Category: Publications

Dominant modes of interannual winter SAT covariability between the Arctic and the Tibetan Plateau: spatio-temporal structures and dynamical linkages

Sun, X., Y. Gao, X.-Q. Yang, Z. Fang, X. Zhang, S. Yuan, N. S. Keenlyside 2025: Dominant modes of interannual winter SAT covariability between the Arctic and the Tibetan Plateau: spatio-temporal structures and dynamical linkages. Clim Dyn. https://doi.org/10.1007/s00382-025-07711-x

Summary: As two highly sensitive climate zones in the world, the Arctic and Tibetan Plateau (TP) regions respectively exhibit significantly uneven spatial variability in surface air temperature (SAT) and greatly influence the Eurasian climate on the interannual timescale. However, despite the synchronized warming trends in these two regions, their interannual spatio-temporal connection remains unclear. In this study, we applied the singular value decomposition (SVD) method to ERA5 wintertime surface air temperature anomalies to explore the dominant modes of SAT covariability between the Arctic and TP. We identified two major interannual modes: the dipolar Arctic-uniform TP (DA-UTP) and the quadrupolar Arctic-dipolar TP (QA-DTP), which together explain 82% of their covariance. The DA-UTP mode resembles the negative phase of the Arctic Oscillation, characterized by a hemispheric-scale pattern of “warm northern North America—cold northern Eurasia—warm TP”, while the QA-DTP mode exhibits a meridional teleconnection in the eastern hemisphere, featuring “warm Barents and Kara Seas—cold Eurasia—warm southern TP”. Both modes primarily draw energy from the North Atlantic Ocean and affect East Asian through the atmospheric Rossby wave train. The corresponding North Atlantic SST anomalies display a tripolar distribution, with the center of the negative SST gradient anomaly in the second mode shifted southward compared to the first. These two climate modes further modulate synoptic and sub-seasonal-to-seasonal winter temperature anomalies in Eurasia by altering the hemispheric-scale temperature gradient. The findings of this study contribute to a deeper knowledge and understanding of the interannual spatial and temporal relationships of wintertime surface temperature anomalies between the Arctic and TP.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Predicting Atlantic and Benguela Niño events with deep learning

Bachèlery ML, Brajard J, Patacchiola M, Illig S, Keenlyside N 2025: Predicting Atlantic and Benguela Niño events with deep learning. Sci. Adv.. https://doi.org/10.1126/sciadv.ads5185

Summary: Atlantic and Benguela Niño events substantially affect the tropical Atlantic region, with far-reaching consequences on local marine ecosystems, African climates, and El Niño Southern Oscillation. While accurate forecasts of these events are invaluable, state-of-the-art dynamic forecasting systems have shown limited predictive capabilities. Thus, the extent to which the tropical Atlantic variability is predictable remains an open question. This study explores the potential of deep learning in this context. Using a simple convolutional neural network architecture, we show that Atlantic/Benguela Niños can be predicted up to 3 to 4 months ahead. Our model excels in forecasting peak-season events with remarkable accuracy extending lead time to 5 months. Detailed analysis reveals our model’s ability to exploit known physical precursors, such as long-wave ocean dynamics, for accurate predictions of these events. This study challenges the perception that the tropical Atlantic is unpredictable and highlights deep learning’s potential to advance our understanding and forecasting of critical climate events.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Amplified wintertime Arctic warming causes Eurasian cooling via nonlinear feedback of suppressed synoptic eddy activities

Yin, M.,  X.-Q. Yang, L. Sun, L. Tao, N. Keenlyside 2025: Amplified wintertime Arctic warming causes Eurasian cooling via nonlinear feedback of suppressed synoptic eddy activities. Sci. Adv.. https://www.science.org/doi/10.1126/sciadv.adr6336

Summary: The amplified wintertime Arctic warming has accelerated in recent decades. However, whether and how the warming can cause Eurasian cooling remains debated. By identifying daily Arctic warming events, we find direct observational evidence that the Arctic warming tends to cause substantial Eurasian cooling and an increase in occurrence frequency of Eurasian cooling events with a roughly 2-day lag. We propose a mechanism explaining the causality. We find that the Arctic warming causes a large suppression in activities of daily weather disturbances (referred to as synoptic eddies) over high-latitude Eurasia. This produces a meridional dipole in geopotential height anomalies characterized by an equivalent-barotropic anomalous low (high) and a lower-level cooling (warming) over mid-latitude Eurasia (the Arctic) via a nonlinear eddy–to–mean flow feedback. The feedback induces near-surface northeasterly anomalies that enlarge the Eurasian cooling via cold advection. Thus, we conclude that the warm Arctic versus cold Eurasia is essentially an intrinsic dipole determined by synoptic eddy–mean flow interaction.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

A novel ensemble-based parameter estimation for improving ocean biogeochemistry in an Earth system model

Singh, T., Counillon, F., Tjiputra, J., Wang, Y. 2025: A novel ensemble-based parameter estimation for improving ocean biogeochemistry in an Earth system model. JAMES. https://doi.org/10.1029/2024MS004237

