Multi-Model Communication and Data Assimilation for Mitigating Model Error and Improving Forecasts

Models for weather and climate prediction are complex, and eachmodel typi-cally has at least a small number of phenomena that arepoorly represented, such as perhaps the Madden-Julian Oscillation(MJO for short) or El Ni~{n}o-Southern Oscillation (ENSO for short)or sea ice. Furthermore, it is often a very challenging task tomodify and improve a complex model without creating newdeficiencies. On the other hand, it is sometimes possible to designa low-dimensional model for a particular phenomenon, such as the MJOor ENSO, with significant skill, although the model may notrepresent the dynamics of the full weather-climate system. Here astrategy is proposed to mitigate these model errors by takingadvantage of each model''s strengths. The strategy involvesinter-model data assimilation, during a forecast simulation, wherebymodels can exchange information in order to obtain more faithfulrepresentations of the full weather-climate system. As an initialinvestigation, the method is examined here using a simplifiedscenario of linear models, involving a system of stochastic partialdifferential equations (SPDEs for short) as an imperfect tropicalclimate model and stochastic differential equations (SDEs for short)as a low-dimensional model for the MJO. It is shown that the MJOprediction skill of the imperfect climate model can be enhanced toequal the predictive skill of the low-dimensional model. Such anapproach could provide a route to improving global model forecastsin a minimally invasive way, with modifications to the predictionsystem but without modifying the complex global physical modelitself.