The study combines hourly and daily observations from Météo-France and the French electricity company EDF across a combined network of 163 stations. The records span the northern and southern French Alpine provinces, where complex topography intersects competing Atlantic and Mediterranean climatic influences. For the climatology, the researchers used recent observations from 1980 through 2022 or 2023. The trend analysis relied on EDF records extending as far back as 1952, including 28 stations with more than 45 years of data and a longest series of 71 years. The long-term network was concentrated in mountain areas, with few EDF stations in the Rhône valley and along the Mediterranean coast, an important limitation when comparing trend patterns with the broader stationary climatology.

Quality control was particularly important because the older EDF material contains only weekly maximum precipitation totals for fixed durations of 1, 2, 3, 6, 12 and 24 hours. For recent hourly series, the researchers compared sums of 24 hourly readings with quality-controlled daily totals and scaled each day’s hourly values when necessary to reproduce the recorded daily precipitation. Days with missing daily observations or discrepancies exceeding 25 percent were invalidated, and visual inspection of seasonal maxima led to the removal of one station with suspicious winter values. Instrumentation and data-processing methods changed during the historical record, including a transition from mechanical to digital registration, so the authors assumed that individual time series were broadly homogeneous while treating known irregularities at the earliest stations as a limitation.

For each station, the team applied extreme value analysis to monthly, seasonal and annual precipitation maxima using a generalized extreme value distribution. A 20-year return level represents rainfall with a 1-in-20 annual chance of being exceeded under the fitted model, making it a measure of a rare, potentially high-impact event rather than the most intense value directly observed. The non-stationary model keeps distribution parameters unchanged before 1985 and allows them to evolve linearly afterward. Separate models permitted trends in the location parameter, the scale parameter or both, with the best-fitting form selected by the Akaike information criterion. A bootstrap procedure resampled each maximum series 500 times and assessed whether the 5 to 95 percent confidence interval for the estimated change included zero.

Because estimating trends from records of only a few decades can be unstable, the model borrowed observations from nearby stations. For each target station, annual maxima from surrounding sites were normalized by their respective means, combined and rescaled to the target station. The neighborhood radius was 0.3 degrees, adding an average of 3.9 neighboring stations, although sites near the regional margins had fewer neighbors. Tests using radii of 0.2 and 0.4 degrees produced broadly similar results. The technique makes regional patterns more robust but assumes that nearby stations share a statistically homogeneous precipitation distribution, an assumption that may fail across steep terrain or distinct weather regimes. The approach also does not explicitly account for topography.

The baseline climatology reveals sharply different seasonal signatures for daily and hourly rainfall. Daily 20-year return levels were generally below 70 millimeters per day from February through August and rose above 200 millimeters per day in parts of Alpes-Maritimes from September through January. The Rhône valley and Mediterranean coast also stood out during the wetter autumn months. Hourly return levels followed a different annual cycle: they were lowest in February, generally below 15 millimeters per hour, and highest in late summer. Estimates reached about 75 millimeters per hour in western Alpes-de-Haute-Provence in July. The warm-season peak is consistent with localized thunderstorms and greater atmospheric moisture and instability, while the dry Mediterranean summer produces exceptionally low totals in the south. Hourly extremes were also higher along southern and western slopes and in the pre-Alps than in the drier inner Alpine zone.

Large-scale circulation helps explain the contrast between daily and hourly events. Atlantic flows dominated northern daily rainfall maxima throughout the year. In the south, Mediterranean and Atlantic influences alternated, with Mediterranean flows especially important from September through November and during the wider seasonal analysis. Daily annual maxima closely resembled autumn maxima, consistent with the importance of Mediterranean storms. Hourly events showed a less robust relationship to the daily weather-pattern classification, reflecting their more localized formation, although anticyclonic conditions became more influential in the south from May through August. The study used an existing classification of eight large-scale weather patterns and grouped three Atlantic and three Mediterranean flow types, together with northeast circulation and anticyclonic conditions.

