The investigation centered on four public datasets. The single-nucleus RNA subset contained healthy nasal mucosa library HC1 and SNSCC tumor library CA5. Bulk RNA sequencing data contained seven SNSCC tumors and seven normal nasal mucosa samples, while a microarray dataset supplied five matched tumor-adjacent-normal pairs. A separate set of 21 nonkeratinizing sinonasal cancer samples was used only as tumor-only context and was not treated as independent tumor-normal validation. The researchers did not substitute head-and-neck cancer data from The Cancer Genome Atlas for missing SNSCC material and did not make an HPV-positive versus HPV-negative inference. After quality control and doublet detection, the single-nucleus dataset contained 14,479 nuclei: 6,817 from healthy tissue and 7,662 from the tumor.

The retained nuclei were annotated into 11 biological cell types, including epithelial, secretory epithelial, basal epithelial, ciliated epithelial, tuft, endothelial, perivascular, fibroblast, myeloid, T and NK, and plasma cells. The largest observed compositional difference was a 44.7-percentage-point increase in epithelial cells in the tumor library relative to the healthy library. Secretory epithelial cells showed the largest decrease, at 56.2 percentage points. These figures describe the two available libraries and do not establish how cell populations change in patients generally. The authors also reported Shannon diversity values of 1.444 for the healthy library and 1.285 for the tumor library, but did not treat cell-level nuclei as independent patient replicates when calculating program scores or reporting comparisons.

Bulk RNA sequencing provided the study’s principal evidence for tumor-versus-normal differential expression. An edgeR robust quasi-likelihood model evaluated 17,677 genes across the 14 bulk libraries and identified 5,732 genes at a false-discovery-rate threshold below 0.05. In an analysis of nine predefined gene programs, cell cycle showed the largest mean standardized shift, increasing by 1.727 units in tumors relative to normal samples. Running-enrichment analysis gave an enrichment score of 0.926 for an interferon-response program and 0.733 for a fibroblast extracellular-matrix program, whereas a myeloid-activation program had a score of −0.424. Despite enrichment of individual interferon-response genes, the composite antigen-presentation score was lower in tumor samples, at −0.238 versus 0.238 in normal tissue, indicating that the elevated genes did not amount to uniform induction of that entire program.

Integration between the bulk and single-nucleus layers showed a moderate correspondence. Exact gene-symbol matching was possible for 16,148 genes, and bulk log fold changes correlated with observed single-nucleus pseudobulk shifts at a Spearman rho of 0.543. Among 1,615 genes in the top 10% by absolute single-nucleus shift, 975 also belonged to the bulk set meeting the false-discovery-rate criterion. Program-level effects across nine predefined sets showed a cross-scale Spearman rho of 0.467. CD74, CCL5, IL32 and HLA-DRA were highlighted because their observed directions were consistent across the layers. Their patterns involved different cellular compartments: CD74 was enriched in myeloid cells in the single-nucleus atlas, while CCL5 and IL32 were concentrated in immune-active populations and HLA-DRA was associated with MHC class II expression.

The paired microarray dataset supplied additional context without extending the claims beyond its measurement limits. A paired limma model tested 60,755 probes and identified 2,620 features at an adjusted probability below 0.05 across the five tumor-adjacent-normal pairs. Because the platform annotation did not permit reliable genome-wide conversion to gene symbols, the researchers retained probe identifiers and did not present these results as gene-level validation of the four candidates. Expression distributions in the 21 tumor-only nonkeratinizing sinonasal cancer samples were also treated as context rather than evidence of a tumor-normal difference or replication of the bulk findings. The cross-layer candidate rankings were filtered for measured expression, bulk statistical significance and concordant direction, making the nominated panel exploratory rather than a comprehensive marker-discovery assay.

Laboratory experiments offered preliminary support for the four leading candidates. The investigators compared two SNSCC cell lines, RPMI 2650 and Detroit 562, with normal bronchial epithelial BEAS-2B cells using quantitative reverse-transcription PCR and enzyme-linked immunosorbent assays. Relative to BEAS-2B cells, RPMI 2650 cells showed mean messenger RNA increases of 4.27-fold for CD74, 3.15-fold for CCL5, 3.68-fold for IL32 and 3.82-fold for HLA-DRA, with all reported P values below 0.01. Detroit 562 cells also showed significant increases, ranging from 2.46-fold for CCL5 to 3.53-fold for CD74, with all P values below 0.01. ELISA measurements showed significantly elevated concentrations of all four proteins in both cancer lines, including a 4.68-fold increase in CD74 protein in RPMI 2650 cells, with P below 0.001. These were cancer-versus-normal cell-culture comparisons, not tests of diagnostic performance or treatment response.

The biological interpretations remain bounded by the design. CCL5 is a chemokine associated with recruitment of T cells, NK cells and other immune effectors, while IL32 is linked to inflammatory signaling and has been implicated in antitumor immune activity. CD74 encodes the MHC class II invariant chain involved in antigen processing, and HLA-DRA is an MHC class II gene, so their coordinated elevation is compatible with selected immune-related responses. However, the assays did not measure immunotherapy response, and neither RPMI 2650 nor Detroit 562 was an established HPV-positive model. The experiments therefore cannot show whether the candidates differ by HPV subtype or predict sensitivity to immunotherapy. Likewise, enrichment of interferon-response genes in the bulk comparison does not establish a clinical response to checkpoint blockade.

Several technical and clinical gaps constrain the conclusions. The single-nucleus atlas represents one library per condition, so its thousands of nuclei cannot substitute for independent patient replication, and library-level differences may reflect factors beyond disease status. Ambient RNA was not estimated because empty-droplet matrices were unavailable, raising the possibility that some low-level signals, including ribosomal signals, were influenced by technical contamination. HPV annotations came from deposited metadata rather than new viral sequencing, E6 or E7 detection, or p16 staining. The immune signatures grouped cells at broad lineage level rather than resolving CD4 and CD8 T-cell subsets, regulatory T cells, distinct B-cell populations, macrophages, dendritic cells or myeloid-derived suppressor cells. Independent cohorts, spatial tissue studies, immunohistochemistry and appropriately powered HPV-matched analyses would be needed to confirm the origins and clinical significance of the signals.

Overall, the reanalysis provides a descriptive map of tumor-normal remodeling in SNSCC and nominates CD74, CCL5, IL32 and HLA-DRA as candidate markers for further study. Its strongest supported result is the widespread bulk differential-expression landscape, including a pronounced cross-layer cell-cycle signal supported by concordant observations in the smaller single-nucleus comparison. The four candidates additionally showed concordant elevation in public transcriptomic layers and increased messenger RNA and protein abundance in two SNSCC cell lines. These results do not establish diagnostic performance, HPV-specific biology or immunotherapy utility. Instead, they identify experimentally tractable signals for validation in larger independent tissue cohorts, with particular attention to immune-cell localization, spatial organization and relationships to HPV status and treatment outcome.

