The danger of leaving weather forecasting to AI
Man has tried anticipating climate efforts for thousands of years, using early folklore - “red skies at night” is an optimistic sight for weather-tired sailors connected to air dry and high pressure over an area - as well as roof views, hand-drawn maps, and local regulations. This guidance on future weather forecasting was based on years of observation and experience.
Then, in the 1950s, there was a group of mathematicians, meteorologists, and computer scientists - led by John von Neumann, a well - known mathematician who had helped the Manhattan Project years before, and Jule Charney, a frequent physicist of atmospheric physics Considering the father of dynamic meteorology — they tested the first automated computer prediction.
Charney, with a team of five meteorologists, divided the United States into medium-sized parcels (by today’s standards), each over 700 kilometers in area. Running a basic algorithm that gave the real-time weight range in each individual unit and advanced it over a day, the team created four 24-hour atmospheric forecasts covering the entire country. It took 33 full days and nights to fulfill the predictions. Though far from perfect, the results were bold enough to revolutionize weather forecasting, shifting the field toward computer modeling.
Over the coming decades, billions of dollars in investments and the evolution of faster, smaller computers led to an increase in predictability. Models are now able to interpret the dynamics of air parcels as small as 3 kilometers in area, and since 1960 these models have been able to input ever more accurate data from weather satellites.
In 2016 and 2018, the GOES-16 and -17 satellites were launched into orbit, delivering a number of enhancements, including higher resolution images and electronic pinpoint detection. The most popular numerical models, the US-based Global Forecasting System (GFS) and the European Center for Medium Weather Forecasts (ECMWF), have undergone a major upgrade this year, with new products and models being developed at faster than ever. At a glance, we can get an amazing weather forecast for our exact location on the Earth's surface.
Today’s electronic speed prediction, the production of advanced algorithms and global data collection, feature one step away from full automation. But they are not perfect yet. Despite the expensive models, a series of advanced satellites, and mega-computers, all human foresters have a unique set of tools. Knowledge - the ability to view and pull links where algorithms cannot - margins those predictions that continue to outperform glossy weather instruments in the most extreme conditions. height.
Although very useful with large image predictions, sensitive models are not, say, the small update in one small landmass that shows water flow creating, according to Andrew Devanas, an operational forecast at the Weather Service office National in Key West, Florida. Devanas lives near one of the most active regions in the world for waters, sea - based tornadoes that can damage ships passing through the Florida Strait # and even come ashore.
The same restriction prevents the prediction of hurricanes, freshwater and onshore tornadoes, such as those that erupt through the Midwest in early December, killing more than 60 people. But when tornadoes happen on land, forecasters often see them by looking for the signature on radar; waterfalls are much smaller and often do not have this feature. In a tropical environment like the Florida Keys, the weather does not change much from day to day, so Devanas and his colleagues had to manually monitor changes in the atmosphere, such as increased wind speeds and humidity. available - changes not made by the algorithms. always pay attention - to see if there was any relationship between certain factors and higher risk from waters. They compared these observations with a model probability index showing the probability of rainfall probability and found that, with the correct combination of atmospheric measurements, the human prediction was “more better ”is a model in all metrics of water flow prediction.
Similarly, research published by NOAA Weather Prediction Service director David Novak and his colleagues shows that while human forecasters may not be able to “beat” the models on the day your normal sun, they still make more accurate predictions than the algorithm- crunchers in bad weather. Over the two decades of data researched by Novak's team, people were 20 to 40 percent more accurate at predicting near-future precipitation than the Global Forecast System (GFS) and the American Mesoscale Weather System North (NAM), the most common national models. People also made statistically significant improvements in temperature prediction across the guidelines of both models. “Oftentimes, we find that it is in the big events when the forecasters can make some further improvements to the automated control,” says Novak.