IBM Can Predict Floods And Droughts Days In Advance

I'll be interested in seeing the results under more extreme conditions. It seems, at least in this region, that conventional forecasts have a window of less than a day when it comes to predicting amounts in heavy rainfall events. Most recently, a forecast of 10-20 mm rain was upgraded to 50 mm only 6 or 8 hours before the event.

I can see that the river behavior can be predicted using the detailed maps if the inputs are known, but for 100 hour lead time this would certainly require better prediction of rainfall amounts as well...

The "Deep Thunder" approach to forecasting is very interesting, in any case. Being able to predict where floods will occur with certain inputs, or where the wind load is a hazard for a specific structure or event, is the kind of thing we are all interested to know. I bet they could sell Deep Thunder reports to prospective home buyers or land users, outlining exactly the conditions in which a disaster would be expected...whether there's a 100 hour window to predict a specific event or not.