Search for: All records

Abstract Atmospheric predictability research has long held that rapid error growth at small spatial scales imposes an intrinsic limit of roughly 2 weeks on deterministic weather forecast skill. We challenge this limit using GraphCast, a machine learning weather model, by optimizing initial conditions for twice-daily forecasts spanning 2020. This approach yields an average error reduction of 86% at 10 days relative to control forecasts from reanalysis initial conditions, with skill lasting beyond 30 days. Mean optimal initial-condition perturbations reveal large-scale, spatially coherent corrections primarily reflecting an intensification of the Hadley circulation. Forecasts using GraphCast-optimal initial conditions in the Pangu-Weather model achieve a 21% error reduction, peaking at 4 days, indicating that analysis corrections reflect adjustments that target both model and analysis error. These results demonstrate the existence of initial conditions producing skillful deterministic forecasts far beyond 2 weeks. Whether such initial conditions can be identified in real time for improving operational weather forecasts remains a topic of future research. Significance StatementOur results contest the long-standing view that rapid error growth at small spatial scales fundamentally limits reliable weather forecasts to about 2 weeks. By applying machine learning–based optimization to atmospheric initial conditions, we achieve skillful deterministic hindcasts to at least 33 days—more than twice the conventional limit. Validating these optimized initial conditions in an independent weather model leads to significant but smaller improvements, highlighting the importance of model error in the procedure. On average, the optimal corrections reveal physically consistent large-scale patterns that align with an intensified tropical overturning circulation. These results demonstrate the existence of initial conditions that considerably extend the current established limit of predictability. Real-time determination of such initial conditions for operational forecasting remains an open question. 

Abstract “Online” data assimilation (DA) is used to generate a seasonal-resolution reanalysis dataset over the last millennium by combining forecasts from an ocean–atmosphere–sea-ice coupled linear inverse model with climate proxy records. Instrumental verification reveals that this reconstruction achieves the highest correlation skill, while using fewer proxies, in surface temperature reconstructions compared to other paleo-DA products, particularly during boreal winter when proxy data are scarce. Reconstructed ocean and sea-ice variables also have high correlation with instrumental and satellite datasets. Verification against independent proxy records shows that reconstruction skill is robust throughout the last millennium. Analysis of the results reveals that the method effectively captures the seasonal evolution and amplitude of El Niño events, seasonal temperature trends that are consistent with orbital forcing over the last millennium, and polar-amplified cooling in the transition from the Medieval Climate Anomaly to the Little Ice Age. 

Abstract Historical observations of Earth’s climate underpin our knowledge and predictions of climate variability and change. However, the observations are incomplete and uncertain, and existing datasets based on these observations typically do not assimilate observations simultaneously across different components of the climate system, yielding inconsistencies that limit understanding of coupled climate dynamics. Here we use coupled data assimilation, which synthesizes observational and dynamical constraints across all climate fields simultaneously, to reconstruct globally resolved sea-surface temperature (SST), near-surface air temperature (T), sea-level pressure (SLP), and sea-ice concentration (SIC), over 1850–2023. We use a Kalman filter and forecasts from an efficient emulator (Linear Inverse Model; LIM) to assimilate observations of SST, land T, marine SLP, and satellite-era SIC. We account for model error by training LIMs on eight CMIP6 models, and we use the LIMs to generate eight independent reanalyses with 200 ensemble members, yielding 1600 total members. Key findings in the Tropics include post-1980 trends in the Walker circulation that are consistent with past variability, whereas the tropical SST contrast (the difference between warmer and colder SSTs) shows a distinct strengthening since 1975. ENSO amplitude exhibits substantial low-frequency variability and a local maximum in variance over 1875–1910. In polar regions, we find a muted cooling trend in the Southern Ocean post-1980 and substantial uncertainty. Changes in Antarctic sea ice are relatively small between 1850 and 2000, while Arctic sea ice declines by 0.5±0.1 (1σ) million km²during the 1920s. 