Summary: Estimating ocean biogeochemistry (BGC) parameters in Earth System Models is challenging due to multiple error sources and interlinked parameter sensitivities. Reducing the temperature and salinity bias in the ocean physical component of the Norwegian Earth System Model (NorESM) diminishes the BGC state bias at intermediate depth but leads to a greater bias increase near the surface. This suggests that BGC parameters are tuned to compensate for the physical ocean model biases. We successfully apply the iterative ensemble smoother data assimilation technique to estimate BGC parameters in NorESM with reduced bias in its physical ocean component. We estimate BGC parameters based on the monthly climatological error of nitrate, phosphate, and oxygen in a coupled reanalysis of NorESM that assimilates observed monthly climatology of temperature and salinity. First, we compare the performance of globally uniform and spatially varying parameter estimations. Both approaches reduce BGC bias obtained with default parameters, even for variables not assimilated in the parameter estimation (e.g., CO2 fluxes and primary production). While spatial parameter estimation performs locally best, it also increases biases in areas with few observations, and overall performs poorer than global parameter estimation. A second iteration further reduces the bias in the near-surface BGC with global parameter estimation. Finally, we assess the performance of global estimated parameters in a 30-year coupled reanalysis produced by assimilating time-varying temperature and salinity observations. This reanalysis reduces error by 10%–20% for phosphate, nitrate, oxygen, and dissolved inorganic carbon compared to a reanalysis done with default parameters.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Impact of Ocean, Sea Ice or Atmosphere Initialization on Seasonal Prediction of Regional Antarctic Sea Ice

Xiu, Y., Wang, Y., Luo, H., Garcia-Oliva, L., Yang, Q. 2025: Impact of ocean, sea ice or atmosphere initialization on seasonal prediction of regional Antarctic sea ice. JAMES. https://doi.org/10.1029/2024MS004382

Summary: This study investigates how the atmosphere, ocean, or sea ice observations affect the seasonal prediction of Antarctic sea ice. We analyze three sets of predictions from the Norwegian Climate Prediction Model, each integrating different data sets of the atmosphere, ocean, or sea ice. Initially, we assess the seasonal cycles, trends, and variability of Antarctic sea ice in these data sets. We found that including atmosphere observations gave the best seasonal cycle compared to the observed sea ice. However, the linear trend in sea ice when including atmospheric data is poorly reproduced in the western Southern Ocean. Regarding variability, including the combined ocean and sea ice data gave the best performance. Next, we assess the accuracy of regional Antarctic sea ice prediction. We found that the accuracy varies with region and season. Austral winter predictions in western Antarctic have some skill up to a year in advance, while those in the eastern Antarctic are less reliable. Predictions based on atmosphere data are generally more accurate than those based on ocean or ocean/sea-ice data, especially when predicting from July or October. Interestingly, once ocean data is used, involving additional sea ice data improves sea ice concentration in the reanalysis but not in the predictions.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Ice-kNN-South: A Lightweight Machine Learning Model for Antarctic Sea Ice Prediction

Lin, Y., Yang, Q., Li, X., Dong, X., Luo, H., Nie, Y., et al. 2025: Ice-kNN-south: A lightweight machine learning model for Antarctic sea ice prediction. JAMES. https://doi.org/10.1029/2024JH000433

Summary: Antarctic sea ice has undergone a transition, with more frequent extreme minimum events observed since 2014, emphasizing the ongoing need for accurate predictions. We developed a lightweight machine learning model called Ice-k-nearest neighbor (kNN)-South to improve Antarctic sea ice prediction. Compared with commonly used benchmarks, such as anomaly persistence, climatology, and the European Centre for Medium-Range Weather Forecasts predictions, the Ice-kNN-South shows skillful predictions for almost 90 lead days, especially in summer. Even in years with extreme minimum sea ice areas, Ice-kNN-South shows strong skill in predicting Antarctic sea ice cover. Additionally, due to its minimal computational resource requirements, Ice-kNN-South shows promise for operational and real-time Antarctic sea ice prediction applications.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Evaluation of the effects of Argo data quality control on global ocean data assimilation systems

Ishikawa I, Fujii Y, de Boisseson E, Wang Y and Zuo H 2024: Evaluation of the effects of Argo data quality control on global ocean data assimilation systems. Front Mar Sci. https://doi.org/10.3389/fmars.2024.1496409

Summary: A series of observing system experiments (OSEs) were conducted in order to evaluate the effects of Argo data quality control (QC), by using the three global ocean data assimilation systems. During the experimental period between 2015 and 2020, some Argo floats are affected by the abrupt salinity drifts, which caused spurious increasing trend of the global mean salinity in the reanalyses using the observations with only real-time QC applied. The spurious trend is mitigated by applying the gray list provided by the Argo Global Data Assembly Centres (GDAC), and further reduced by assimilating the delayed-mode Argo data of the Argo GDAC instead of the real-time Argo data. These impacts of the Argo QC are generally consistent among the three ocean data assimilation systems. Further investigations in the JMA’s system show that errors in the analyzed salinity with respect to the delayed-mode Argo data are smaller in the OSE with more rigorous QC, and the spatiotemporal variations in the sea-surface dynamic height are reproduced better. Additionally, QC impacts on the analyzed temperatures are shown not to directly reflect the difference in temperature observations among OSEs, and may be affected by difference in the salinity observations among OSEs through the cross-covariance relationship in the data-assimilation systems.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Improving subseasonal forecast skill in the Norwegian Climate Prediction Model using soil moisture data assimilation

