Abstract
This study aims to evaluate the impact of the in-situ ocean observations on seasonal forecasts. A series of seasonal reforecasts have been conducted for the period 1993-2015, in which different sets of ocean observations were withdrawn in the production of the ocean initial conditions, while maintaining a strong constrain in sea surface temperature (SST). By comparing the different reforecast sets, it is possible to assess the impact on the forecast of ocean and atmospheric variables. Results show that the in-situ observations have profound and significant impacts on the mean state of forecast ocean and atmospheric variables, which can be classified into different categories: i) impact due to local air-sea interaction, as direct consequence of changes in the mixed layer in the ocean initial conditions, and visible in the early stages of the forecasts; ii) changes due to different ocean dynamical balances, most visible in the Equatorial Pacific in forecasts initialized in May, which amplify and evolve with forecast lead time; iii) changes to the atmospheric circulation resulting from changes in large scale SST gradients; these are non-local, mediated by the atmospheric bridge, and they are obvious from the visible impact of the removing in-situ observations on the Atlantic basin only in the global atmospheric circulation; iv) changes in the atmospheric tropical deep convection associated with the structure of the warm pools. The ocean observations have also a significant impact on the representation of the trends of the ocean initial conditions, which affect the trends in the seasonal forecasts of ocean and atmospheric variables. The impact of the ocean observing system in the Atlantic and extratropics appears dominated by Argo, but this is not the case in the Tropical Pacific, where the other ocean observing systems play a role in constraining the ocean state.