Ajit Tyagi, former director general (2008-12), India Meteorological Department (IMD), spoke to Sandip Das on various aspects of the complex weather forecasting system. Excerpts:
How has the IMDs monsoon forecast evolved over the past decade or so
Research and development have been an integral part of long-range forecasting work at the IMD. Research work is done to 1) understand monsoon variability; 2) identify useful predictors; 3) develop more accurate statistical models and 4) assess the skills of various atmospheric and coupled models. These studies have identified new tele-connection patterns and mechanisms of monsoon variability, and new predictors for long-range forecasting. Many new statistical techniques like Ensemble Multiple Regression, Artificial Neural Network, Canonical Correlation Analysis, Discriminant Analysis and Pursuit Projection Regression have been used to develop statistical models. The IMD has started, in addition to the deterministic rainfall forecast based on the Multiple Regression Technique, probabilistic forecast based on the Ensemble Multiple Regression Technique. It also takes into consideration inputs provided by dynamical models.
How do you assess the IMDs performance in giving forecast (in terms of accuracy)
While the IMD has not been able to predict drought (in fact, no forecasting centre in the world has been able to), its performance for other years has been satisfactory within the model errors. The right interpretation of probabilistic forecast does provide some indication of likely drought.
What are the technologies used by the IMD in its forecast
As of now, the operational IMD
system is primarily based on statistical techniques. It runs experimental dynamical models and also takes into consideration the forecast of dynamical models of leading centres of the world.
Will dynamic models help the IMD in improving its forecasting system
The future lies with the dynamical model. The ministry of earth sciences, under the Monsoon Mission, is working to improve the coupled atmospheric and oceanic model for monsoon predictions. Once the forecast skills of the dynamical models become better than the statistical ones, the IMD is expected to use dynamical models for operational monsoon forecasts.
What are the key areas where monsoon forecast is used
There are many areas where monsoon forecast is used more efficiently and effectively. One such example is the weekly crop weather watch meetings organised by the ministry of agriculture. These help in continuously evaluating and updating forecasts for operational application in agriculture. Coping with the drought of 2009 and the difficult months of June and July 2012 are some success stories of such interactions. Similar close coordination and cooperation will be useful with other ministries such as water, power and state governments.
Another area where our focus should be is to provide forecast for extended range (two-three weeks) at the state level. The IMD should start providing experimental forecasts and inputs on extended range to state governments.