Subject of Research: Exploratory multi-platform transcriptomic analysis of tumor-normal dysregulation in sinonasal squamous cell carcinoma

Article Title: Reanalysis of public bulk and single-nucleus transcriptomes identifies widespread tumor-normal dysregulation and a cross-layer cell-cycle signal in sinonasal squamous cell carcinoma

Article References: Yuan, J., Wang, Y., Lin, K., & Sun, L. (2026). Reanalysis of public bulk and single-nucleus transcriptomes identifies widespread tumor-normal dysregulation and a cross-layer cell-cycle signal in sinonasal squamous cell carcinoma. Frontiers in Genetics, 17, Article 1918021. https://doi.org/10.3389/fgene.2026.1918021

Image Credits: AI Generated

DOI: 10.3389/fgene.2026.1918021

Keywords: SNSCC, single-nucleus RNA sequencing, bulk RNA sequencing, cell-cycle regulation, CD74, CCL5, IL32, HLA-DRA, tumor immunology, biomarker discovery, transcriptomics, Reanalysis

The researchers combined reanalysis data with simulations representing the present day and the Last Glacial Maximum. Their analysis indicates that PDO regime shifts regulate spring precipitation by changing where subtropical troughs sit and altering the associated wind patterns. The results therefore provide a proposed mechanism connecting decadal climate variability, regional moisture transport and drought extent, rather than showing that the PDO alone causes every change in rainfall.

During a negative PDO phase, the India-Burma trough deepens and the base of its circulation shifts about 1,330 km equatorward. This configuration produces northeasterly anomalies east of the trough, weakening the transport of moisture into East Asia and suppressing precipitation. At the same time, the North American subtropical trough becomes shallower and its base moves about 560 km poleward, encouraging ridge development that reduces precipitation in the water-stressed western part of North America.

The comparison between Last Glacial Maximum and present-day simulations suggests that subsequent warming has enlarged the geographic footprint of drying linked to negative-PDO conditions. The reported 160% and 57% increases refer specifically to affected area under the study’s negative-PDO drying effect; they are not estimates of total regional drought area or direct measurements of future drought risk. The simulations represent a warming interval similar in magnitude to that projected for this century, offering a basis for assessing how PDO-related circulation changes may interact with a substantially warmer climate.

Subject of Research: How warming since the Last Glacial Maximum amplified Pacific Decadal Oscillation-driven drought expansion

Article Title: Warming since the Last Glacial Maximum amplifies Northern Hemisphere drought expansion driven by the Pacific Decadal Oscillation

Article References: Zhang, C., Duan, A., Liang, M., Cao, S., Xie, Y., & Dai, A. (2026). Warming since the Last Glacial Maximum amplifies Northern Hemisphere drought expansion driven by the Pacific Decadal Oscillation. Communications Earth & Environment. https://doi.org/10.1038/s43247-026-04186-4

Image Credits: AI Generated

DOI: 10.1038/s43247-026-04186-4

Keywords: Pacific Decadal Oscillation, East Asian drought, North American drought, subtropical troughs, spring precipitation, Last Glacial Maximum, climate warming, moisture transport, climate models, reanalysis, Warming, since

The study began as a comparison of lifestyle and noncommunicable disease feature importance between Indian (n = 528) and Japanese (n = 70) university students, using cross-sectional data collected from June to August 2024. India and Japan present contrasting epidemiological and dietary profiles that made the comparison appealing. India faces a rapid nutrition transition, with sedentary behaviour independently associated with obesity risk among urban Indian adults, while Japan has undergone decades of continuous dietary westernization alongside a strong cultural link between lifestyle behaviour change and subjective well-being. Body mass index also carries different metabolic implications in South Asian populations than in Western reference populations, making BMI-based thresholds an imperfect, population-dependent proxy for adiposity rather than a universal standard.

But the preliminary analysis contained three specific, checkable pitfalls that the authors identified in their own work. First, the outcome of obesity, defined as BMI ≥ 25 kg/m², was tautologically predicted by including BMI itself as a model input. Second, model evaluation was performed without a held-out test set, meaning the models were evaluated on the same data they were trained on. Third, feature-importance comparisons ignored sampling variability from an eightfold difference in cohort size. The team addressed each problem using penalized logistic regression, Random Forest, and gradient-boosted trees with proper train/test splitting, held-out permutation importance, and bootstrap confidence intervals.

The most dramatic result came from quantifying the BMI-obesity circularity. With BMI included as a predictor, all three India-obesity models achieved perfect discrimination on the held-out test set, with an AUC of 1.000 for logistic regression, Random Forest, and XGBoost alike. BMI’s Gini importance was 0.91 with a bootstrap 95% confidence interval of [0.84, 0.92], and its held-out permutation importance was 0.50, meaning permuting BMI alone destroys half of the model’s discriminative AUC. This is definitional, not empirical: obesity is BMI ≥ 25 by construction. When BMI is removed from the predictor set, test AUC collapses to 0.568 for logistic regression, 0.526 for Random Forest, and 0.503 for XGBoost, indistinguishable from chance. Lifestyle features carry essentially no independent discriminative signal for obesity once the tautological predictor is removed.

For the remaining viable outcome, elevated self-rated stress, held-out predictive performance was modest at best. In India, test AUC was 0.67 for logistic regression, 0.67 for Random Forest, and 0.71 for XGBoost, with a bootstrap 95% confidence interval of [0.49, 0.72]. In Japan, the models did not clearly exceed chance, with AUCs of 0.57, 0.57, and 0.49, and a very wide bootstrap confidence interval of [0.29, 0.82] reflecting Japan’s small test set of only 14 observations. None of the AUC values in the 0.92 to 0.98 range obtained under the preliminary, non-held-out evaluation approach were reproduced once a genuine held-out test set was used.

The reanalysis also revealed a critical discrepancy in how feature importance was measured. Gini importance, computed on the training fit, reflects how much a variable reduced impurity during tree construction, while permutation importance on held-out data reflects how much actual predictive information the model loses without that variable. These two measures frequently disagreed. Most notably, sleep duration’s Gini importance in Japan was 0.22, superficially close to the 0.47 value initially reported as headline evidence for a Japanese psychosocial phenotype, but its held-out permutation importance was −0.004, indicating no measurable out-of-sample discriminative contribution at all. By contrast, BMI with a permutation importance of 0.067 and exercise frequency at 0.040 contributed more to real, held-out predictive performance than sleep duration did. The claim that sleep duration is the primary stress predictor in Japan is not supported once importance is assessed out-of-sample.

The study’s central contribution is a sample-size sensitivity analysis that directly tests whether the India-Japan feature-importance difference exceeds what unequal cohort size alone predicts. The team repeatedly subsampled the Indian cohort, without replacement, down to Japan’s analytic sample size of 70, refitting the Random Forest pipeline on each of 300 draws and recording the resulting distribution of feature importances. For every feature examined, including age, sex, BMI, sleep duration, exercise frequency, water intake, and all food-preference indicators except one near-zero-importance category, Japan’s actual observed Gini importance value fell inside the 95% range produced by this subsampling procedure alone. In other words, drawing a random subsample of 70 Indian participants and fitting the identical pipeline would routinely observe importance values as different from India’s full-sample estimate as Japan’s actual values are.

The reanalysis also uncovered a measurement-instrument mislabeling that the authors disclose transparently. Perceived stress was intended to be assessed using the Perceived Stress Scale (PSS-10), a validated 10-item instrument. But on auditing the raw survey exports, no PSS-10 items exist in either dataset. The actual instrument administered at both sites was a single self-rated item asking participants to rate their stress or anxiety level on a scale of 0 to 10, which was binarized at 5 or above to define elevated stress. A single-item stress rating has not been validated for cross-cultural measurement equivalence, and the authors treat all stress results as exploratory accordingly. Similarly, once the study’s own chronic-illness exclusion criterion was correctly applied, positive diabetes cases fell to zero in both cohorts, and Japan’s survey instrument never asked a hypertension-related question at all.

