Abstract

The predictability of the duration of La Nina is assessed using the Community Earth System Model Version 1 (CESM1), a coupled climate model capable of simulating key features of the El Nino/Southern Oscillation (ENSO) phenomenon, including the multi-year duration of La Nina. Statistical analysis ... Show moreThe predictability of the duration of La Nina is assessed using the Community Earth System Model Version 1 (CESM1), a coupled climate model capable of simulating key features of the El Nino/Southern Oscillation (ENSO) phenomenon, including the multi-year duration of La Nina. Statistical analysis of a 1800 year long control simulation indicates that a strong thermocline discharge or a strong El Nino can lead to La Nina conditions that last 2 years (henceforth termed 2-year LN). This relationship suggest that 2-year LN maybe predictable 18 to 24 months in advance. Perfect model forecasts performed with CESM1 are used to further explore the link between 2-year LN and the "Discharge" and "Peak El Nino" predictors. Ensemble forecasts are initialized on January and July coinciding with ocean states characterized by peak El Nino amplitudes and peak thermocline discharge respectively. Three cases with different magnitudes of these predictors are considered resulting in a total of six ensembles. Each "Peak El Nino" and "Discharge" ensemble forecast consists of 30 or 20 members respectively, generated by adding a infinitesimally small perturbation to the atmospheric initial conditions unique to each member. The forecasts show that the predictability of 2-year LN, measured by the potential prediction utility (PPU) of the -3.4 SST index during the second year, is related to the magnitude of the initial conditions. Forecasts initialized with strong thermocline discharge or strong peak El Nino amplitude show higher PPU than those with initial conditions of weaker magnitude. Forecasts initialized from states characterized by weaker predictors are less predictable, mainly because the ensemble-mean signal is smaller, and therefore PPU is reduced due to the influence of forecast spread. The error growth of the forecasts, measured by the spread of the -3.4 SST index, is independent of the initial conditions and appears to be driven by wind variability over the southeastern tropical Pacific and the western equatorial Pacific. Analysis of observational data supports the modeling results, suggesting that the "thermocline discharge" and "Peak El Nino" predictors could also be used to diagnose the likelihood of multi-year La Nina events in nature. These results suggest that CESM1 could provide skillful long-range operational forecasts under specific initial conditions. Show less