Abstract A deep learning (DL) model, based on a transformer architecture, is trained on a climate‐model data set and compared with a standard linear inverse model (LIM) in the tropical Pacific. We show that the DL model produces more accurate forecasts compared to the LIM when tested on a reanalysis data set. We then assess the ability of an ensemble Kalman filter to reconstruct the monthly averaged upper ocean from a noisy set of 24 sea‐surface temperature observations designed to mimic existing coral proxy measurements, and compare results for the DL model and LIM. Due to signal damping in the DL model, we implement a novel inflation technique by adding noise from hindcast experiments. Results show that assimilating observations with the DL model yields better reconstructions than the LIM for observation averaging times ranging from 1 month to 1 year. The improved reconstruction is due to the enhanced predictive capabilities of the DL model, which map the memory of past observations to future assimilation times. 

Paleoclimates provide examples of past climate change that inform estimates of modern warming from greenhouse-gas emissions, known as Earth’s climate sensitivity. However, differences between past and present climate change must be accounted for when inferring climate sensitivity from paleoclimate evidence. The closest paleoclimate analog to near-term warming from greenhouse-gas emissions is the Pliocene (5.3 to 2.6 Ma), a warm epoch with atmospheric CO₂concentrations similar to today. Recent reconstructions indicate the Pliocene was 1 °C warmer than previously thought, implying higher climate sensitivity, which is also supported by recent reconstructions showing more cooling with reduced CO₂at the Last Glacial Maximum (LGM; 19 to 23 thousand years ago). However, large-scale patterns of paleoclimate temperature change differ strongly from modern projections. Climate feedbacks and sensitivity depend on temperature patterns, and such “pattern effects” must be accounted for when using paleoclimates to constrain modern climate sensitivity. Here we combine data-assimilation reconstructions with atmospheric general circulation models to show Earth’s climate is more sensitive to Pliocene forcing than modern CO₂forcing. Pliocene ice sheets, topography, and vegetation alter patterns of ocean warming and excite destabilizing cloud feedbacks, and LGM feedbacks are similarly amplified by the North American ice sheets. Accounting for paleoclimate pattern effects produces a best estimate (median) for modern climate sensitivity of 2.8 °C and 66% CI of 2.4 to 3.4 °C (90% CI: 2.1 to 4.0 °C), substantially reducing uncertainty in projections of 21st-century warming. 

Abstract The traditional method for estimating weather forecast sensitivity to initial conditions uses adjoint models, which are limited to short lead times due to linearization around a control forecast. The advent of deep‐learning frameworks enables a new approach using backpropagation and gradient descent to iteratively optimize initial conditions, minimizing forecast errors. We apply this approach to the June 2021 Pacific Northwest heatwave using the GraphCast model, yielding over 90% reduction in 10‐day forecast errors over the Pacific Northwest. Similar improvements are found for Pangu‐Weather model forecasts initialized with the GraphCast‐derived optimal, suggesting that model error is an unimportant part of the perturbations. Eliminating small scales from the perturbations also yields similar forecast improvements. Extending the length of the optimization window, we find forecast improvement to about 23 days, suggesting atmospheric predictability at the upper end of recent estimates. 

Abstract Deep learning (DL)–based general circulation models (GCMs) are emerging as fast simulators, yet their ability to replicate extreme events outside their training range remains unknown. Here, we evaluate two such models—the hybrid Neural General Circulation Model (NGCM) and purely data‐driven Deep Learning Earth System Model (DLESyM)—against a conventional high‐resolution land–atmosphere model (HiRAM) in simulating land heatwaves and coldwaves. All models are forced with observed sea surface temperatures and sea ice over 1900–2020, focusing on the out‐of‐sample period (1900–1960). Both DL models generalize successfully to unseen climate conditions, broadly reproducing the frequency and spatial patterns of heatwave and coldwave events during 1900–1960 with skill comparable to HiRAM. An exception is over portions of North Asia and North America, where all models perform poorly during 1940–1960. Due to excessive temperature autocorrelation, DLESyM tends to overestimate heatwave and coldwave frequencies, whereas the physics–DL hybrid NGCM exhibits persistence more similar to HiRAM. 