The authors emphasize that both cohorts are convenience samples recruited from a single institution in each country with substantially different response patterns. India retained 54.0% of initial respondents as age-eligible and not excluded, while Japan retained only 20.2%. Any apparent difference between them may reflect institutional, disciplinary, or self-selection factors as readily as national culture. The authors also note that they did not conduct formal cross-cultural measurement-invariance testing for any instrument used in the study, meaning that interpreting comparable constructs across sites remains a working assumption rather than a demonstrated fact. Self-reported height and weight typically underestimate true BMI, introducing further measurement uncertainty.

Despite the negative findings, the authors regard the analysis as a legitimate and useful contribution in its own right. The pattern illustrated here, including small paired cohorts, no held-out evaluation, circular outcome definitions, Gini importance treated as ground truth, and an oversampling step that was described but never actually implemented, is not unique to this dataset. The checks the authors describe, including quantifying circularity by ablating the offending predictor, computing held-out permutation importance alongside Gini importance, and running a sample-size-matched subsampling sensitivity analysis, are general-purpose and inexpensive to apply before any cross-cultural machine learning feature-importance claim is published. The authors suggest that future studies routinely report an explicit list of which outcome-defining variables are excluded from the predictor set, held-out permutation importance alongside any impurity-based importance measure, and a sample-size-matched subsampling sensitivity check before interpreting any between-population difference as substantive. A genuine test of cross-cultural machine learning transferability would train a model on one cohort and evaluate its discrimination and calibration on the other, directly measuring transfer performance rather than comparing independently fitted feature importances. With complete, runnable code accompanying the paper, every number in the study can be independently verified, and the authors hope this transparency will help digital health research avoid mistaking statistical artifacts for cultural insight.

Subject of Research: Do machine learning models really encode cross-cultural lifestyle-health differences? A leakage-free reanalysis of two university cohorts in India and Japan

Article Title: Do machine learning models really encode cross-cultural lifestyle-health differences? A leakage-free reanalysis of two university cohorts in India and Japan

Article References: Takaoka, E., Ruban, S., Mohammed Siraj, B., Vinay, V., Shanthini, H., & Dias, A. (2026). Do machine learning models really encode cross-cultural lifestyle-health differences? A leakage-free reanalysis of two university cohorts in India and Japan. Frontiers in Digital Health, 8, Article 1791066. https://doi.org/10.3389/fdgth.2026.1791066

Image Credits: AI Generated

DOI: 10.3389/fdgth.2026.1791066

Keywords: machine, learning, models, really, encode, cross-cultural, lifestyle-health, differences, leakage-free, reanalysis, university, cohorts

The stakes of this audit are higher than they might appear. Ground-based ozone records are the backbone of nearly everything we know about the ozone layer: they underpin the detection of long-term trends, they measure whether the Montreal Protocol is actually healing the stratosphere, and they serve as the yardstick against which every satellite retrieval and reanalysis product is validated. If those yardsticks drift, the entire edifice of ozone recovery science wobbles with them. Yet until now, the most comprehensive performance assessment of the global network dated from 2008, before a wave of new satellite missions, new reanalysis products, and the explosive growth of the Pandonia Global Network reshaped the observing system.

The study covers six instrument types with very different personalities. The Dobson spectrophotometer, invented in the 1920s, measures ozone by comparing ultraviolet light at wavelength pairs that ozone absorbs differently, using a double monochromator to suppress stray light. The Brewer spectrophotometer, designed in the 1970s to automate the process, uses a diffraction grating to measure four wavelengths between 306 and 320 nanometers and can operate unattended in direct sun, moonlight, and zenith-sky modes. Together these are the World Meteorological Organization’s reference instruments, with accuracy better than one percent, traceable through a strict calibration hierarchy from world primary standards down to field instruments. Filter ozonometers, compact devices using fixed interference filters near 306 and 328 nanometers, offer operational simplicity at lower precision. Zenith-sky DOAS instruments, including the SAOZ systems that dominate polar monitoring, retrieve ozone from scattered twilight sunlight and carry a typical accuracy around five percent. Fourier-transform infrared spectrometers retrieve ozone from solar absorption spectra alongside a dozen other gases, and the newest network, Pandora, records direct-sun spectra from 280 to 525 nanometers with central processing through the Pandonia Global Network.

To grade these instruments fairly, the team built a harmonized statistical framework. For every station and every five-year window, they calculated the daily percentage difference between ground-based observations and comparison data from satellite overpasses and four reanalyses: ERA5, MERRA-2, JRA-3Q, and MSR2. Five metrics captured different failure modes: the mean bias, the standard deviation of daily and monthly differences, the amplitude of the seasonal cycle in the differences, and the range of annual means. A well-calibrated instrument should show a stable seasonal signature against fixed benchmarks; drifts in that signature can betray errors in extraterrestrial constants, stray light, or the effective temperature of the ozone layer itself.

The methodological innovation lies in how stations are classified. Rather than imposing one-size-fits-all limits, the team derived percentile-based thresholds from the internal statistics of each network against each benchmark. This matters because the networks differ in precision: the Brewer and Dobson networks show standard deviations of daily differences around two percent, while Filter and UVVIS instruments spread slightly wider, under three percent, especially at high latitudes. Applying Brewer-grade criteria to a UVVIS instrument would artificially condemn data that is perfectly acceptable for its class. The thresholds label each station-period as high quality, medium quality with minor issues, or not assured with major issues, and in most five-year periods fewer than twenty percent of stations worldwide fell into the not-assured category.

The headline numbers are striking. Ground-based annual means generally agree with satellite and reanalysis benchmarks within plus or minus two percent, with typical variability near two percent. The two WMO reference networks agree with each other and with the benchmarks mostly within one percent in recent decades. The world reference Dobson at Boulder and the world reference Brewer in Toronto, along with the European regional reference at Izaña, all show clean long-term records, confirming that the anchor points of the global system are holding steady. The Pandora network, despite being the youngest, shows statistical consistency comparable to the reference networks, with a standard deviation of daily mean differences of 1.83 percent, actually slightly better than Brewer’s 1.97 and Dobson’s 2.04, albeit over a shorter record.

Reanalysis datasets proved both a blessing and a caution. Because most reanalyses assimilate satellite ozone but not ground-based data, they serve as genuinely independent referees, and they allowed the assessment to extend back before the satellite era began in the 1970s. But the pre-satellite portions of ERA5 and JRA-3Q, where ozone is only indirectly constrained by meteorological observations, show biases of up to minus five percent relative to Dobson measurements. The team also caught a sudden two percent shift in MERRA-2 around 2005, when the assimilation system switched its satellite inputs from TOMS and SBUV to OMI and MLS, a reminder that reanalyses carry artifacts of their own. To handle conflicting verdicts from the four reanalyses, the researchers merged their assessments using the Dempster-Shafer evidence theory, which penalizes datasets that repeatedly disagree with their peers.

The audit also exposed the shifting geography of ozone monitoring. The Dobson network peaked at 76 active sites during 1985 to 1994 and has since declined to 51, raising concerns about long-term continuity even as the remaining stations perform superbly. The Brewer network has grown to 87 active sites, making it the largest reference network. Filter ozonometers have contracted from 52 sites to just 20, mostly in Russia. The NDACC FTIR and UVVIS networks report ozone at 22 and 17 sites respectively, while the Pandonia Global Network has surged since 2015 to 138 active sites. Notably, the study found and corrected a data-reporting error in the Filter network archive affecting observations from 2007 to 2024, and documented a confirmed hydrogen lamp failure at the Princess Elisabeth Station in Antarctica that degraded Brewer data in the 2022-23 and 2023-24 seasons before the instrument was recalibrated and returned to service in December 2024.

