Title: Long-term climate forecasts
Authors: Lola Gadea - University of Zaragoza (Spain) [presenting]
Jesus Gonzalo Munoz - Universidad Carlos III de Madrid (Spain)
Abstract: Climate is a long-run phenomenon, and then the interesting forecast should convey certain long-term flavour (5, 10, 25, 50, 100 years ahead). The existence of a trend locally (time series data from Central England) as well as globally (cross-section stations from NH) in all the temperature distributional characteristics (not only the mean) has been previously detected. This is a definition of local or global warming. What it was not provided is the type of trend those distributional characteristics contain. We propose different trend models to capture the evolution of the temperature distribution and, by running a forecast competition, we choose the trend model that forecasts best. We also provide temperature long-run forecast from the best model, as well as from a combination of forecasts. One of the problems of long-run forecast is the length of the confidence intervals. We propose to construct the forecast confidence intervals by forecasting the different quantiles of the temperature distribution. These intervals are smaller than the ones constructed by standard forecast methods. These forecasts provide another picture of how serious is the local-global warming.