Macroweather temperature forecasts

点击翻译按钮获取中文摘要

Our lab has developed a new system for making monthly, seasonal and annual temperature and other forecasts: the STOChastic Seasonal and Interannual Prediction System (StocSIPS). Whereas conventional forecast systems are based on the state of the entire atmosphere and ocean at the moment of the forecast, StocSIPS instead exploits the atmosphere and ocean's long memory by using long series of past data that capture both the memory as well as (implicitly) many of the spatial dependencies (e.g. teleconnections). Since the forecast depends on the real world climate, for periods of months and longer, it is more accurate that alternatives that depend on imperfect model climates. StocSIPS is ideally suited for exploiting local - even hyper local - data. This project thus meshes perfectly with the requirements of our partner MétéoGlobale that already provides its client with various weather scale (several days or less) products and adapts them to their particular concerns. These include highly local - especially municipal scale - products that are pertinent to winter viability, in particular that help with predicting near term road conditions with implications for snow removal. Other customers benefit from dynamic marketing products that use similar information for inventory planning. At the moment, the services are restricted to very short time scales, less than one week. The specific problem that will be addressed is to provide much longer term forecasts especially for monthly and seasonal averages. MétéoGlobale is an early stage company with clients in Canada. By expanding features in its product lines, MétéoGlobale can continue with its growth not only in Canada but abroad as well. These potential product improvements can be implemented within a few years of validation. Finally having more accurate and reliably forecasts supplied to MétéoGlobale's municipal clients will directly benefit Canadians, especially in the winter months.