The practical payoff is a quality-assessed, station-level dataset with flags that data centres can use immediately. The World Ozone and Ultraviolet Radiation Data Centre and its counterparts can now steer users toward the most reliable records for climatologies, trend analyses, and satellite validation, identify sites and periods needing reprocessing, and prioritize calibration campaigns where they matter most. The audit also flags structural vulnerabilities: seventeen percent of Pandora sites are run by university researchers, whose instruments often retire with them, and vast regions including Africa and parts of South America remain thinly covered. A companion paper will turn the tables, using these quality-assured ground records to evaluate the satellites and reanalyses themselves. For a planet betting its ultraviolet protection on a handful of aging spectrophotometers and a rising generation of new instruments, knowing exactly which numbers to trust is the quiet foundation on which the ozone recovery story rests.

Subject of Research: Performance assessment of global ground-based total column ozone monitoring networks from 1940 to 2024

Article Title: Measurement report: Global Total Ozone Records – Part 1: Ground-based monitoring networks performance assessment and status review

Article References: Zhao, X., Fioletov, V., Petropavlovskikh, I., Velazco, V., Redondas, A., Solomatnikova, A., Vigouroux, C., Strong, K., Van Roozendael, M., Pazmino, A., Hanisco, T. F., Cede, A., Tiefengraber, M., Labow, G., Fujiwara, M., van der A, R., Griffin, D., McLinden, C., Kralidis, T., … Lee, S. C. (2026). Measurement report: Global Total Ozone Records – Part 1: Ground-based monitoring networks performance assessment and status review. Atmospheric Chemistry and Physics, 26(19), 13693-13720. https://doi.org/10.5194/acp-26-13693-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-13693-2026

Keywords: ozone layer, total column ozone, Dobson spectrophotometer, Brewer spectrophotometer, Pandora, satellite validation, reanalysis, Montreal Protocol, WOUDC, NDACC, atmospheric monitoring, data quality

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

Reanalyses occupy a special place in climate science. They are essentially the most honest simulations a modelling system can produce: a numerical model runs forward in time while a data assimilation scheme continuously nudges it back toward reality using observations. The result is a physically consistent, gap-free record of the past. Global products such as ERA5 for the atmosphere or GLORYS12 for the ocean do this superbly at planetary scales, but their horizontal resolution is too coarse to capture the Mediterranean’s notoriously complex geography, where steep mountains, narrow straits, and intense land-sea contrasts generate mesoscale processes that global grids simply blur out. Regional reanalyses exist for the Mediterranean, but until now they have been either ocean-only or atmosphere-only, never both at once.

MESMAR-R closes that gap by coupling three model components through the OASIS3-MCT coupler at half-hourly intervals. The atmospheric core is the Weather Research and Forecasting model running on a grid of roughly 15 kilometres with 41 vertical levels, while the ocean component uses the NEMO model at approximately 7 kilometres with 72 vertical levels spanning the entire basin from the Atlantic box to the Dardanelles. A hydrological discharge model routes surface and subsurface runoff collected by the atmospheric model directly into coastal ocean points, meaning rivers genuinely interact with the sea rather than being prescribed from climatology. Air-sea fluxes of momentum, heat, and freshwater are computed within the atmospheric model to guarantee consistency with its boundary-layer physics, then passed conservatively to the ocean, while sea surface temperature and currents flow back the other way.

The assimilation strategy is deliberately asymmetric. The ocean receives a full three-dimensional variational analysis every three days, ingesting temperature and salinity profiles from Argo floats, gliders, moorings, and expendable bathythermographs, along with along-track satellite altimetry, weak surface relaxations, and a large-scale bias correction for the deep layers. The atmosphere, by contrast, is constrained through spectral nudging toward ERA5: only spatial scales longer than roughly 850 kilometres are relaxed, with timescales of about an hour for winds and temperature and a day for humidity. This leaves the regional model free to develop its own mesoscale weather, including Mediterranean tropical-like cyclones, while remaining anchored to the large-scale circulation of the global reanalysis. The authors describe the system as weakly coupled, since observations in one medium do not instantly correct the other, but information propagates between components dynamically through the exchanged fluxes during each model integration.

The evaluation against observations is where the new product earns its credentials. Over the common comparison period of 2000 to 2020, MESMAR-R delivers the lowest root-mean-square error of any tested product for sea surface height, with typical errors below four centimetres basin-wide, and it more than halves the sea surface salinity error relative to the global GLORYS12 reanalysis. For sea surface temperature it is essentially indistinguishable from the dedicated regional ocean reanalysis MEDREA24, and both substantially outperform the global product. In the upper ocean, the layer that matters most for air-sea interaction, MESMAR-R shows the lowest temperature errors in the 20 to 60 metre range and leads or ties for best salinity performance in the top 210 metres. Bias profiles reveal only a small fresh bias near the surface and near-zero thermal drift below 200 metres, indicating that the system stays honest at depth.

The atmospheric component, handicapped by its simplified nudging-based assimilation, naturally trails the dedicated ERA5 and CERRA reanalyses in area-averaged skill. Root-mean-square errors for two-metre temperature, relative humidity, wind speed, and pressure are moderately higher, and the gap is largest for precipitation over land. But the comparison maps tell a subtler story: MESMAR-R achieves localized improvements near complex coastlines and orography, precisely where the 15-kilometre model resolves gradients that coarser reference datasets smooth away. Crucially, the atmospheric fields remain internally consistent with the ocean beneath them, which is the entire point of a coupled product. For studying compound events, where a marine heatwave coincides with an atmospheric heat dome, that consistency matters more than shaving a few hundredths of a degree off an error metric.

Sensitivity experiments disentangle the contributions of coupling and data assimilation, and they deliver one of the study’s most instructive findings: coupling alone does not improve a reanalysis. A fully coupled simulation with no data assimilation showed the largest error growth, particularly in the subsurface ocean, while an uncoupled ocean reanalysis forced by ERA5 slightly outperformed the full coupled system for most parameters. Yet coupling measurably improved sea surface salinity, especially in the Aegean Sea, where the interactive exchange of freshwater fluxes enhances realistic evaporation and cools and salinifies the surface in ways prescribed forcing cannot replicate. Ocean data assimilation, meanwhile, proved to be the dominant control on basin-scale heat content, with the reanalysis reproducing the observed warming rate of roughly 1.3 watts per square metre in the upper 700 metres, against an observational estimate of about 1.4 watts per square metre, and correlations with observed heat content reaching 0.93 to 0.94.

The long-term climate trends extracted from the reconstruction read like a diagnosis of a basin under stress. Total Mediterranean sea level rose at approximately 3.06 millimetres per year over 1993 to 2024, in excellent agreement with independent estimates, and a striking 94 percent of that rise is mass-driven, reflecting the basin’s response to Atlantic sea-level rise through the Strait of Gibraltar rather than local thermal expansion. Beneath that total, a powerful thermo-halo compensation plays out: the halosteric component of sea level falls at nearly 3.9 millimetres per year basin-wide as the sea salinifies, almost exactly cancelling the thermosteric rise from warming. The eastern basin warms fastest in heat content, consistent with the documented abrupt warming and salinification of Levantine Intermediate Water since the early 2000s, while the western basin shows the strongest salinity-driven density increase, a hotspot centred on the Alboran Sea and the western gyre.

