Using multi-regression-based statistics on data collected between 1979-1993 on tens of thousands of forecast points, Professor Pinhas Alpert from Tel Aviv University and his team were able to quantify the causes - human-made and natural - for weather prediction inaccuracies.
"I have been looking for a way to quantify the dominant factors that cause errors in forecasting. Until now, there has been no comprehensive analysis of these factors. They have been studied separately, but not in combination," said Alpert.
Using statistical analysis of meteorological data over thousands of locations and the course of 15 years, Alpert identified unique factors affecting forecasts in Europe, North Africa, the Mediterranean, Asia, and East Asia.
The researchers found the dominant factors clouding the accuracy of predictions comprised land-use changes (ie an area that had been covered in forest is suddenly bare), topography, particles in the atmosphere and population density.
More From This Section
"It is difficult for forecasters to incorporate changes like this. In effect, this single land-cover change altered the entire local climate over the Northern Negev, and existing forecast models had difficulty accommodating this, leading to erroneous predictions," said Alpert.
Alpert and his team also created a table of "factor prioritisation" - gold, silver, and bronze labels to identify dominant and less dominant factors for different regions in the world.
For example, they found that in the eastern Mediterranean, particles in the atmosphere were the most important cause of forecast fallacies, followed by land cover change. They also found topography to be the most influential factor affecting weather around the world.
"The only tool the weather forecaster has is his model, and the only choice he or she has is to look at different models, each of which has strengths and weaknesses," said Alpert.