The Claros model itself relies upon what we believe to be the best* set of weather data in existence. Our data is processed, cleaned and adjusted both automatically and manually by experienced meteorologists. Delphi then maps out the predicted movements of overarching weather-affecting patterns called teleconnections (the most famous is El Nino/Southern Oscillation or ENSO). Where Claros comes into its own is its understanding of the interplay between all of these patterns – this requires a massive amount of data processing and the input of meteorologists with a possibly unparalleled understanding of the effects that teleconnections have on our weather (we rate our team pretty highly!). It uses this understanding to deliver something akin to a hyper-complex blend of analogue-derived forecasting**, machine learning and numerical weather predictions*** that can deliver predictions and parameter-based probabilities out to 24 months without significant decay to efficacy. The results show a step-change improvement over traditional statistical modelling, AI models or the longest of the long range numerical forecasts. It represents an advance in the field of meteorology that we’re really quite proud of!

• this is totally subjective and there are many incredible sets of meteorological data. But we couldn’t do what we do with any of the others.

** a technique where historical teleconnection behaviour is compared with what’s expected in the forecast period, to identify the most similar points in the past (analogues), on the basis that these provide the most useful guidance for the forecast. Importantly, our approach is far more advanced than a simple filtering process.

*** NWP – numerical weather prediction, which uses mathematical representations of the atmosphere to model the forward evolution of weather patterns. Extremely intensive, requiring power-hungry supercomputers to produce results within hours rather than days.