The atmospheric trends confirm the region’s reputation as a climate change amplifier. Winter warming averages about 0.040 degrees Celsius per year across the Mediterranean domain, accelerating to roughly 0.047 degrees per year in summer, with the strongest signals over southern Europe. Precipitable water increases almost everywhere, by about 0.023 percent per year in winter and a remarkable 0.100 percent per year in summer, yet relative humidity near the surface declines across large parts of Europe and mean precipitation does not follow the moisture upward. The picture is a warming, drying regime in which the atmosphere holds more water but delivers it in fewer, more intense bursts, favouring longer dry spells punctuated by extreme events. Winter wind trends even leave fingerprints in the ocean, with localized strengthening over the Gulf of Lion deepening the mixed layer while the southeastern basin shoals as buoyancy loss weakens.

The authors are candid about limitations. Observational coverage is not homogeneous across the 32 years, and because the atmospheric component is nudged toward ERA5, it partly inherits the time-varying information content of the global observing system, so decadal trends should be interpreted with that caveat in mind. Mid-depth temperature skill trails MEDREA24, and localized degradations near the northern Adriatic and the Dardanelles point to riverine and boundary-condition challenges. Future versions may adopt observation-space atmospheric assimilation and fully balanced air-sea corrections. Even so, MESMAR-R stands as one of the first regional coupled reanalyses ever produced, and its balanced, physically coherent reconstruction of the Mediterranean’s recent past arrives at a moment when the region needs it most, offering researchers, forecasters, and even machine-learning emulators a single dataset in which the sea and the sky finally tell the same story.

Subject of Research: A coupled regional atmosphere-ocean-hydrology reanalysis of the Mediterranean Sea covering 1993-2024, evaluated against observations and used to quantify regional climate trends

Article Title: Evaluation of a coupled regional reanalysis for the Mediterranean region covering the period 1993–2024

Article References: Storto, A., de Toma, V., & Yang, C. (2026). Evaluation of a coupled regional reanalysis for the Mediterranean region covering the period 1993–2024. Ocean Science, 22(5), 2809-2834. https://doi.org/10.5194/os-22-2809-2026

Image Credits: AI Generated

DOI: 10.5194/os-22-2809-2026

Keywords: Mediterranean Sea, reanalysis, data assimilation, coupled modelling, climate change, sea level rise, ocean heat content, salinification, air-sea interactions, NEMO, WRF, marine heatwaves

The research team confronted a long-standing puzzle. Central Europe sits in a climatological transition zone, wedged between the drying Mediterranean and the wetting Nordic countries, and different studies have reached contradictory conclusions about whether the region is becoming wetter or drier. Part of the disagreement stems from the metrics used. The researchers tackled this by employing two complementary drought indices computed over three-month accumulation windows: the Standardized Precipitation Index, which considers rainfall alone, and the Standardized Precipitation Evapotranspiration Index, which also accounts for atmospheric evaporative demand, the combined thirst of the atmosphere for water from soil and vegetation. The gap between these two measures turned out to be the key to the entire story.

To extend the record back to the pre-industrial era, the team harnessed three independent reanalysis datasets, each blending numerical weather models with historical observations. ERA5, the most modern and high-resolution product, assimilates satellites, aircraft, ships and station data but only reaches back to 1940. The Twentieth Century Reanalysis version 3 relies primarily on surface pressure observations and offers temporal consistency from 1836 onward, while the Modern Era Reanalysis incorporates proxy records, documentary sources and early instrumental measurements. Before trusting any of them, the researchers rigorously evaluated each dataset against century-long weather station records from De Bilt, Frankfurt am Main, Paris-Le Bourget and Uccle, using moving-window error metrics and correlation coefficients to identify the periods where each reanalysis could be considered reliable.

The evaluation revealed telling weaknesses. ERA5 showed spurious dryness in its earliest years, likely because upper-air temperature observations were scarce before 1946, while the Twentieth Century Reanalysis displayed excessive wetness before the late 1910s, when the number of assimilated Northern Hemisphere observations jumped sharply. ModE-RA proved dependable from 1844 onward. The final analysis therefore combined reliable windows from each dataset: 1947 to 2023 for ERA5, 1918 to 2015 for the Twentieth Century Reanalysis, and 1844 to 2008 for ModE-RA. Encouragingly, the datasets agreed well with one another, with correlation coefficients for drought-event identification reaching 0.79 between ERA5 and ModE-RA.

The historical perspective delivered some surprising verdicts. The most severe drought in the 180-year record was not the celebrated disaster of 2018 but the great drought of 1921, which persisted for an extraordinary fourteen months in the ModE-RA reconstruction, producing a severity more than double the threshold defining the worst ten percent of events. Comparable clusters of successive dry years to the 2011-2023 period occurred during 1857-1874 and 1947-1960, separated by conspicuously wet intervals in 1912-1920 and 1977-1988. This pronounced multidecadal oscillation means that any assessment of recent trends must reckon with the phase of natural variability, a caution the authors emphasize throughout.

Yet one finding stood out with unsettling clarity. When evaporative demand was included, the 2010s emerged as the driest decade since the pre-industrial era, with 35 drought months in the ERA5 record, nearly a third of the decade, and western Germany locally exceeding 40 drought months. When only precipitation was considered, that same decade looked entirely unremarkable, with just 21 drought months. This divergence is the fingerprint of warming. Since the 1990s, western Central Europe has warmed at 0.42 degrees Celsius per decade, faster than the global average, and the resulting rise in atmospheric evaporative demand has progressively offset the region’s increasing precipitation. In practical terms, droughts that once required a rainfall deficit now can be manufactured, at least in part, by heat alone.

The seasonal anatomy of this shift is equally revealing. Winter and autumn have become wetter, as robust precipitation increases outpace the modest warming-driven rise in evaporative demand during the energy-limited cold season. Summer, by contrast, shows the strongest increase in evaporative demand, which largely cancels out rainfall gains and produces drying in the water-balance index. Most strikingly, spring, historically an intermediate season, has recently tipped toward drying as well, with precipitation rising far below the roughly seven percent per degree expected from basic thermodynamic scaling while evaporative demand climbed more than twice as fast. Spring drying, the analysis shows, is a distinctly recent development, absent from the longer historical record.

To explain these patterns dynamically, the researchers turned to the atmosphere itself. They applied a k-means clustering algorithm to seasonal anomalies of 500 hectopascal geopotential height, the height of the pressure surface roughly five to six kilometers aloft, for every drought event across the Euro-Atlantic sector. Four recurrent circulation patterns emerged: a Baltic High, a British Isles High, a North-South Dipole combining high pressure over Northern Europe with low pressure to the south, and a European High centered directly over the study region. Each carries a distinct hydroclimatic signature. The Baltic High draws warm continental air from the southeast, inflating evaporative demand, while the British Isles High brings cool northerly flow and its droughts are driven almost purely by rainfall deficits. The North-South Dipole dominates winter droughts and, linked to cold easterly air masses, produces the weakest events.

The European High is the villain of the modern chapter. Droughts associated with this pattern, which combines strong anticyclonic subsidence, reduced cloud cover and the largest positive evaporative demand anomalies, are the most intense in the record, particularly over western Belgium and northern France, and they occur predominantly in spring. Crucially, the fraction of droughts tied to the European High has been rising across all reanalyses and both indices, with moderate to high statistical confidence, while associations with the British Isles High and the North-South Dipole have declined since 1947. This dynamical shift dovetails neatly with the observed hydroclimatic trends: the fading winter dipole aligns with winter wetting under increasingly positive North Atlantic Oscillation phases, and the ascending European High provides a coherent explanation for the emergence of spring drying.

