Climate Variability and Data Assimilation Techniques

Climate variability encompasses natural fluctuations in temperature, precipitation and atmospheric circulation across timescales ranging from seasons to centuries. Key drivers include ocean–atmosphere interactions such as the El Niño–Southern Oscillation, shifts in large-scale pressure systems and variability in ocean currents. Understanding these fluctuations is essential for anticipating extreme events, managing water resources and informing adaptation strategies. Data assimilation techniques bridge observations—whether from satellites, weather stations or palaeoclimate proxies—and numerical models to produce coherent estimates of the state of the climate system. Approaches such as variational methods and ensemble filters allow assimilation both online (within a model run) and offline (against pre-computed ensembles), yielding reanalyses that underpin modern forecasting, climate monitoring and palaeoclimate reconstruction. Recent advances in machine learning and high-resolution modelling have improved the capacity to resolve subweekly to centennial variability, offering new insights into the mechanisms driving extremes and enhancing the skill of future projections.

Research from Nature Portfolio

Recent studies have reconstructed six centuries of Atlantic–European jet stream behaviour using monthly and daily atmospheric fields. By analysing jet strength, latitude and tilt against flood and drought reconstructions, researchers have shown that recent jet shifts fall within historical bounds and modulate European hydroclimate extremes on seasonal to annual scales. Volcanic forcing emerges as an important influence on jet variability.

Innovative machine-learning methods have demonstrated that a recurrent neural network can non-linearly reconstruct over 400 years of global monthly temperature anomalies from sparse pseudo-station data. The approach delivers realistic spatial patterns at low computational cost, rivalling conventional reconstruction techniques while flexibly adapting to different regions, periods and variables.

Research from all publishers

A global monthly palaeo-reanalysis covering 1421–2008 has been developed by blending ensemble atmospheric simulations with natural proxies and instrumental records. This offline data assimilation framework provides field estimates and an observation feedback archive, enabling separation of the roles played by proxy data and model forcings in driving reconstructed climate variability and extreme events.

Analyses of 40 CMIP6 climate models reveal distinct Arctic and Eurasian winter responses to two types of El Niño events. Central Pacific El Niño induces pan-Arctic warming and weaker Siberian cooling, whereas Eastern Pacific El Niño leads to strengthened Arctic pressure patterns and Eurasian surface air cooling, highlighting persistent uncertainties in reanalysis data sampling.

Intercomparison of multiple atmospheric reanalyses shows significant increases in subweekly temperature variability over Southern Hemisphere midlatitude landmasses during recent decades. Trends are linked to enhanced horizontal temperature advection and vary seasonally, with robust changes identified over South Africa, South America and Australia, illustrating both the strengths and biases of different assimilation systems.

Climate Variability and Data Assimilation Techniques publication trend

The graph below shows the total number of articles in climate variability and data assimilation techniques across all publications each year (not limited to Nature Index journals).

Technical terms

Data assimilation: A set of methods for merging observational data with model forecasts to produce a consistent estimate of the climate or weather state.

Reanalysis: A retrospective climate dataset created by applying data assimilation to historical observations within a numerical model framework.

Proxy record: An indirect climate indicator—such as tree rings, ice cores or sediment layers—that provides information on past environmental conditions.

Ensemble Kalman filter: An assimilation technique that uses a collection of model simulations to estimate error covariances and update state estimates sequentially.

El Niño–Southern Oscillation (ENSO): A coupled ocean–atmosphere phenomenon in the tropical Pacific that drives interannual climate variability worldwide.

References

1. Past hydroclimate extremes in Europe driven by Atlantic jet stream and recurrent weather patterns. Nature Geoscience (2025).
2. Artificial intelligence achieves easy-to-adapt nonlinear global temperature reconstructions using minimal local data. Communications Earth & Environment (2023).
3. ModE-RA: a global monthly paleo-reanalysis of the modern era 1421 to 2008. Scientific Data (2024).
4. Distinct impacts of two types of El Niño events on northern winter high-latitude temperatures simulated by CMIP6 climate models. Environmental Research Letters (2023).
5. Seasonally dependent increases in subweekly temperature variability over Southern Hemisphere landmasses detected in multiple reanalyses. Weather and Climate Dynamics (2024).