The starting point is COSMO-REA6, a regional reanalysis for the European CORDEX domain developed by the Hans-Ertel-Centre for Weather Research together with the universities of Bonn and Cologne and the German Meteorological Service. Reanalyses blend numerical weather models with historical observations to produce a physically consistent record of past atmospheric conditions, and COSMO-REA6 does so on a grid of roughly six kilometers with forty vertical levels. But even at this resolution, the model’s deterministic gust diagnostic struggles with the chaotic, localized nature of wind gusts, which are produced by turbulent eddies deflecting fast-moving air parcels down to the surface or by convective downdrafts spreading horizontally when they strike the ground. These mechanisms, first formalized in diagnostics by researchers such as Brasseur and Nakamura, are deterministic by design and carry no measure of uncertainty, which is precisely what forecasters and emergency managers need most.

Ertz and Friederichs attack this problem with the tools of extreme value statistics, the branch of mathematics built specifically for rare, high-impact events. Because a gust measurement represents the highest three-second average wind speed within an hour, the theoretically appropriate model for daily maxima is the generalized extreme value distribution. After fitting single-station models and finding that the shape parameter showed no clear signal or leaned slightly negative, the authors deliberately fixed it at zero, yielding a Gumbel distribution. This choice avoids imposing an artificial upper bound on gust speeds, a property of the Weibull-type distribution that would be physically implausible for storm-driven extremes, while keeping parameter estimates stable.

The real innovation lies in how the model handles space. At the top level of the hierarchy, gust observations follow a Gumbel distribution whose location and scale parameters vary in both space and time. These parameters are expressed as linear regressions on predictor variables drawn from COSMO-REA6, chiefly the maximum ten-meter wind speed and the mean horizontal wind, plus a covariate capturing the difference between the station altitude and the model’s representation of that altitude. Crucially, the regression coefficients themselves are not treated as fixed numbers. Instead, each coefficient is a realization of a two-dimensional Gaussian random field with a constant mean and an isotropic Matérn covariance function, meaning that stations close to one another tend to share similar gust behavior. This latent-process structure allows information to be pooled across the entire network of 109 synoptic stations, stabilizing estimates at data-sparse locations and enabling kriging-based interpolation to any unobserved point in the domain.

Perhaps the most elegant trick in the paper is a correction to the distance metric itself. Mountain stations such as the Zugspitze or the Brocken can sit within a few kilometers of valley towns like Braunlage, yet their gust climates are worlds apart. Under a purely geographic distance measure, the model would wrongly treat these stations as near neighbors, contaminating the estimated spatial dependency structure for the whole region. The authors solve this by adding a scaled elevation difference to the great-circle distance between stations, effectively pushing mountain and valley locations apart in the covariance calculation. The scaling factor is treated as an unknown parameter with a prior centered on the ratio of buoyancy to Coriolis forces, a value of roughly one hundred derived from quasigeostrophic scale analysis. The result is that mountaintop observations can remain in the training data without distorting the picture elsewhere, and the estimated correlation lengths become larger and more stable.

Training such a model is no small computational feat. The authors used the Stan probabilistic programming language via its Python interface, running the No-U-Turn Sampler, an adaptive form of Hamiltonian Monte Carlo, for chains of 1500 iterations with 500 discarded as burn-in. The full spatial model required about an hour of sampling time, and the subsequent interpolation of parameter fields across a 121-by-161 grid cell domain over Germany took roughly three hours on an ordinary desktop computer, using an iterative scheme that simulates horizontal slices of the country one after another, each conditioned on its southern neighbor. The authors emphasize that the entire pipeline is reproducible on standard hardware and that the code and preprocessed data are openly available.

To find out whether all this statistical machinery actually pays off, the team subjected the model to a demanding verification regime. Data from odd-numbered years between 2001 and 2018 trained the model, while even-numbered years served for evaluation, and a leave-one-out cross-validation ensured that the station being predicted was always excluded from training. Performance was judged with proper scoring rules: the Brier score for probabilities of exceeding warning thresholds of 14, 18, 25, 29 and 33 meters per second, and the quantile score for predictive quantiles ranging from the 75th to the 99.9th percentile. Both scores were decomposed into miscalibration, resolution and uncertainty components using modern isotonic-regression-based estimators, allowing the authors to see not just whether the model was skillful but why.

