Introducing our new and improved weather forecasting system.
It’s the end of the year and even though the farming season is coming to an end, there’s so much to look forward to in the coming year including… Drum roll please… better and more accurate weather forecasts!
Our team of meteorologists and scientific developers have been working round the clock for the last year to improve our forecasting systems and we are proud to say that they’ve done it!
As a background, the operational forecaster’s main role is to use his/her expertise, experience and pattern recognition to improve on the forecasts generated by numerical prediction systems. This role is particularly important in a tropical weather climate such as ours to enable the most thought-through and reliable service to our end-users, as traditional numerical prediction faces challenges predicting tropical weather.
To mitigate the problem of determining exactly where and when a thunderstorm will strike, Ignitia’s predictive algorithm consists of taking data from about 40 ensemble members generated by our prediction model and running them up until 72 hours into the future.
These members differ in various ways, such as how initial conditions are
generated and what physics settings are used to describe various processes. The results are then combined to form a baseline probability of rainfall for every point in West Africa. A specific algorithm then estimates the confidence of this forecast and automatically adjusts the probability if it seems off compared to a set of specific tests. Qiang, our scientific developer, runs these equations on our supercomputer and produces a best-guess forecast.
Andreas Vallgren, who is our chief science officer, thought of the methodology and after very large amounts of experiments, our meteorologist, Daniel Pinto, developed scripts in order to make a weighted and fine-tuned weighting of the members based on their performance, and comparing the outputs of these models in terms of precipitation probability against the actual rain observation from satellite observations.
From these we are able to perform several statistical measurements, make experiments and evaluate which models are best and at which times, as well as attribute the percentages given to each forecast category.
'Through this improved process, we are now able to improve the accuracy of the algorithm to further optimise the process to such a point that the second day of our 48-hour weather forecast is now almost as accurate as the first day'
Looking into the next year, we intend to expand the forecast window past Day 2, allowing Ignitia’s forecasters to focus more on Day 1 where satellite observation and experience are still of great value. We further plan to improve the quality of the dataset of the initial atmospheric conditions, so that our already good algorithm can be even further improved.
Translation: 2019 will be a great year for accurate and reliable weather forecasts from iska. We reached a huge milestone in further improving our forecasting model, making us even more advanced than global models.