The broader implication is that western Central Europe’s droughts are entering a new regime. Severity alone no longer tells the full story, because the 1921 benchmark can still be matched or exceeded by persistence, while the character of droughts, increasingly hot, evaporative and spring-rooted, has shifted beneath the apparent stability of the long-term water balance. The authors note that compound hot and dry conditions, which inflict impacts exceeding either hazard alone, are becoming the regional norm, and that multi-year droughts of the kind seen in 2018 are projected to intensify further as evaporative demand continues to climb. For the roughly 100 million people who live in this economic heartland of Europe, and for the croplands, grasslands and forests it supports, the message of the past 180 years is clear: the droughts of the future will not simply be repeats of the past, and monitoring systems that ignore the atmosphere’s growing thirst risk underestimating the threat.

Subject of Research: Long-term drought variability and associated atmospheric circulation patterns in western Central Europe since 1844

Article Title: Droughts in western Central Europe and associated atmospheric circulation patterns since 1844

Article References: Neimry, E., Goosse, H., & Jonard, M. (2026). Droughts in western Central Europe and associated atmospheric circulation patterns since 1844. Weather and Climate Dynamics, 7(3), 1759-1778. https://doi.org/10.5194/wcd-7-1759-2026

Image Credits: AI Generated

DOI: 10.5194/wcd-7-1759-2026

Keywords: drought, western Central Europe, atmospheric circulation, evaporative demand, climate change, reanalysis, SPEI, anticyclone, European High, multidecadal variability, spring drying, hydroclimate

The heart of the problem lies in what oceanographers call reanalysis products. These are computer reconstructions of the ocean’s past state, blending models with satellite and in situ observations, and they are the standard physical backdrop against which biogeochemical models are run and calibrated. Yet current global reanalyses struggle to resolve ocean dynamics at horizontal scales below roughly one hundred kilometres, and at the sea surface they typically resolve features only down to scales of one to two degrees. Those unresolved mesoscale and submesoscale motions, eddies and fronts a few kilometres to a few tens of kilometres across, exert a disproportionate influence on nutrient supply and plankton blooms. When a biogeochemical model is calibrated against data generated with slightly misplaced physics, the resulting parameter estimates can be badly biased, distorting simulated bloom timing and amplitude even under relatively benign conditions.

To attack the problem in a controlled setting, the researchers built an Observing System Simulation Experiment, a kind of numerical testbed in which the true answer is known by construction. They implemented a one-dimensional version of the classic NNPZD model, which tracks nitrate, ammonium, phytoplankton, zooplankton and detritus through a vertical water column extending to 350 metres, driven by solar irradiance at the surface and by time-varying vertical mixing coefficients. The physical forcings were drawn from a high-resolution, two-kilometre simulation of the North Atlantic subpolar gyre centred on the Porcupine Abyssal Plain, a well-studied open-ocean site for carbon export research. Because the framework was written using differentiable programming, gradients could flow through the entire model, enabling both variational data assimilation and neural network training on the same footing.

The team generated 3,100 one-year simulations by sampling the model’s sixteen biogeochemical parameters within plus or minus twenty percent of their reference values, of which 2,000 served as training data for a neural network, 1,000 for validation and 100 for testing. Crucially, they contaminated the physical forcings with realistic errors: a random, time-evolving spatial shift of the sampling location, mimicking the way reanalyses misplace mesoscale features, at three uncertainty levels ranging from about 0.3 to 1 degree of horizontal displacement. They also simulated four observing configurations, from an idealised scenario in which all five state variables are measured at every depth, down to sparse CTD-only sampling that leaves ammonium and zooplankton entirely unobserved, with sampling intervals ranging from daily to every sixty days.

Against this benchmark, the authors compared three approaches. The first was a weakly constrained four-dimensional variational data assimilation scheme, the workhorse of operational systems, which adjusts model parameters and initial states by minimizing the mismatch between simulated and observed concentrations. The second was an end-to-end deep learning scheme built on a UNet architecture, a multi-scale convolutional network originally developed for image segmentation, here trained to map noisy forcings and sparse observations directly to corrected forcings, reconstructed state variables and estimated parameters. The third was a hybrid that feeds the neural network’s outputs into the variational scheme as a refined first guess. In error-free physics, the variational method remains the sharper tool, but the moment forcing uncertainty enters the picture, the ranking reverses dramatically.

The numbers are striking. Under the reference scenario with low-level forcing uncertainty and a ten-day sampling rate, the neural scheme achieved a mean correlation of 0.99 between reconstructed and true nitrogen stocks, against 0.88 for the variational method, whose worst case fell to minus 0.49. Seasonal blooms reconstructed with variational estimates could be shifted by up to twenty-eight days, while the neural scheme kept shifts within about ten days. Parameter estimation errors spanned a range of roughly 0.78 for the variational approach but only 0.38 for the neural one, with the standard deviation of the error reduced by a factor of 2.8. Perhaps most tellingly, the neural scheme’s performance barely degraded as forcing uncertainty tripled, whereas the variational calibration, which implicitly assumes error-free physics, could overestimate or underestimate bloom amplitudes by factors exceeding 1.5.

The sensitivity analyses carry practical weight for anyone designing ocean observing campaigns. Sampling frequency proved decisive: at one-day and ten-day intervals the neural calibration reached mean correlations of 0.994 and 0.987, but performance dropped to 0.97 at thirty days and 0.957 at sixty days, with parameter uncertainty roughly tripling. The reason is temporal: the simulated blooms are single-peak events lasting a few weeks, so monthly or two-monthly sampling simply misses their duration and amplitude. Vertical structure mattered too. Dropping ammonium and zooplankton observations, as a CTD-only strategy would, degraded parameter estimates by factors of 1.4 to 3.5 compared with full-depth sampling, confirming earlier findings that zooplankton data, hard to collect as they are, are disproportionately valuable for constraining ecosystem models.

The hybrid scheme then delivered the study’s most elegant result. When the neural network’s corrected forcings, estimated states and parameters were used to initialize and drive the variational assimilation, reconstruction errors for phytoplankton and zooplankton concentrations were halved relative to either method alone, with normalized mean squared errors falling from 0.085 to 0.047 and from 0.017 to 0.008 respectively. The hybrid also produced the most honest uncertainty estimates: its ensemble spread encompassed 80.6 percent of true values, compared with 42.1 percent for the variational scheme and 65.8 percent for the neural scheme alone. In other words, the physics-informed refinement not only sharpened the answer but made the accompanying confidence intervals trustworthy, a property that matters enormously for operational monitoring and carbon accounting.

The authors are careful about the limits. The neural network is scenario-dependent: it must be retrained for each observing configuration and uncertainty level, and its skill is bounded by the distribution of errors represented in its training data. The testbed also assumes that the model equations are correct and that state variables are directly observable, whereas in reality phytoplankton is inferred from chlorophyll fluorescence, zooplankton from nets and cameras that capture only part of the community, and model structural errors can dwarf calibration errors. Scaling to three-dimensional global models poses a computational challenge, since generating thousands of training simulations with a system like NEMO-PISCES would demand computing resources comparable to century-scale ensemble runs. Neural emulators of ocean biogeochemistry, now emerging for sea surface chlorophyll and global circulation, may dissolve that bottleneck. For now, the study stands as a compelling demonstration that supervised learning, trained on a few thousand synthetic years of ocean life, can outperform classical assimilation precisely where the real ocean is messiest, and that the best answer comes from teaching the neural network and the physics to work together.