The verdict is a qualified but meaningful success for the spatial approach. Against a spatially constant baseline model, which itself already achieved skill scores of 10 to 45 percent relative to local climatology, the spatial Bayesian model delivered median improvements of 2 to 5 percent for prediction quantiles, with the largest gains at the extreme upper quantiles that matter most for warnings. When the evaluation was restricted to the top quartile of forecast gust events, the spatial model’s advantage grew even more pronounced, and for the 99.9th percentile it nearly matched the skill of a model trained locally at each station, the theoretical ceiling of the approach. The score decomposition revealed that the gains came primarily from better resolution rather than better calibration, meaning the spatial model discriminates more effectively between windy and calm situations. For threshold exceedance probabilities the improvement was marginal, though the spatial model showed a slight edge at the highest 33 meters per second threshold.

The study is candid about its limits. At sheltered inland locations with weak mean winds, exemplified by Garmisch-Partenkirchen, even the locally trained model has no skill against climatology, because the reanalysis predictors simply lack explanatory power in complex, low-wind terrain. The spatial interpolation can also introduce bias where a station’s gust character differs sharply from its surroundings. Yet the overall message is compelling: by letting the data itself determine how gust behavior varies across the landscape, and by treating mountains as the statistical islands they truly are, the Bonn team has produced a proof of concept that spatially aware extreme value modeling can extract more reliable gust information from reanalysis archives. With extensions already envisioned for ensemble forecasts, diurnal and annual cycles, and land-sea contrasts, the framework could soon help weather services issue wind warnings that are not only more accurate but accompanied by the probabilities that decision-makers increasingly demand.

Subject of Research: Spatial Bayesian hierarchical extreme value post-processing of wind gusts from the COSMO-REA6 reanalysis over Germany

Article Title: Post-processing of wind gusts from COSMO-REA6 with a spatial Bayesian hierarchical extreme value model

Article References: Ertz, P., & Friederichs, P. (2025). Post-processing of wind gusts from COSMO-REA6 with a spatial Bayesian hierarchical extreme value model. Advances in Statistical Climatology, Meteorology and Oceanography, 11(2), 229-256. https://doi.org/10.5194/ascmo-11-229-2025

Image Credits: AI Generated

DOI: 10.5194/ascmo-11-229-2025

Keywords: wind gusts, extreme value statistics, Bayesian hierarchical model, COSMO-REA6, reanalysis, post-processing, spatial statistics, Gaussian random fields, Gumbel distribution, forecast verification, Germany, meteorology

The stakes of this evaluation could hardly be higher. In the summer of 2020, record-breaking heavy precipitation swept across South Korea, claiming 46 lives and leaving people missing, while inflicting property damage of roughly 860 million dollars. The memory of that disaster, alongside the catastrophic July 2021 rainfall in China’s Henan Province that killed 398 people and caused about 19 billion dollars in direct economic losses, frames the urgency of the research. With climate change expected to increase the frequency of extreme precipitation events, the question of whether AI models can be trusted to warn of such hazards has become one of the most consequential in modern meteorology.

GraphCast represents a fundamentally different approach to forecasting than the numerical weather prediction models that have dominated the field for decades. Rather than solving the physical governing equations of the atmosphere, GraphCast uses graph neural networks arranged in an encoder-processor-decoder configuration to learn how atmospheric states evolve, drawing on more than four decades of historical reanalysis data from 1979 to 2017. Trained on the European Centre for Medium-Range Weather Forecasts’ ERA5 dataset, the model autoregressively predicts meteorological variables, including precipitation, at six-hour intervals with a horizontal resolution of about 25 kilometers. Its computational speed is extraordinary, producing forecasts in minutes on a single processor where traditional models require supercomputers running for hours.

