Sensitivity of the tropical Pacific Ocean to precipitation-induced freshwater flux

We have performed experiments using an ocean model to study the sensitivity of tropical Paci"c Ocean to variations in precipitation induced freshwater #uxes. Variations in these #uxes arise from natural causes on all time scales. In addition, estimates of these #uxes are uncertain because of di!erences among measurement techniques. The model used is a quasiisopycnal model, covering the Paci"c from 40 3S to 40 3N. The surface forcing is constructed from observed wind stress, evaporation, precipitation, and sea surface temperature (SST) "elds. The heat #ux is produced with an iterative technique so as to maintain the model close to the observed climatology, but with only a weak damping to that climatology. Climatological estimates of evaporation are combined with various estimates of precipitation to determine the net surface freshwater #ux. Results indicate that increased freshwater input decreases salinity as expected, but increases temperatures in the upper ocean. Using the freshwater #ux estimated from the Microwave Sounding Unit leads to a warming of up to 0.6 3C in the western Paci"c over a case with zero net freshwater #ux. SST is sensitive to the discrepancies among di!erent precipitation observations, with root-mean-square di!erences in SST on the order of 0.2}0.3 3C. The change in SST is more pronounced in the eastern Paci"c, with di!erence of over 1 3C found among the various precipitation products. Interannual variation in precipitation during El Nin! o events leads to increased warming. During the winter of 1982}83, freshwater #ux accounts for about 0.4 3C (approximately 10}15% of the maximum warming) of the surface warming in the central-eastern Paci"c. Thus, the error of SST caused by the discrepancies in precipitation products is more than half of the SST anomaly produced by the interannual variability of observed precipitation. Further experiments, in which freshwater #ux anomalies are imposed in the western, central, and eastern Paci"c, show that the in#uence of net freshwater #ux is also spatially dependent. The imposition of freshwater #ux in the far western Paci"c leads to a trapping of salinity anomalies to the surface layers near the equator. An identical #ux imposed in the central Paci"c produces deeper and o!-equatorial salinity anomalies. The contrast between these two simulations is consistent with other simulations of the western Paci"c barrier layer formation.