Subject of Research: Machine learning-based calibration and reanalysis of a one-dimensional ocean biogeochemical model under physical forcing uncertainty

Article Title: Solving calibration and reanalysis challenges of ocean biogeochemical dynamics with neural schemes: a 1D vertical model case-study

Article References: Littaye, J., Memery, L., & Fablet, R. (2026). Solving calibration and reanalysis challenges of ocean biogeochemical dynamics with neural schemes: a 1D vertical model case-study. Biogeosciences, 23(19), 6879-6914. https://doi.org/10.5194/bg-23-6879-2026

Image Credits: AI Generated

DOI: 10.5194/bg-23-6879-2026

Keywords: ocean biogeochemistry, deep learning, data assimilation, carbon cycle, NNPZD model, UNet, reanalysis, phytoplankton bloom, observing system simulation experiment, parameter calibration, North Atlantic, Biogeosciences

A new study published in the journal Weather and Climate Dynamics by Ninghui Li of Tsinghua University and colleagues now offers a sharper lens on this problem. The team analyzed twenty-one years of warm-season data, from 2000 to 2020, drawn from three of the most modern atmospheric reanalysis products available: the European Centre for Medium-Range Weather Forecasts’ ERA5, NASA’s MERRA-2, and the Japan Meteorological Agency’s JRA-3Q. By combining a sophisticated vortex-tracking algorithm with a purpose-built diagnostic based on potential vorticity tendencies, they have produced one of the most detailed mechanistic accounts to date of how the anticyclone is born, moves, strengthens, and decays.

The first headline finding concerns structure. For years, scientists debated whether the anticyclone was fundamentally bimodal, alternating between centers over the Iranian Plateau and the Tibetan Plateau, or whether it simply drifted between them. The new analysis, performed on the 370 Kelvin isentropic surface using the Montgomery stream function, a dynamically consistent measure of flow on surfaces of constant potential temperature, confirms that the system is robustly trimodal. Three preferred centers emerge: one near 50 degrees east over the Iranian Plateau, one near 90 degrees east over the Tibetan Plateau, and a third over the Western Pacific near 135 to 150 degrees east, corresponding to the well-known Bonin High. All three reanalysis products agree on this threefold structure, and the agreement is particularly strong between JRA-3Q and ERA5.

The tracking algorithm itself is elegant in its simplicity. The researchers first locate the zero-zonal-wind ridgeline, the line separating easterly tropical flow from westerly subtropical flow, and then search for local maxima in the Montgomery stream function along that line. Each candidate center must pass strict filters: the relative vorticity at the center and its neighbors must be anticyclonic, the winds poleward must be westerly while those equatorward must be easterly, and the ridgeline must be long enough to exclude small-scale noise. Vortices that persist for more than eighteen hours and move less than ten degrees between successive six-hour snapshots are classified as persistent; the rest are transient. Persistent vortices accounted for more than 65 percent of all identified centers in ERA5 and JRA-3Q, underscoring how stable these features can be.

The seasonal choreography of the three modes is striking. Tibetan Plateau centers dominate in May, Iranian Plateau centers take over from mid-June to mid-July, and the balance shifts back toward the Tibetan Plateau as the Bonin High emerges in August and September, coinciding with the maturation of the Western North Pacific monsoon. The latitude of the vortices also migrates, drifting northward to about 40 degrees north between mid-July and mid-August before retreating southward. Many features show evidence of westward propagation, though eastward motion appears early and late in the season, and time-longitude diagrams reveal tails of low potential vorticity stretching in both directions, the telltale signature of eddy shedding, in which subsidiary anticyclones break away from the main circulation.

The heart of the study, however, lies in its diagnostic innovation. Potential vorticity, a quantity that combines rotation and stratification, is conserved in adiabatic, frictionless flow, making it a powerful tracer of atmospheric motion. But instantaneous potential vorticity fields cannot distinguish between a vortex that is moving and one that is intensifying in place. To resolve this ambiguity, the team developed a decomposition of the potential vorticity tendency, separating dynamical contributions from horizontal advection of potential vorticity by the winds from thermodynamic contributions arising from diabatic heating, both latent heating from condensation in convective clouds and radiative heating from sunlight and infrared exchange.

The results reveal a characteristic tripole pattern in the potential vorticity tendency around each anticyclonic vortex. An advective dipole, with positive tendencies to the west of the center and negative tendencies to the east, governs the eastward translation of the vortex: positive tendencies erode the low-potential-vorticity anomaly on the western flank while negative tendencies reinforce it on the eastern flank, nudging the entire core eastward. Superimposed on this dipole is a central monopole, driven mainly by diabatic heating and its vertical gradient, which modulates the vortex’s intensity without displacing it. A negative central tendency strengthens the anticyclone; a positive one weakens it. This separation of propagation from intensification is precisely what earlier diagnostics could not achieve.

The heating story is nuanced. Deep convective latent heating, concentrated below and to the east of the vortex centers, generates negative potential vorticity tendencies that intensify the anticyclone, particularly for the Tibetan Plateau mode, where heating along the southern slopes of the Himalayas is strongest. Radiative heating, by contrast, generally acts to spin the anticyclone down: because radiative heating increases with height through the upper troposphere and lower stratosphere, its vertical divergence produces positive potential vorticity tendencies that erode the low-potential-vorticity core. Latent heating partially offsets this radiative damping near the vortex centers. Meanwhile, the composite analysis shows that diabatic heating exerts a southward pull on the vortices, tugging them back toward the convective core of the monsoon and inhibiting their escape from it.

When it comes to propagation, the winds rule. Mean zonal advection by the subtropical westerly jet pushes all three vortex modes eastward relative to the surface, but meridional advection, the dominant term in Rossby wave propagation, pushes them westward relative to the background flow, and the two effects largely offset each other in the net. The balance of terms differs across the reanalysis products, with MERRA-2 showing stronger radiative damping and weaker advective tendencies than ERA5 or JRA-3Q, differences the authors attribute in part to how each system represents anvil cloud radiative effects and prognostic cloud ice. These inter-product discrepancies are a sobering reminder that mechanistic attribution in the upper troposphere and lower stratosphere still carries real uncertainty.

The implications extend well beyond academic bookkeeping. Because the anticyclone controls the leakage of Asian monsoon pollution and moisture into the global stratosphere, a reliable framework for separating its propagation from its intensification could improve predictions of stratospheric composition and, ultimately, of monsoon-related climate variability itself. The authors suggest that their potential vorticity tendency diagnostic could help clarify the differences between westward- and eastward-shedding eddies, processes first linked to baroclinic instability more than two decades ago but never fully quantified. Future work, they note, will need objective criteria for identifying eddy shedding events and a fuller accounting of data assimilation effects and parameterized momentum sources. For now, the study establishes a unified theoretical framework for one of the atmosphere’s most important summer circulations, showing that winds set the anticyclone’s course while heating sets its strength, a division of labor now made visible for the first time.

Subject of Research: Dynamics and potential vorticity tendency diagnostics of the Asian Summer Monsoon Anticyclone in the upper troposphere and lower stratosphere

Article Title: Dynamics of the Asian Summer Monsoon Anticyclone: insights from potential vorticity tendency diagnostics

Article References: Li, N., Wright, J. S., Rupp, P., Ming, A., Zhang, S., & Gao, J. (2026). Dynamics of the Asian Summer Monsoon Anticyclone: insights from potential vorticity tendency diagnostics. Weather and Climate Dynamics, 7(3), 1853-1874. https://doi.org/10.5194/wcd-7-1853-2026

Image Credits: AI Generated

DOI: 10.5194/wcd-7-1853-2026

Keywords: Asian Summer Monsoon Anticyclone, potential vorticity, upper troposphere, lower stratosphere, diabatic heating, vortex tracking, reanalysis, Bonin High, Tibetan Plateau, eddy shedding, monsoon circulation, atmospheric dynamics

The team, led by Rhidian Thomas of the National Centre for Atmospheric Science and the University of Reading, together with colleagues from the University of Edinburgh, the University of Reading, and NOAA’s Physical Sciences Laboratory, used a sophisticated technique built on the NOAA-CIRES-DOE Twentieth Century Reanalysis version 3, known as 20CRv3. A reanalysis system combines a numerical weather model with historical observations to reconstruct the atmospheric state of the past. In this case, the system assimilated more than a century-old archive of surface pressure readings from land stations, ships, and tropical cyclone reports, using an Ensemble Kalman Filter to blend those observations with the NOAA GFS atmospheric model. The result is an 80-member ensemble of three-hourly atmospheric fields covering all of 1903, a year that ranks as the third coldest in the 175-year HadCRUT5 temperature record, partly because of the cooling influence of the Santa Maria volcanic eruption of 1902.