To assess whether that speed comes at the cost of accuracy, the research team selected the ten days in the summer of 2020 with the largest accumulated precipitation over South Korea, spanning late June through mid-August. These cases featured daily maximum precipitation exceeding 130 millimeters and countrywide averages ranging from roughly 30 to 70 millimeters. Crucially, the team evaluated the model against observations from 549 rain gauges operated by the Korea Meteorological Administration, a ground-truth network widely regarded as the most reliable source of precipitation data. This choice matters, because most previous evaluations of AI weather models relied on reanalysis or analysis products, which can themselves carry biases that flatter models trained on similar data.

For comparison, the researchers ran the Weather Research and Forecasting model, a physics-based numerical system, at three horizontal resolutions of 27, 9, and 3 kilometers. For each heavy precipitation day, they performed 27 WRF simulations using different combinations of microphysics, boundary-layer, and radiation schemes, averaging the results to account for the model’s sensitivity to parameterization choices. Both the AI and numerical models were initialized with the same ERA5 data, creating as fair a head-to-head comparison as the profound architectural differences between the two systems allow.

The results on broad precipitation patterns were genuinely impressive for GraphCast. Using metrics that measure overall spatial structure, including normalized mean bias, normalized centered root-mean-square error, and spatial correlation coefficient, the AI model matched or beat the numerical simulations. Its normalized mean bias of about negative 14 percent was comparable to the 27-kilometer WRF’s roughly negative 13 percent, and its centered error was the smallest of all the models tested. Its spatial correlation of about 0.54 exceeded the 0.46 of the similarly resolved WRF, though the 3-kilometer WRF reached higher at about 0.62. GraphCast also proved remarkably adept at capturing the synoptic setup of the storms, reproducing the strong low-pressure systems, the monsoon-frontal pressure gradients, and the atmospheric-river-like moisture transports exceeding 1250 kilograms per meter per second that funnel water vapor toward the Korean Peninsula.

But the picture darkened when the team examined precipitation by intensity. Using the equitable threat score, neighborhood versions of that score, and contingency-table metrics across thresholds from 5 to 100 millimeters, GraphCast showed a systematic distortion of the rainfall distribution. It overpredicted light precipitation at or below 20 millimeters, painting wider areas of appreciable rain than actually occurred, while severely underpredicting heavy precipitation at or above 70 millimeters. Averaged across the 70 to 100 millimeter thresholds, its frequency bias was a mere 0.22, compared with 0.61 for the 27-kilometer WRF, and its probability of detection was just 0.16. In eight of the ten cases, GraphCast missed the observed precipitation peaks above 100 millimeters entirely, producing smoothed fields that washed out the most dangerous local maxima.

A clever resolution-matching experiment revealed that this weakness was not simply an artifact of the model’s 25-kilometer grid. When the higher-resolution WRF outputs were degraded to 27 kilometers, their advantage at heavy thresholds persisted, indicating that GraphCast’s deficit lies in its ability to represent precipitation peaks with spatial scales of roughly 25 to 100 kilometers, features that are resolvable even at coarse resolution. The researchers trace this limitation to two likely culprits. First, the ERA5 training data themselves struggle to capture heavy precipitation, meaning the model may have learned a systematically muted version of extreme rainfall. Second, the use of mean squared error as the training objective tends to blur small-scale variability through the well-known double penalty effect, in which a correctly predicted feature that is slightly misplaced is penalized twice, once as a false alarm and once as a miss.

Perhaps the most novel contribution of the study is its examination of physical consistency, an aspect of AI weather models that had not previously been tested for precipitation. The team focused on the Sobaek Mountains, where topographic lifting enhances rainfall during warm-season events. Observations showed the terrain-adjacent region received 20 millimeters more precipitation than nearby lowland areas on average. GraphCast reproduced only 42 percent of that enhancement, or an 8-millimeter difference, while the WRF models captured 63 to 90 percent. Strikingly, although both GraphCast and the 27-kilometer WRF simulated nearly identical synoptic conditions and moisture transport intensities, the upward motions over the mountains were considerably weaker in GraphCast, failing to exceed the 0.5 pascals per second seen in the numerical model, despite nearly identical mountain heights in both simulations. This suggests the AI model, unconstrained by physical equations, fails to fully translate moisture-laden flow over terrain into the orographic lifting that drives enhanced precipitation.

