Ensemble Kalman filter-based targeted observation is one of the best methods for determining the optimal observational array for oceanic buoy deployment. This study proposes a new algorithm suitable for a ‘cross-region and cross-variable’ approach by introducing a projection operator into the optimization process. A targeted observational analysis was conducted for El Niño–Southern Oscillation (ENSO) events in the tropical western Pacific for the Tropical Pacific Observation System (TPOS) 2020. The prediction target was at the Niño 3.4 region and the first 10 optimal observational sites detected reduced initial uncertainties by 70%, with the best observational array located where the Rossby wave signal dominates. At the vertical level, the most significant contribution was derived from observations near the thermocline. This study provides insights into understanding ENSO-related variability and offers a practical approach to designing an optimal mooring array. It serves as a scientific guidance for designing a TPOS observation network.

In this study, we investigate the changes in the multiyear predictability of the Pacific decadal oscillation (PDO) since 1900. A newly developed method, the model-analog method, is used to hindcast the PDO from 1900 to 2015. Model-analog hindcast provides comparable PDO prediction skills with the traditional assimilation-initialized forecast but with much lower computational costs. Our results show that PDO prediction skills have significantly changed over time. Specifically, the PDO could skillfully be predicted up to 5 years in advance in 1910–1960, but it can only be predicted 2–3 years in advance after 1960. We attribute these changes to the strength of the re-emergence process, where deep oceanic mixed layer temperature anomalies from one winter reappear in the following winters. In the high-prediction-skill period, the stronger re-emergence process provides more predictable information, leading to enhanced multiyear prediction skills. Our study offers new insights into the multiyear predictability of the PDO over a long-time frame and sheds light on the reasons behind the low prediction skill of the PDO.

In this paper, we employ reanalysis data to systematically investigate the development of the Indian Ocean dipole (IOD), thereby distinguishing the SST–wind causal relationship during IOD development. The results indicate that the variations in sea surface temperature anomalies (SSTA) are particularly important during IOD development. SSTAs over the eastern Indian Ocean (EIO) lead to variations in Sumatran coastal winds and equatorial zonal winds, whereas SSTAs over the western Indian Ocean (WIO) lag behind these variations. On this basis, the Community Earth System Model (CESM) is adopted to examine the influences of different atmospheric physical processes and model resolutions on the simulation of the IOD evolution. For this purpose, four sets of sensitivity experiments are carried out involving two versions of the Community Atmospheric Model (CAM4 or CAM5) and two atmospheric model resolutions (0.9° × 1.25° or 1.9° × 2.5°). The CAM5 simulation experiments better capture the detailed characteristics of IOD development, especially the wind–SST causal relationship, than the CAM4 experiments. Moreover, increasing the resolution of the atmospheric model can effectively reduce the simulation bias, thus benefiting the simulation of the SST–wind relationship.

In this study, we employ the Model‐based Analog Forecast approach to conduct Indian Ocean Dipole (IOD) hindcasts from 1982 to 2017, using 18 CMIP6 models. We focus on the skill diversity among different climate models, with particular attention to how the cold tongue bias affects IOD predictions. Our findings reveal a significant diversity in IOD prediction skills across the CMIP6 models. Skillful predictions are observed for lead times ranging from 1 to 4 months, depending on the model. Also, this study identifies a direct relationship between cold tongue bias and IOD prediction skills. Models that exhibit a more pronounced cold tongue bias tend to show weaker El Niño‐Southern Oscillation influences over the tropical Indian Ocean, which in turn leads to a reduction in IOD prediction skills. This study provides valuable insights into the factors driving the diversity in IOD predictions and highlights the potential for improving IOD forecasting skills.

The model‐dependency has been a challenging issue for traditional data assimilation‐based targeted observational method. This study developed a new strategy to address this challenge using multiple‐model prediction ensemble. It was found that while the ensemble size reaches a sufficiently large number the optimal observational sites detected tend to stable and model‐independent. This new finding answers the long‐standing challenge question on the model dependence in targeted observational analysis, offering an efficient and objective way to identify optimal observational sites. With this strategy, we designed an optimal observational array in the tropical Pacific for the El Niño‐Southern Oscillation (ENSO) prediction using the multiple historical simulation data sets from Coupled Model Intercomparison Project Phase 6 and reanalysis data sets. Sensitive experiments show that while number of data sets reaches 12, a robust optimal observational array is obtained. The first 10 optimal observational sites, mostly located in the equatorial central eastern Pacific, can reduce initial uncertainties by 67%. This was further confirmed by the observation system simulation experiments, which is implemented by the Ensemble Adjustment Kalman Filter assimilation system developed in the Community Earth System Model. This newly developed model‐independent strategy makes it feasible to design a robust oceanic observational network for ENSO prediction even using the current targeted observational algorithm, well serving the goal of international Tropical Pacific Observation System 2020 project.

The boreal summer Indian Ocean Basin Mode (IOBM) profoundly impacts the summer climate over South and East Asia, yet its significant decadal variability poses a major challenge for seasonal predictions. Traditional dynamical prediction models are typically initiated from 1980 onwards due to computational constraints, limiting the comprehensive study of decadal changes in IOBM predictability. To address these limitations, this study investigates decadal changes in boreal summer IOBM predictability (1948–2022) using the Model-based Analog Forecast (MAF) method, based on a library of 29 CMIP6 models. The hindcast results reveal a distinct regime shift in predictive skill. Change-point analysis identifies transitions in prediction skill at 1978 and 2000, based on which three equal-length 21-year periods (1958–1978, 1980–2000, and 2002–2022) are selected to represent the early low-skill, high-skill, and recent low-skill regimes, respectively. This decadal shift is primarily attributable to the decadal modulation of the El Niño-Southern Oscillation (ENSO)–IOBM teleconnection. During the high-skill period, El Niño events induce significant southwestern Indian Ocean (SWIO) warming. This, in turn, activates a robust wind-evaporation-SST (WES) feedback, which maintains the basin-wide warming into summer, thereby providing an enhanced signal component for IOBM predictions. In contrast, during the low-skill period, weaker ENSO forcing fails to sustain this feedback, leading to weakened IOBM signals and consequently lower forecast skill. These findings demonstrate that boreal summer IOBM predictability is nonstationary and reveal that accurately representing the ENSO–IOBM teleconnection is essential for advancing forecast skill.