An Assessment of the Tropical Pacific Latent Heat Flux Simulated by BCC_CSM 1.1(m)

The simulated tropical Pacific annual mean latent heat flux by BCC_CSM1.1(m) as well as 14 other CMIP5 models are analyzed and compared with observations from objectively analyzed air-sea fluxes (OAFlux). Some possible causes for annual latent heat flux trend biases in BCC_CSM1.1(m) are investigated.Biases of annual average latent heat flux between observations and BCC_CSM1.1(m) in the tropical ocean and west boundary current area is larger, while in mid-high latitudes is smaller. Annual average latent heat flux is larger than observations, and annual mean latent heat flux variance is smaller than observations. The tropical Pacific annual and zonal mean latent heat flux is quite different in different latitudes. Simulation results of BCC_CSM1.1(m) near 10°N and 8°S have relatively large biases, while the biases are rather small in equator. So BCC_CSM1.1(m) needs to focus on improving the simulation of Pacific latent heat flux near 10° in each hemisphere.Among 15 CMIP5 models, NorESM1_M gives the best simulation result, and the root mean square error is the smallest, while the root mean square error of GISS_E2_R result is the largest. The root mean square error of BCC_CSM1.1(m) result is 22.9 W·m^-2, ranking eighth among all models, which indicates a moderate simulating ability.The trend of the tropical Pacific annual mean latent heat flux in BCC_CSM1.1(m) has biases comparing with the observation, and the cause can be concluded in 3 aspects. First, the local contribution horizontal wind speed to latent heat flux trend is underestimated in BCC_CSM1.1(m). Second, there are large biases for simulated non-local contribution of horizontal wind speed in BCC_CSM1.1(m). Finally, the response to the global warming of horizontal wind speed in BCC_CSM1.1(m) has large biases as well. Therefore, the main cause for trend biases of tropical Pacific annual mean latent heat flux is the large simulation deviation of horizontal wind speed in BCC_CSM1.1(m), and therefore the model needs improving in horizontal wind speed simulation.