The authors are careful to frame these findings not as a verdict on AI versus physics in general, but as a measurement of where the current generation of AI models stands. Their results echo earlier reports that machine learning models struggle with mesoscale variability in temperature, wind, and intense precipitation, and they point toward promising remedies: better objective functions than mean squared error, high-resolution open-access training datasets from convection-permitting models, and hybrid strategies that pair AI models for locating precipitation features with numerical models for quantifying their intensity. Given GraphCast’s enormous computational advantages and its demonstrated skill with large-scale patterns, the study suggests the future of high-impact weather forecasting may lie not in choosing between silicon and physics, but in designing systems that combine the strengths of both, ensuring that the deadliest details of a storm are no longer smoothed away.

Subject of Research: Skill assessment of the AI-based weather prediction model GraphCast in forecasting heavy precipitation over South Korea

Article Title: Skill assessment of AI-based weather model in heavy precipitation prediction over South Korea

Article References: Hong, S.-H., Park, K., Kim, D.-H., Baik, J.-J., & Jin, H.-G. (2026). Skill assessment of AI-based weather model in heavy precipitation prediction over South Korea. Theoretical and Applied Climatology, 157(9), Article 606. https://doi.org/10.1007/s00704-026-06536-w

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06536-w

Keywords: GraphCast, AI weather prediction, heavy precipitation, South Korea, WRF model, forecast verification, topographic precipitation, extreme weather, machine learning, ERA5, equitable threat score, physical consistency

The team focused on what forecasters call the subseasonal range, roughly two weeks to a month or more ahead, a notoriously awkward window that sits beyond the reach of weather forecasts but before the timescales where seasonal climate outlooks shine. To probe this gap, the researchers turned to the European Centre for Medium-Range Weather Forecasts Subseasonal to Seasonal, or S2S, ensemble reforecast dataset, a treasure trove of hindcast simulations produced by one of the world’s leading prediction systems. They compared these forecasts against high-quality sea surface temperature observations from the National Oceanic and Atmospheric Administration’s Optimum Interpolation Sea Surface Temperature product, evaluating performance across the tropical Indian Ocean for the twenty-year period from 2001 to 2020.

The headline finding is a hard limit on predictability: the effective sea surface temperature forecast lead time over the tropical Indian Ocean is about 31 days. Beyond roughly a month, the model’s temperature forecasts lose the skill needed to meaningfully constrain marine heatwave prediction. That number is more than a technical footnote. It defines the practical horizon within which fisheries managers, aquaculture operators, and conservation agencies could realistically act on a forecast before an oceanic heat event arrives, and it sets the benchmark against which future improvements to prediction systems will be measured.

But the study’s most striking results concern not just how far ahead forecasts work, but where and how they go wrong. When the researchers broke down the deterministic forecasts, meaning forecasts that give a single best estimate of future conditions, they found a sharp geographic split in the nature of the errors. In the near-equatorial region, the model systematically overestimated both the number of marine heatwave days and the cumulative intensity of events. The culprit was an excess of false positives, or false alarms, cases where the model predicted heatwave conditions that never actually materialized in the observations.

Off the equator, the story flipped. In these off-equatorial regions, the forecasts underestimated marine heatwave days and cumulative intensity, primarily because of false negatives, meaning misses where real heatwave events occurred but the model failed to flag them. These biases were not static; they grew steadily worse as the forecast lead time increased. At the maximum effective lead of 31 days, the overestimation of marine heatwave days in the near-equatorial region reached up to 82 percent, with cumulative intensity overestimated by as much as 62 percent. In the off-equatorial zones, the underestimation of days climbed to 47 percent and of cumulative intensity to 56 percent at the same lead time.