The crucial innovation lies in what the researchers did next. They re-ran the entire data assimilation system, feeding it exactly the same pressure observations but with two deliberate changes: sea surface temperatures were uniformly raised by 2 degrees Celsius, and in a second, novel experiment, atmospheric carbon dioxide was lifted to 530 parts per million on top of the ocean warming. Because the same pressure data constrain the atmospheric flow in every experiment, all three simulations experience essentially identical weather patterns, the same storms tracking the same paths, the same anticyclones parking over the same regions. Any difference between the runs therefore reflects the pure thermodynamic response to a warmer, moister atmosphere, isolated from the uncertain changes in large-scale circulation that dominate disagreements between climate models.

This storyline approach, related to pseudo-global-warming and nudging methods used in event attribution, has a powerful advantage: it sidesteps the largest source of uncertainty in regional climate projections. Regional forecasts of temperature and rainfall depend heavily on how storm tracks, jet streams, and blocking patterns might shift, questions on which models diverge. By anchoring the analysis in weather patterns that demonstrably occurred, the reanalysis experiments answer a cleaner question: if the atmosphere’s circulation stayed the same, how would the same weather feel in a warmer world? The team focused on four regions where 1903’s pressure observations were dense enough to tightly constrain the circulation: northwestern Europe, the western Mediterranean, the East Coast of the United States, and southeastern Australia.

The temperature results are unambiguous. In the warmer experiments, 99 percent of days became warmer than in the original reconstruction, with typical days over land warming by roughly 2 to 2.5 degrees Celsius for a 2-degree ocean warming. But the warming was far from uniform across the temperature distribution. The largest increases, 4 to 5 degrees Celsius, fell on the coldest days, those dipping below freezing. The mechanism is a classic snow-albedo feedback: in the warmer simulations, less of the land surface is covered in snow, so proportionally more sunlight is absorbed by darker ground rather than reflected back to space, amplifying warming precisely on the days when snow would otherwise lie. Southeastern Australia, where freezing days are rare, was the lone region without this cold-day amplification.

Hot days warmed disproportionately too, with increases of 3 degrees Celsius or more on days above 30 degrees in the western Mediterranean, the US East Coast, and southeastern Australia. The underlying physics varies by region. On the US East Coast, the hottest days coincide with dry soils, and the simulations show a sharp reduction in evaporative cooling above the 97th temperature percentile, meaning the land surface loses its ability to shed heat through evaporation, a well-known mechanism that intensifies heatwaves. In southeastern Australia, by contrast, the amplification appears tied to reduced cloud cover and greater incoming solar radiation. Notably, the direct radiative effect of the added carbon dioxide contributed an extra 0.2 to 0.4 degrees of warming on typical days, and as much as a fifth of the total warming over the arid western United States, demonstrating that carbon dioxide warms the surface directly, not merely indirectly through ocean warming.

Rainfall told an equally counterintuitive story. Globally, total precipitation rose in the warmer experiments, but far more slowly than atmospheric moisture, which climbed at roughly 8.5 percent per degree of warming, in line with the Clausius-Clapeyron relation governing how much water vapor air can hold. The reason is the atmosphere’s energy budget: latent heating from rain must balance longwave cooling, and additional carbon dioxide suppresses longwave emission to space, blunting the precipitation increase. Regionally, the pattern echoed canonical model projections, with wetting across much of the extratropics and drying over the Mediterranean and subtropical oceans. Even in the western Mediterranean, where annual rainfall declined, daily rainfall variability increased, meaning larger swings from wet day to dry day, a signature consistent with emerging observational trends.

The most vivid finding concerns how the extra rain is delivered. Across all four regions, the number of days with light, drizzly rain decreased, while days with moderate to heavy rain grew exponentially more frequent as rain rates increased. The cutoff scale separating ordinary from heavy events shifted upward roughly in step with Clausius-Clapeyron scaling, yet the time-mean increase in rainfall was concentrated on fewer than 1 day in 10. In other words, even where total annual rainfall rises, most wet days actually become drier, and the annual surplus arrives in a handful of intense bursts. Dry days also increased in three of the four regions, likely because reduced relative humidity over land raises convective inhibition, lengthening the gaps between rain events.

October 1903 provides the study’s showpiece. That month remains the wettest calendar month on record for the United Kingdom, with 220 millimeters in the national series back to 1836. The reanalysis reproduced the observed geography of the deluge, and in the warmer experiments, rainfall over the UK and Ireland rose by 12 to 17 percent, equivalent to 6 to 8.8 percent per degree of local warming, well above the annual-mean scaling. On sub-daily timescales the amplification was even more dramatic: one intense three-hour event on 6 October intensified by 31 percent, around 15 percent per degree, exceeding Clausius-Clapeyron scaling as previous studies have found for short-duration extremes. Scaling to today’s roughly 1.5 degrees of warming since 1903, the authors estimate that the same weather pattern would now deliver around 240 millimeters across the UK, comfortably surpassing the previous record of 216.3 millimeters set in December 2015.

The study’s limitations are candidly acknowledged. The experiments are idealised, with uniform ocean warming, unchanged sea ice, and a single year of weather, and the assumptions that 1903’s patterns could occur in a warmer world, and that the assimilation remains consistent under perturbed boundary conditions, are supported but not proven; the warmer runs did reject somewhat more pressure observations, particularly over densely observed northwestern Europe. Circulation changes, if they occur, could amplify or modify these thermodynamic signals, and long-standing model biases in light rainfall temper confidence in the drizzle decline. Yet the fact that a single pair of counterfactual reanalysis runs recovers so many robust patterns seen across generations of climate models, amplified warming of hot and cold extremes, rising precipitation variability, fewer rainy days, and heavier downpours, makes this a powerful, computationally efficient new line of evidence. It suggests that the everyday weather of the past is already a preview of a different daily experience in our warming future, one written into every frost-free night and every intense burst of rain.

Subject of Research: Thermodynamic impacts of climate change on everyday temperature and precipitation, studied by reanalysis experiments translating the weather of 1903 into a warmer world

Article Title: Everyday weather in a warmer world

Article References: Thomas, R., Compo, G. P., George, S., Hegerl, G. C., Schurer, A., Shepherd, T. G., Slivinski, L. C., Thompson, V., & Hawkins, E. (2026). Everyday weather in a warmer world. Earth System Dynamics, 17(5), 1365-1380. https://doi.org/10.5194/esd-17-1365-2026

Image Credits: AI Generated

DOI: 10.5194/esd-17-1365-2026

Keywords: climate change, reanalysis, 20CRv3, extreme weather, precipitation, temperature extremes, Clausius-Clapeyron, storyline attribution, 1903, hydroclimate, snow albedo feedback, Mediterranean drying