Tropical Indian Ocean surface salinity bias in Climate Forecasting System coupled models and the role of upper ocean processes

In the present study sea surface salinity (SSS) biases and seasonal tendency over the Tropical Indian Ocean (TIO) in the coupled models [Climate Forecasting System version 1 (CFSv1) and version 2 (CFSv2)] are examined with respect to observations. Both CFSv1 and CFSv2 overestimate SSS over the TIO throughout the year. CFSv1 displays improper SSS seasonal cycle over the Bay of Bengal (BoB), which is due to weaker model precipitation and improper river runoff especially during summer and fall. Over the southeastern Arabian Sea (AS) weak horizontal advection associated with East Indian coastal current during winter limits the formation of spring fresh water pool. On the other hand, weaker Somali jet during summer results for reduced positive salt tendency in the central and eastern AS. Strong positive precipitation bias in CFSv1 over the region off Somalia during winter, weaker vertical mixing and absence of horizontal salt advection lead to unrealistic barrier layer during winter and spring. The weaker stratification and improper spatial distribution of barrier layer thickness (BLT) in CFSv1 indicate that not only horizontal flux distribution but also vertical salt distribution displays large discrepancies. Absence of fall Wyrtki jet and winter equatorial currents in this model limit the advection of horizontal salt flux to the eastern equatorial Indian Ocean. The associated weaker stratification in eastern equatorial Indian Ocean can lead to deeper mixed layer and negative Sea Surface Temperature (SST) bias, which in turn favor positive Indian Ocean Dipole bias in CFSv1. It is important to note that improper spatial distribution of barrier layer and stratification can alter the air–sea interaction and precipitation in the models. On the other hand CFSv2 could produce the seasonal evolution and spatial distribution of SSS, BLT and stratification better than CFSv1. However CFSv2 displays positive bias in evaporation over the whole domain and negative bias in precipitation over the BoB and equatorial Indian Ocean, resulting net reduction in the fresh water availability. This net reduction in fresh water forcing and the associated weaker stratification lead to deeper (than observed) mixed layer depth and is primarily responsible for the cold SST bias in CFSv2. However overall improvement of mean salinity distribution in CFSv2 is about 30 % and the mean error has reduced by more than 1 psu over the BoB. This improvement is mainly due to better fresh water forcing and model physics. Realistic run off information, better ocean model and high resolution in CFSv2 contributed for the improvement. Further improvement can be achieved by reducing biases in the moisture flux and precipitation.