Why does the equator behave so differently from its surroundings in the model’s eyes? The researchers traced these regional differences closely to the spatial distribution of sea surface temperature anomaly forecast biases. In other words, where the forecast model systematically runs too warm or too cold relative to reality, the marine heatwave diagnostics inherit those errors. A warm bias near the equator pushes the model over the heatwave threshold too often, manufacturing spurious events, while cold biases elsewhere suppress the model’s ability to detect genuine warming episodes. This connection between background temperature bias and extreme event detection is a crucial insight, because it suggests that improving the mean state of the forecast should directly improve heatwave prediction.

There was one metric, however, where the errors were more uniform. The mean intensity of marine heatwaves, as opposed to their cumulative intensity or duration, was underestimated across the entire tropical Indian Ocean, regardless of latitude. This underestimation grew from about 10 percent at a one-day lead time to roughly 15 percent at 31 days. The implication is sobering: even when the model correctly identifies that a heatwave is underway, it tends to paint the event as milder than it truly is. For ecosystems sitting near their thermal tolerance limits, such as coral reefs where a single extra degree of sustained warmth can trigger mass bleaching, an intensity underestimate could translate into dangerously complacent risk assessments.

Deterministic forecasts tell only half the story, and the study’s probabilistic analysis offers a more encouraging picture. Rather than asking whether the model predicts a heatwave or not, probabilistic verification asks how well the forecast probabilities discriminate between events and non-events. Here the researchers used two standard tools: the Brier Skill Score, which measures whether a probabilistic forecast beats a naive reference forecast, and the Area Under the Curve, or AUC, which quantifies discrimination skill. The results showed that marine heatwave forecasting exhibits relatively high skill at lead times of one to seven days, with predominantly positive Brier Skill Scores and AUC values approaching 0.80, a level indicating strong ability to separate heatwave conditions from ordinary ones.

The practical stakes of this work are considerable. Marine heatwaves have been linked to devastating ecological outcomes worldwide, including coral bleaching, shifts in species distributions, and cascading effects on biodiversity and the ecosystem services that oceans provide. Their socioeconomic footprint is equally real, touching fisheries yields, aquaculture operations, and coastal livelihoods. In the Indian Ocean specifically, these events interact with the monsoon system and with climate modes such as the Indian Ocean Dipole and the Madden-Julian Oscillation, making the region both a hotspot for oceanic extremes and a critical piece of the global climate puzzle. A reliable early warning capability, even one limited to a few weeks of lead time, could allow fishers to adjust operations, aquaculture farms to deploy mitigation measures, and reef managers to prioritize interventions before the thermal stress peaks.

The study also maps out a clear agenda for improvement. Because the heatwave forecast errors track the underlying sea surface temperature anomaly biases, efforts to reduce those biases in subseasonal prediction models, whether through better ocean-atmosphere coupling, improved initialization, or higher resolution, should pay dividends for extreme event prediction. The finding that probabilistic skill remains respectable in the first week, while deterministic biases compound with lead time, suggests that forecast products for the tropical Indian Ocean may be most useful when they communicate uncertainty explicitly rather than offering a single deterministic answer. As global warming continues to push ocean temperatures upward and marine heatwaves grow longer, more frequent, and more intense, knowing precisely where our forecasting window ends, and where the blind spots lie within it, is an essential first step toward seeing the ocean’s next fever coming in time to act.

Subject of Research: Subseasonal forecast skill of marine heatwaves in the tropical Indian Ocean

Article Title: Evaluation of the subseasonal forecast skill of marine heatwaves in the tropical Indian ocean

Article References: Meng, L., Li, X., & Yan, Y. (2026). Evaluation of the subseasonal forecast skill of marine heatwaves in the tropical Indian ocean. Climate Dynamics, 64(11), Article 449. https://doi.org/10.1007/s00382-026-08374-y

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08374-y

Keywords: marine heatwaves, tropical Indian Ocean, subseasonal forecasting, sea surface temperature, ECMWF S2S, forecast verification, Brier Skill Score, SST anomaly bias, climate dynamics, coral bleaching, ocean warming, probabilistic forecasting

