Proxy system models (PSMs) are an important bridge between climate simulations and climate records prior to the period where instrumental observations are available. PSMs help to interpret what proxies show and how they record climate. Although previous studies have evaluated PSMs for individual sites, their systematic evaluation on a global scale has not yet been conducted. This study evaluated the performance of PSMs for stable water isotopes in ice cores, corals, and tree‐ring cellulose for the period 1950–2007. Spatial distributions of the mean state were well simulated for all proxy types, albeit with a bias for tree‐ring cellulose. Interannual variability was well simulated for corals and tree‐ring cellulose. These results indicate that the models represent key mechanisms for the proxies. In contrast, the reproducibility of interannual variability in ice cores was markedly lower than that for the other proxies. Although the reproducibility was limited by the atmospheric forcing used to drive the model, the results suggest that the PSM may be missing postdepositional processes, such as sublimation for ice cores on the interannual time scale.

Water isotopes measured in Antarctic ice cores enable reconstruction at the first order of the past temperature variations. However, the seasonality of the precipitation and episodic events, including synoptic‐scale disturbances, influence the isotopic signals recorded in ice cores. In this study, we adopted an isotope‐enabled atmospheric general circulation model from 1981 to 2010 to investigate variations in climatic factors in δ¹⁸O of precipitation (δ¹⁸O_p) at Dome Fuji, East Antarctica. The Southern Annular Mode (SAM), the primary mode of atmospheric circulation in the southern midhigh latitudes, significantly contributes to the isotope signals. Positive δ¹⁸O_p anomalies, especially in the austral winter, are linked to the negative polarity of the SAM, which weakens westerly winds and increases the southward inflow of water vapor flux. Daily variations in temperature and δ¹⁸O_p in Dome Fuji are significantly small in the austral summer, and their contribution to the annual signals is limited. The isotope signals driven by the SAM are a locational feature of Dome Fuji, related to the asymmetric component of the large‐scale atmospheric pattern.

The products from the Stable Water Isotope Intercomparison Group, Phase 2, are currently used for numerous studies, allowing water isotope model‐data comparisons with various isotope‐enabled atmospheric general circulation model (AGCMs) outputs. However, the simulations under this framework were performed using different parameterizations and forcings. Therefore, a uniform experimental design with state‐of‐the‐art AGCMs is required to interpret isotope observations rigorously. Here, we evaluate the outputs from three isotope‐enabled numerical models nudged by three different reanalysis products and investigate the ability of the isotope‐enabled AGCMs to reproduce the spatial and temporal patterns of water isotopic composition observed at the surface and in the atmospheric airborne water. Through correlation analyses at various spatial and temporal scales, we found that the model's performance depends on the model or reanalysis we use, the observations we compare, and the vertical levels we select. Moreover, we employed the stable isotope mass balance method to conduct decomposition analyses on the ratio of isotopic changes in the atmosphere. Our goal was to elucidate the spread in simulated atmospheric column δ¹⁸O, which is influenced by factors such as evaporation, precipitation, and horizontal moisture flux. Satisfying the law of conservation of water isotopes, this budget method is expected to explain various fractionation phenomena in atmospheric meteorological and climatic events. It also aims to highlight the spreads in modeled isotope results among different experiments using multiple models and reanalyses, which are primarily dominated by uncertainties in moisture flux and precipitation, respectively.

Data assimilation (DA) has been successfully applied in the field of paleoclimatology to reconstruct past climate. However, data reconstructed from proxies have been assimilated, as opposed to the actual proxy values. This prevented full utilization of the information recorded in the proxies. This study examined the feasibility of proxy DA for paleoclimate reconstruction. Isotopic proxies (δ¹⁸O in ice cores, corals, and tree-ring cellulose) were assimilated into models: an isotope-enabled general circulation model (GCM) and forward proxy models, using offline data assimilation. First, we examined the feasibility using an observation system simulation experiment (OSSE). The analysis showed a significant improvement compared with the first guess in the reproducibility of isotope ratios in the proxies, as well as the temperature and precipitation fields, when only the isotopic information was assimilated. The reconstruction skill for temperature and precipitation was especially high at low latitudes. This is due to the fact that isotopic proxies are strongly influenced by temperature and/or precipitation at low latitudes, which, in turn, are modulated by the El Niño–Southern Oscillation (ENSO) on interannual timescales. Subsequently, the proxy DA was conducted with real proxy data. The reconstruction skill was decreased compared to the OSSE. In particular, the decrease was significant over the Indian Ocean, eastern Pacific, and the Atlantic Ocean where the reproducibility of the proxy model was lower. By changing the experimental design in a stepwise manner, the decreased skill was suggested to be attributable to the misrepresentation of the atmospheric and proxy models and/or the quality of the observations. Although there remains a lot to improve proxy DA, the result adequately showed that proxy DA is feasible enough to reconstruct past climate.

Data assimilation (DA) has been applied to estimate the time‐mean state, such as annual mean surface temperature for paleoclimate reconstruction. There are two types of DA for this purpose: online‐DA and offline‐DA. The online‐DA estimates both time‐mean states (analyses) and initial conditions for subsequent DA cycles, while the offline‐DA only estimates the time‐mean analyses. If there is sufficiently long predictability in the system of interest compared to the temporal resolution of the observations, online‐DA is expected to outperform offline‐DA by utilizing information in the initial conditions. However, previous studies failed to show the superiority of online‐DA when time‐averaged observations are assimilated, and the reason has not been investigated thoroughly. This study compares online‐DA and offline‐DA and investigates the relation to the predictability using an intermediate complexity general circulation model with perfect‐model observing system simulation experiments. The result shows that the online‐DA outperforms offline‐DA when the length of predictability is longer than the averaging time of the observations. We also found that the longer the predictability, the more skillful the online‐DA. Here, the ocean plays a crucial role in extending predictability, which helps online‐DA to outperform offline‐DA. Interestingly, the observations of near‐surface air temperature over land are highly valuable to update the ocean variables in the analysis steps, suggesting the importance of using cross‐domain covariance information between the atmosphere and the ocean when online‐DA is applied to reconstruct paleoclimate.