Real time prediction of heat waves two to three weeks in advance is now possible, thanks to the extended range prediction system developed by researchers at Pune’s Indian Institute of Tropical Meteorology (IITM). The prediction system has shown promise in indicating the date of onset, duration and demise of heat waves, with a small spatial and temporal error.
Transferred to IMD
It is only now that heat waves are being predicted in real time two-three weeks in advance. The system can predict heat waves with 70% accuracy, which is quite good considering that no real time prediction existed earlier. The prediction system has been successfully transferred to the Indian Meteorological Department (IMD), which has been using it since 2017. IMD used this system to predict heat waves this year.
Based on observation data available from 1981 to 2017 and model-run data available from 2003, the researchers found that places in northwest India and southeast coastal regions are prone to heat waves conditions. Places in these two regions experience heat waves for more than eight days during summer. The study found that the southeast coast region has become more vulnerable to heat waves in the recent years.
In India, heat waves are generally seen in Rajasthan, Punjab, Haryana, Chandigarh, Delhi, west Madhya Pradesh, west and east Uttar Pradesh, Chhattisgarh, Orissa and Vidarbha, parts of Gangetic West Bengal, coastal Andhra Pradesh and Telangana. These regions experience heat waves for more than six days in a year.
‘From March 1’
“The extended range forecast can predict heat wave condition two weeks in advance with 70% probability. The short-range forecast two-three days before the onset of heat wave condition can provide more reliable information about the time and location of the heat wave condition,” says Dr. Susmitha Joseph from IITM and co-author of a paper published in the journal Scientific Reports.
“The heat wave forecast starts from March 1 till monsoon covers the whole country,” she says.
A heat wave day is identified when the maximum temperature is above 44 degree C or when all the three following conditions are seen — when the region is facing unusually hot temperature, when the temperature is more than 36 degree C and when the departure from the normal temperature is more than 3.5 degree C.
“While the IMD considers 45 degree C maximum temperature to classify it as heat wave, we used 44 degree C because we were using gridded data,” said Dr. A.K. Sahai from IITM Pune, corresponding author of the paper. “We also modified the criteria developed by the IMD to classify heat wave to fit for gridded data.”
Events identified
Based on the new criteria, between 1981 and 2018, the researchers identified 22 heat wave events in northwest region, and 14 in southeast coastal region. In addition, the researchers found nine events in a new region called northwest-southeast — when heat wave spells in northeast and southeast regions have at least one day overlap.
The researchers verified the model for heat wave using three select events in all the three heat wave regions. The model verified the onset, duration and end phase of heat wave events by the first three and last three consecutive days, and the total spell of the event.
Not precise location
For the northwest region, the heat wave event of June 2-11, 2014 that caused more than 1,500 deaths was used. The model was able to capture the onset even a week in advance but not the intensity. It was able to correctly predict the duration and end of the heat wave event but not the precise location about three weeks in advance. The predicted location of the event was off to the observed location.
For the southeast region the heat wave even of May 15-22, 2008 was taken for testing. The model could predict the onset of the heat wave even about two weeks in advance, and the duration and end of the event about a week in advance. Again, the model could predict the location of the event one week in advance.
For the northwest-southeast region the May 18-31, 2015 event was chosen. The model could predict the onset about 10 days in advance. The duration and end phase of the event was also captured reasonably well.