The study evaluated NOAA’s Seamless System for Prediction and Earth System Research, known as SPEAR, and the National Centers for Environmental Prediction’s Climate Forecast System version 2, or CFSv2, across the contiguous United States from 1991 to 2022. SPEAR is the newer of the two, developed at NOAA’s Geophysical Fluid Dynamics Laboratory with an atmospheric resolution of roughly 50 kilometers, and it runs a 15-member ensemble initialized from an ensemble coupled data assimilation system that blends in subsurface ocean temperature and salinity observations. CFSv2, operational since 2011, uses a coarser atmospheric grid of about 0.94 degrees and a 24-member lagged ensemble built from forecasts launched at four different times of day. Both systems were verified against the PRISM observational dataset, a high-resolution 4-kilometer gridded precipitation product built from quality-controlled weather stations, with sea surface temperatures checked against NOAA’s OISST and circulation fields against the ERA5 reanalysis.

The researchers deployed a battery of verification metrics designed to probe different aspects of forecast quality. Percentage bias quantified how much the models systematically over- or under-predicted rainfall, while the anomaly correlation coefficient measured whether forecasts tracked the year-to-year ups and downs of real precipitation. For the first time in this context, they also applied quantile-based categorical metrics, the critical success index, probability of detection, and false alarm ratio, to assess how well the models captured wet extremes of seasonal-mean precipitation, defined as seasons exceeding the 80th and 90th percentile thresholds of the climatological distribution. Statistical significance of model differences was tested with bootstrap resampling over 10,000 iterations, ensuring that the reported advantages of one model over the other were not artifacts of a small sample of hindcast years.

The headline finding is a stark seasonal asymmetry. During spring, both models exhibit modest but regionally coherent skill, with positive anomaly correlations across the West Coast, the north-central states, and parts of the Southwest. SPEAR’s CONUS-average spring correlations ranged from about 0.25 to 0.30, compared with roughly 0.18 to 0.28 for CFSv2, and the bootstrap test showed that SPEAR’s advantage was statistically significant only at the shortest one-month lead time. By summer, skill collapsed in both systems. SPEAR’s correlations fell to between 0.2 and 0.3 while CFSv2 hovered around 0.15, and over the South-Central United States, a case-study region encompassing Kansas, Oklahoma, and Texas, summer correlations dropped to nearly zero, exposing a profound inability of either system to anticipate summer rainfall anomalies in a region where drought and deluge both carry enormous economic stakes.

The bias analysis added further texture to the picture. In spring, both models were too wet over much of the central and eastern United States, but the problem was far worse in CFSv2, whose area-averaged gridded spring bias over the full continent reached 62 to 78 percent, compared with 31 to 33 percent for SPEAR. Over the South-Central region specifically, CFSv2’s spring wet bias climbed as high as 73 percent while SPEAR remained nearly neutral. Summer flipped the sign: SPEAR showed a robust dry bias over the continent and a severe 38 to 43 percent underestimation over the South-Central region, while CFSv2’s local wet and dry errors partially cancelled in the regional mean. The authors emphasize that how bias is summarized matters enormously, because compensating local errors can hide behind a seemingly benign regional average.

Perhaps the most practically troubling results concern wet extremes. When the researchers asked whether the models could identify seasons whose precipitation landed in the top 20 or top 10 percent of the historical distribution, the answer was largely no, especially in summer. Probability of detection declined sharply from the 80th to the 90th percentile threshold, while false alarm ratios climbed toward very high values across much of the country during June through August. Crucially, the spatial patterns of detection and false alarms mirrored the bias maps: in wet-biased regions the models trigger too many threshold exceedances, inflating both hits and false alarms simultaneously, whereas dry-biased regions suppress detections altogether. High detection rates, in other words, do not guarantee reliable forecasts when they come bundled with a flood of false alarms.

