44% of heavy rainfall alerts for Mumbai were wrong in 2017, admits IMD
The information was revealed by the weather bureau in its reply to a reply to right to information querymumbai Updated: Mar 26, 2018 12:14 IST
The India Meteorological Department (IMD), Mumbai, issued 32 rainfall predictions in 2017, of which 14 turned out to be incorrect. This means 44% of the predictions of heavy, very heavy or extremely heavy rainfall warnings issued by IMD were incorrect. The information was revealed by the weather bureau in its reply to a reply to right to information (RTI) query. The southwest monsoon is active over Mumbai between June and September.
Meteorologists and independent advisor to the state said international weather agencies at major cities such as London, Miami, New York, and Dubai ensure they maintain 80% to 85% accuracy throughout the year.
Based on an RTI filed by activist Chetan Kothari in December, the IMD (Mumbai) responded to his query on Friday. “In a city like Mumbai, which is no stranger to extreme weather events, accurate predictions are very important for lakhs of commuters travelling daily. However, over the past two years, the accuracy for such predictions has been really poor. It is the same reason why lakhs of farmers are suffering in rural Maharashtra,” said Kothari.
IMD officials said variability of weather factors plays a major role during predictions, especially in tropical areas. “Every monsoon is different from the next one. It cannot be generalised as there is a difference of amount of rainfall, intensity variability of weather events and number of heavy rainfall days, among others. However, there is always scope for improvement,” said KS Hosalikar, deputy director general, western region, IMD. “Since it is forecast, there is a tendency that it may get deviated due to changing weather factors. This is just data from one year. Some years might not have challenges and forecasts have been 80% accurate.”
The verification of weather forecasts are further based on three separate categories – correct, moderate and unusable, he said. “We issue details from these three categories post our forecast for the state, citizens, commuters, and farmers. to use appropriately. As opposed to aviation forecasts which are very specific to flying zones and are more accurate, general forecasts are for larger areas or sub-divisions where the chances of variability are much more,” said Hosalikar.
For weather predictions, IMD currently uses surface observations over land and sea that are received from different observatories, including that of the state government, automatic weather stations, upper air observational network where balloons are sent for data collection, satellite observations, and numerical weather predictions based weather models and surface charts. Forecasts vary from 0 to 2 hours – nowcasting, short range (12 to 72 hours), medium range forecasts (5 to 7 days), extended outlook (15-30 days), monthly forecast and long range forecast (for summer and monsoon).
Akshay Deoras, meteorologist and independent meteorological advisor to the Maharashtra government, said false alarms mainly happen in India because forecasters blindly rely on numbers instead of estimating impacts. “IMD Mumbai’s probability of correctly issuing a rainfall warning last year has been as good as getting a head or tail on tossing a coin. This kind of forecasting is dangerous as it leads to confusion among citizens due to randomness in false alarms. In many cases, thunderstorms act as game changers as models cannot accurately tell if they will produce 50 or 100mm rainfall. However, a smart forecaster should be able to make sense by checking other parameters,” he said.
Deoras added that IMD predictions on July 17, August 29 and September 19 advisories were so ambiguous that they gave all possibilities at the same time. “IMD’s verification is based on 24 hours rainfall accumulation and not the intensity. International agencies like the UK Met office, however, focuses more on issuing impact specific weather advisories, which are backed by regular research and assessment of changing climatic and environmental scenarios. They have an accuracy rate of 80-85%, sometimes even up to 90%,” he said.