoon rain, about 150 milimetres for the following 24 hours. Instead, 550 milimeters of rain fell in some areas between 9 pm on 25 May and 5 am on 26 May.

“The low pressure system just changed so suddenly, there was no time for anyone to communicate, issue warnings or effect evacuations,” he said. “It was so sudden and quick.”

As the disaster unfolded, the DMC began sending out mass text messages to warn of floods in different areas. The Irrigation Department and the National Building Research Organisation also issued alerts. But no alerts were issued, and no evacuations were carried out before the storm arrived.

The Irin Report goes on to talk about the inadequacy of sensors.

Here, there is confusion:

Kodippilli said the DMC relies on information from by the Irrigation Department on floods and the National Building Research Organisation on landslides, but his agency received no information from either body.

Assume the Irrigation Department has all the sensors it needs in all the reservoirs and waterways under its control. Can this help PREDICT floods? It can tell us that there are elevated water levels in particular places. It can inform us about the likelihood of floods downstream from those locations, precisely if good models are used and roughly, even if not. But on floods caused by excessive rainfall (550 mm in 8 hours), Irrigation Department sensors are useless. NBRO has installed rain gauges that automatically communicate rainfall data to their head office. This does not predict floods or rain, but when fed into the right models can tell us about the likelihood of landslides in specific locations.

To predict (that is to know the quantum of rainfall and where it falls, hours before the rain falls), we need sensors located where the storm is hours earlier, which is the middle of the Indian Ocean. After the storm makes landfall and the sensors are in place to pick up the data, we will not have hours but minutes, to warn the coastal areas. Weather prediction is a complex business, so challenging that the first head of the UK Meteorological Office, Robert FitzRoy took his own life due to the strain. Peter Moore’s Weather Experiment provides a good description of the challenges of weather prediction in islands where the weather comes from the sea.

There is no doubt that we can do more to improve our forecasts and prepare to cope with the extreme weather events that are the results of climate change. But that must be based on a realistic assessment of the existing science and technology, not on raw emotion.

Comments are closed.