To understand why the models struggle, the team turned to the physics of the El Niño-Southern Oscillation, the dominant source of seasonal predictability for North American rainfall. An empirical orthogonal function analysis confirmed that the leading mode of tropical Pacific sea surface temperature variability during the study period was the canonical eastern-Pacific El Niño pattern, explaining 31.1 percent of SST variance and correlating at 0.95 with the standard Niño-3.4 index. Observations showed that the link between this tropical signal and South-Central U.S. rainfall is strongly seasonal: robust during winter and early spring, weakening through late spring and summer as regional convection, land-atmosphere feedbacks, and the Great Plains Low-Level Jet, which funnels Gulf of Mexico moisture into the continent’s interior, take center stage. Both models broadly reproduced this seasonal ebb and flow, though CFSv2 tended to overstate the SST-precipitation correlation in many seasons.

The centerpiece of the diagnostic work was a singular value decomposition analysis coupling tropical Pacific SST to CONUS precipitation. In spring, the observed leading mode explained 34.2 percent of the squared covariance and traced a coherent pathway: an El Niño-like SST pattern driving a Pacific-North American ridge-trough couplet in the mid-troposphere, with an anomalous ridge over the northeastern Pacific and a trough over the southeastern United States that steers Gulf moisture northward. Both models captured the qualitative structure of this spring teleconnection, but their 500-hectopascal geopotential height anomalies were only 40 to 60 percent of the observed amplitude. Strikingly, the ensemble-mean versions of the models showed squared covariance fractions of 70.5 percent for SPEAR and 87.7 percent for CFSv2, far above observations, yet when the decomposition was repeated for individual ensemble members the values fell back near the observed level. The inflation, the authors show, is a statistical artifact of ensemble averaging, which filters out chaotic internal atmospheric noise and isolates the forced SST signal rather than evidence that the models are genuinely over-dominated by El Niño.

In summer, the teleconnection machinery visibly breaks down in the models. The observed leading coupled mode shifts away from the canonical equatorial El Niño pattern toward western and off-equatorial Pacific anomalies, and the associated circulation features a deep anomalous trough over the central and eastern United States paired with rising motion that feeds continental rainfall. SPEAR retained some spatial fidelity, with a pattern correlation of 0.24 against the observed height regression, but CFSv2’s pattern correlation collapsed to 0.01, indicating a fundamental displacement of the summer circulation response. Vertical-velocity cross sections told the same story: the models produced weaker, smoother, and less vertically coherent ascent and descent anomalies than observations, consistent with known deficiencies in convective parameterizations and the simulation of Rossby wave propagation. An El Niño-like SST signal alone, the study concludes, is simply not enough; the atmospheric bridge that translates ocean warmth into rain must also be simulated faithfully.

The implications reach well beyond model scorecards. Seasonal forecasts underpin drought preparedness, reservoir operations, and agricultural planning, and the South-Central United States, a transition zone between the humid Southeast and the arid Southwest, sits squarely where those decisions hurt most when forecasts fail. The authors argue that future gains will require more than better ocean prediction: improved representation of SST-forced teleconnections, regional atmospheric variability, and additional warm-season drivers such as land-surface memory, soil moisture feedbacks, Gulf moisture transport, and modes of variability beyond ENSO. Until then, the study offers forecast users a candid calibration of what to trust: spring rainfall anomalies over the southern plains deserve guarded confidence, while summer outlooks, particularly for unusually wet seasons, should be treated with the skepticism that three decades of hindcast evidence now firmly justify.

Subject of Research: Seasonal forecast skill of the SPEAR and CFSv2 climate models for warm-season precipitation over the contiguous United States and its relation to ENSO teleconnections

Article Title: Relative skills of SPEAR and CFSv2 in foretelling warm-season precipitation anomalies over the conterminous United States and potential sources of errors

Article References: Nguyen, H., Zhang, Y., Bakhtar, A., Lu, F., & Fernando, N. (2026). Relative skills of SPEAR and CFSv2 in foretelling warm-season precipitation anomalies over the conterminous United States and potential sources of errors. Climate Dynamics, 64(9), Article 406. https://doi.org/10.1007/s00382-026-08357-z

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08357-z

Keywords: seasonal forecasting, SPEAR, CFSv2, precipitation, ENSO, teleconnections, climate models, drought, Great Plains, sea surface temperature, forecast verification, Climate Dynamics