Hindcast-based estimates of recent climate trends
Despite considerable progress in model development over recent decades, some persistent discrepancies exist in large-scale circulation trends between coupled models and observations. How might these discrepancies change if model biases could be reduced, and if each coupled model member matched observed patterns of multidecadal variability? To answer these questions, we explore the trends predicted by the Met Office’s coupled decadal prediction model. The model hindcasts are initialised from observed conditions and are used to predict the climate at seasonal to annual lead times. By calculating trends between successive hindcast predictions for each year in the satellite era, a distribution is built of plausible model-based trends that are constrained by observed variability. I will show that while the initialisation strongly constrains trends at low latitudes, the role of internal variability is much greater at high latitudes, even at short lead times. Key to understanding many of the hindcast trends appears to be changes in the zonal SST gradient in the tropical Pacific, and I will discuss some ongoing work to better understand the global impact of this trend.