A Study On ENSO Simulations/Predictions By Parameterizing Subsurface Entrainment Temperature

There are some common bias in the current OGCMs and OGCM-based coupledmodels (HCMs and CGCMs), e.g., the underestimated SST variability in the easternequatorial Pacific and along the coast of South America, and too weak and too highfrequent ENSO variability in coupled models, and so on. In view of these problemsin the current OGCMs, a scheme is put forward to improve their ENSO simulations,which is focused on the subsurface entrainment temperature (T'_e). In particular,following the intermediate ocean models (IOMs), a separate SSTA submodel, inwhich T'_e is parameterized by an empirical nordocal scheme, is embedded intoOGCMs. By this way, ENSO simulations are improved notably in a tropical PacificOGCM (IAP-TPOGCM) and a coupled model based on it. What' more important,ENSO predictions are also improved by this way. In addition, we also examined therole of T'_e in ENSO on deeadal scale. In detail, the main results are given asfollows:1. Following IOMs, a T'_e parameterization scheme is applied in an OGCM. Inthis parameterization scheme, an empirical relationship is developed between T'_eand sea level variability, so that T'_e can be parameterized by SL anomlies explicitly.By this scheme, the vertical process can be described in OGCMs better, and weakenthe common bias in the current OGCMs, i.e., too week feedback of the thermoclinevariability on surface mixed layer.2. Improved ENSO simulation in an OGCM and OGCM-based coupled modelby parameterizing the subsurface entrainment temperature (T'_e). Similar to manycurrent OGCMs and OGCM-based coupled models, IAP-TPOGCM shows manybiases, e.g., too low SST variability in the far eastern Pacific and too high frequentENSO oscillation in the coupled model (quasi—biennial oscillation). In contrast,when the subsurface entrainment temperature (T'_e) parameterization scheme isapplied, SSTA simulations are improved notably on the uncoupled basis.Furthermore, ENSO simulations in coupled model are also improved notably, e.g., realistic frequency, realistic SST anomalies pattern and so on. Further analysis showsthat the simulated ENSO in the improved coupled model is consistent with the fourfashionable ENSO theories from different aspects, which suggests that manymechanisms can operate simultaneously in ENSO cycle.3. Checked the possibility of improving ENSO predictions by parameterizingthe subsurface entrainment temperature (T'_e), and developed a new tool for ENSOforecast. Long-term hindcast experiments are explored by the standard and improvedcoupled models, in which ocean components are all initialized by an ocean dataassimilation system (OVALS). It is shown that the skillful period in standard coupledmodel is only 5-6 months equivalent with persistence forecast; and consistent withthe bias in ENSO simulations, the skillful area is confined in a narrow band alongthe equator, and the skill is depressed sharply along the far eastern Pacific oceanwith the leading times extended. In contrast, the prediction skill is improved notablydue to the empirical T'_e parameterization, anomaly correlation in which is enhancedby 0.1-0.2 compared to the standard one, and the skillful period has been prolongedto 9 months; furthermore, the large prediction errors in the far eastern Pacific oceanare also put down.4. Examined the role of T'_e in ENSO Decadal Changes and its possiblemechanism. Several sensitive experiments are explored by a HCM. It is shown thatby modifying the relative strength of the zonal advective feedback and thethermocline feedback in the coupled system, the changes in the structure of T'_e canmodulate the ENSO properties (i.e., the oscillation period and propagationcharacteristic of SSTAs along the equator), which are consistent with the behaviorshift of ENSO observed in the late 1970s. Further experiments show that the T'_echanges have dominant role in the ENSO decadal changes as observed in the late1970s, while the effect of oceanic background changes is not clear in our models.The main innovative points are summarized as follows:1. Following IOMs, a T'_e parameterization scheme is applied in an OGCM.By this scheme, the vertical process can be described in OGCMs better, and ENSO simulations are improved on uncoupled- and coupled- basis. This scheme issuggestive for other OGCMs.2. By conducting long-term ENSO hindcast experiments, we examined theidea that ENSO predictions may be improved by the T'_e parameterization scheme.Furthermore, a new tool for ENSO forecast has been developed. This result suggeststhat this scheme may be a choice for improving ENSO predictions and it may besuggestive for other operational forecast systems.3. Examined the role of T'_e in ENSO Decadal Changes and its possiblemechanism and put forward a new explanation for it.