Has something changed with ECMWF??

COMK Analysis

Failures are stepping stones to Success is a famous proverb which is normally used for motivation. In the context of weather forecasting understanding failures becomes critical to become successful as feedback is an integral part of numerical weather models. As most know by this time some of the most common & popular numerical models used by not only Weather Bloggers but also Weather Agencies are Global scale models. So what happens when as a weather blogger you end up way of the scale in your estimates. You go back and try to understand what is going wrong. This post is one attempt in understanding that.

With the advent of Satellite derived precipitation estimates that gives a fair outlook on past rainfall pattern it has become easy for even weather bloggers to compare what models estimated & what was the reality. To begin with today we will understand how ECMWF, without doubts the most popular weather model among the weather bloggers of Chennai and also the best performing weather model by a mile, has behaved over the last 10 days or so. During this period not only we have had thunderstorms ebb & flow but also Cyclone Amphan started developing in Bay.

It appears over the past 10 days or so ECMWF seems to have had issues estimating thunderstorms over Tamil Nadu area with consistently missing out some very active & intense thunderstorms

  • 21st May Storms near Pondicherry / Villuppuram / Cuddalore
  • 19th May North TN storms which were fairly widespread was mostly missed with the storms over Tiruvallur, Vellore & adjoining parts of South AP completely missed
  • 18th May South TN storms were completely missed with just one area around Delta showing up in the model estimates
  • In the same context it also appears ECMWF seems to have over estimates possible rains over Kerala as well. Many a times fairly widespread rainfall estimates have ended up as just isolated rainfall reality.

As weather bloggers we understand summer thunderstorms are notorious to predict and have in the past created lot of instances of failed inferences. So we decided to run a similar analysis for ECMWF over same period last year. One has to keep in mind while the overall seasonal weather was pre monsoon the day to day weather events do not follow calendar year pattern. Nevertheless it is interesting to note Last year ECMWF seems to be much closer in terms of inferences with Kerala not under Fairly wide Spread rains every day. While scale of rains seems to be still leaning towards under estimation except for couple of instances widely off the mark estimates are lesser.

It appears something seems to have changed for the world’s best performing model which is resulting in inconsistent rainfall estimates. It will be interesting to see if the arrival of Southwest Monsoon changes things around for ECMWF