Abstract Paleoclimate reconstructions are increasingly central to climate assessments, placing recent and future variability in a broader historical context. Paleoclimate reconstructions are increasingly central to climate assessments, placing recent and future variability in a broader historical context. Several estimation methods produce plumes of climate trajectories that practitioners often want to compare to other reconstruction ensembles, or to deterministic trajectories produced by other means, such as global climate models. Of particular interest are “offline” data assimilation (DA) methods, which have recently been adapted to paleoclimatology. Offline DA lacks an explicit model connecting time instants, so its ensemble members are not true system trajectories. This obscures quantitative comparisons, particularly when considering the ensemble mean in isolation. We propose several resampling methods to introduce a priori constraints on temporal behavior, as well as a general notion, called plume distance, to carry out quantitative comparisons between collections of climate trajectories (“plumes”). The plume distance provides a norm in the same physical units as the variable of interest (e.g. °C for temperature), and lends itself to assessments of statistical significance. We apply these tools to four paleoclimate comparisons: (1) global mean surface temperature (GMST) in the online and offline versions of the Last Millennium Reanalysis (v2.1); (2) GMST from these two ensembles to simulations of the Paleoclimate Model Intercomparison Project past1000 ensemble; (3) LMRv2.1 to the PAGES 2k (2019) ensemble of GMST and (4) northern hemisphere mean surface temperature from LMR v2.1 to the Büntgen et al. (2021) ensemble. Results generally show more compatibility between these ensembles than is visually apparent. The proposed methodology is implemented in an open-source Python package, and we discuss possible applications of the plume distance framework beyond paleoclimatology. 

Abstract Global deep learning weather prediction models have recently been shown to produce forecasts that rival those from physics-based models run at operational centers. It is unclear whether these models have encoded atmospheric dynamics or simply pattern matching that produces the smallest forecast error. Answering this question is crucial to establishing the utility of these models as tools for basic science. Here, we subject one such model, Pangu-Weather, to a set of four classical dynamical experiments that do not resemble the model training data. Localized perturbations to the model output and the initial conditions are added to steady time-averaged conditions, to assess the propagation speed and structural evolution of signals away from the local source. Perturbing the model physics by adding a steady tropical heat source results in a classical Matsuno–Gill response near the heating and planetary waves that radiate into the extratropics. A localized disturbance on the winter-averaged North Pacific jet stream produces realistic extratropical cyclones and fronts, including the spontaneous emergence of polar lows. Perturbing the 500-hPa height field alone yields adjustment from a state of rest to one of wind–pressure balance over ∼6 h. Localized subtropical low pressure systems produce Atlantic hurricanes, provided the initial amplitude exceeds about 4 hPa, and setting the initial humidity to zero eliminates hurricane development. We conclude that the model encodes realistic physics in all experiments and suggest that it can be used as a tool for rapidly testing a wide range of hypotheses. 

Abstract Despite increased Atlantic hurricane risk, projected trends in hurricane frequency in the warming climate are still highly uncertain, mainly due to short instrumental record that limits our understanding of hurricane activity and its relationship to climate. Here we extend the record to the last millennium using two independent estimates: a reconstruction from sedimentary paleohurricane records and a statistical model of hurricane activity using sea surface temperatures (SSTs). We find statistically significant agreement between the two estimates and the late 20th century hurricane frequency is within the range seen over the past millennium. Numerical simulations using a hurricane-permitting climate model suggest that hurricane activity was likely driven by endogenous climate variability and linked to anomalous SSTs of warm Atlantic and cold Pacific. Volcanic eruptions can induce peaks in hurricane activity, but such peaks would likely be too weak to be detected in the proxy record due to large endogenous variability.