Nair, A.S., Counillon, F., Keenlyside, N. 2024: Improving subseasonal forecast skill in the Norwegian Climate Prediction Model using soil moisture data assimilation. Clim Dyn. https://doi.org/10.1007/s00382-024-07444-3

Summary: This study shows the importance of soil moisture (SM) in subseasonal-to-seasonal (S2S) predictions at mid-latitudes. We do this through introducing the Norwegian Climate Prediction Model Land (NorCPM-Land), a land reanalysis framework tailored for integration with the Norwegian Climate Prediction Model (NorCPM). NorCPM-Land assimilates blended SM data from the European Space Agency’s Climate Change Initiative into a 30-member offline simulation of the Community Land Model with fluxes from the coupled model. The assimilation of SM data reduces error in SM by 10.5 % when validated against independent SM observations. It also improves latent heat flux estimates, illustrating that the adjustment of underlying SM significantly augments the capacity to model land surface dynamics. We evaluate the added value of land initialisation for subseasonal predictions, by comparing the performance of hindcasts (retrospective prediction) using the standard NorCPM with a version where the land initial condition is taken from NorCPM-Land reanalysis. The hindcast covers the period 2000 to 2019 with four start dates per year. Land initialisation enhances SM predictions, reducing error by up to 2.5-month lead time. Likewise, the error for precipitation and temperature shows improvement up to a lead time of 1.5-month. The largest improvements are observed in regions with significant land-atmospheric coupling, such as the Central United States, the Sahel, and Central India. This method further enhances the prediction of extreme temperature variations, both high and low, with the most notable improvements seen in regions at mid and high latitudes, including parts of Europe, the United States, and Asia. Overall, our study provides further evidence for the significant role of SM content in enhancing the accuracy of subseasonal predictions. This study introduces a technique for improved land initialisation, utilising the same model employed in climate predictions.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Coupled data assimilation for climate prediction: a focus on ocean-atmosphere coupling (PhD thesis)

Lilian Carolina Garcia Oliva (2024-10-17): Coupled data assimilation for climate prediction: a focus on ocean-atmosphere coupling. PhD thesis, University of Bergen, Bergen, Norway. https://hdl.handle.net/11250/3157446

Summary: Seasonal-to-Decadal (S2D) climate predictions can provide decision-making information for diverse sectors, such as food security, energy and climate adaptation. The initial condition of the ocean is fundamental for providing skilful S2D predictions. A method to estimate the ocean’s initial condition is by merging the model and observations through a process called Coupled Data Assimilation (CDA). Ocean observations have demonstrated their potential to achieve skilful prediction. The Norwegian Climate Prediction Model (NorCPM) features an advanced Ocean Data Assimilation (ODA) scheme based on an ensemble method. This thesis outlines our efforts to improve S2D predictions within the NorCPM using atmospheric observations.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.

Hybrid covariance super-resolution data assimilation

Barthélémy, S., Counillon, F., Brajard, J., Bertino, L. 2024: Hybrid covariance super-resolution data assimilation. Ocean Dynamics. https://doi.org/10.1007/s10236-024-01643-6

Summary: The super-resolution data assimilation (SRDA) enhances a low-resolution (LR) model with a Neural Network (NN) that has learned the differences between high and low-resolution models offline and performs data assimilation in high-resolution (HR). The method enhances the accuracy of the EnKF-LR system for a minor computational overhead. However, performance quickly saturates when the ensemble size is increased due to the error introduced by the NN. We therefore combine the SRDA with the mixed-resolution data assimilation method (MRDA) into a method called “Hybrid covariance super-resolution data assimilation” (Hybrid SRDA). The forecast step runs an ensemble at two resolutions (high and low). The assimilation is done in the HR space by performing super-resolution on the LR members with the NN. The assimilation uses the hybrid covariance that combines the emulated and dynamical HR members. The scheme is extensively tested with a quasi-geostrophic model in twin experiments, with the LR grid being twice coarser than the HR. The Hybrid SRDA outperforms the SRDA, the MRDA, and the EnKF-HR at a given computational cost. The benefit is the largest compared to the EnKF-HR for small ensembles. However, even with larger computational resources, using a mix of high and low-resolution members is worth it. Besides, the Hybrid SRDA, the EnKF-HR, and the SRDA, unlike the MRDA, prevent the smoothing of dynamical structures of the background error covariance matrix. The Hybrid SRDA method is also attractive because it is customizable to available resources.

Link to publication. You are most welcome to contact us or the corresponding author(s) directly, if you have questions.