Even if they are not always accurate, current weather forecasts are, most often, reliable ten days ahead. They are provided by computer models that integrate the various weather components and can thus predict the change in weather But thanks to a new technique, this limit could be pushed back. The limit of reliable predictions could be pushed back from ten to fourteen or fifteen days, an increase of nearly 30%. Reduce the uncertainty of current conditions To do this, scientists at the University of Pennsylvania incorporated fewer variables associated with weather forecasting. "Reducing the uncertainty of the current exact conditions can increase reliable forecasts to five additional days," says the team that conducted the study. Thus, each weather forecast is primarily based on current weather conditions. But the more you go in the future, the greater the uncertainty because the number of variables involved increases. These initial conditions are collected by many different measuring instruments: balloons, satellites, weather stations. And the data is far from perfect. As a result, forecasting models suffer from these imperfections. Almost perfect data The researchers behind this new study decided to pretend that these data were almost perfect, using ten times fewer variations to begin their predictions. Thus, the more precise the data at the beginning, the less there will be variables to integrate then. They averaged the data collected, ignoring those that might appear to be outliers. These, which were previously taken into account, could complicate the task of computer simulators. By testing this approach with data from previous 2016 weather reports, the team was able to predict, with fairly reliable accuracy, what the weather would be two weeks later. With very different models from those of three years ago. A technique yet to be perfected Nevertheless, there is still a lot of work to be done to complete this new approach. Thus, some scientists argue that this technique is far too speculative to actually be used. The 14-day barrier seems to be a glass ceiling on which meteorologists stumble. Already, half a century ago, research had proven that predictions became useless beyond two weeks, even before the arrival of supercomputers in the twenty-first century. And if the technology continues to improve, it seems that, for now, two weeks is a realistic limit.Check out the video above for more!