Daily trends were generally between minus 10 and plus 10 millimeters per day per decade, but their seasonal and regional structure was coherent. The northern Alps showed significant increases in May and decreases in June and September. In the south, daily extremes decreased in August and September but increased from October through December. Seasonal trends were negative in winter and summer and positive in spring and autumn. Annual return levels in the south rose by as much as 24 millimeters per day per decade, equivalent to an estimated increase of up to 50 percent from 1990 to 2020, whereas northern annual trends reached minus 8 millimeters per day per decade, or about minus 20 percent over the same comparison. The strongest southern increases coincided with stronger Mediterranean circulation, while the opposite northern pattern was consistent with changes in Atlantic influence, although the observational study alone cannot establish causation.

Hourly trends were less consistent and statistically significant at fewer stations. Northern estimates were usually centered near zero, apart from relatively small negative September trends. In the south, trends leaned positive from April through June and from October through December, with slight negatives in August and September. November was the clearest exception: a spatially coherent positive pattern reached 3.5 millimeters per hour per decade in the south, corresponding to an estimated increase of up to 100 percent between 1990 and 2020. Annual hourly trends were also centered near zero in the north and only slightly positive in the south, with significant estimates at just two southern stations having opposite signs. A possible northeast-southwest gradient appeared in summer and annual results, with stronger values in the high Alps than in lower southwestern terrain, but the researchers did not present it as a definitive regional pattern.

The physical causes of the hourly pattern remain hypotheses rather than demonstrated mechanisms. The authors suggest that warming may increase the energy available for convection in high mountain areas, potentially aided by shorter snow persistence and associated albedo changes, while increasing aridity may reduce moisture for local recycling in lower areas. Seasonal shifts may also matter: the widespread decline in September daily extremes may indicate a later onset of autumn rainfall, while increases in November are consistent with that timing change. The study does not directly test these explanations because the EDF rain-gauge stations lack temperature measurements, and its hourly records cannot resolve subhourly behavior. More fundamentally, return-level estimates depend on record length, sampling and measurement quality, while the model assumes linear change after 1985 and statistical homogeneity within neighborhoods. Overall, the results show that daily and hourly hazards cannot be treated as interchangeable in the French Alps: circulation changes strongly organize daily extremes, whereas short-duration events reflect a more complicated balance among moisture, energy, convection and topography.

Subject of Research: Observed climatology and trends of daily and hourly precipitation extremes in the French Alps

Article Title: Climatology and trends of observed daily and hourly extreme precipitation in the French Alps

Article References: Berghald, S., Blanchet, J., Blanc, A., & Penot, D. (2026). Climatology and trends of observed daily and hourly extreme precipitation in the French Alps. Hydrology and Earth System Sciences, 30(20), 6421-6442. https://doi.org/10.5194/hess-30-6421-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-6421-2026

Keywords: French Alps, extreme precipitation, rainfall extremes, hourly rainfall, daily rainfall, extreme value analysis, Mediterranean storms, climate trends, flash floods, weather patterns, Climatology, trends

ASPIRE is a planned research project rather than a completed forecast study, and the announcement does not report numerical forecast improvements or operational results. Its stated aim is to shorten the path from raw atmospheric data to physical insight. The project will build a coordinated AI system that combines curated data, findings from existing scientific literature and regional atmospheric simulations. Researchers intend to use that system to study relationships between small-scale physical processes and the large-scale weather patterns associated with severe storms, snow droughts and other extreme precipitation events.

A central element of the planned system is agentic AI, which the announcement describes as autonomous systems capable of carrying out complex, multistep tasks from high-level instructions. In a possible research workflow, a scientist could ask an AI agent to investigate the conditions behind an extreme snowfall event such as the December 2021 Colorado “Santa Slammer.” The agent could synthesize relevant literature, assemble datasets, design and run numerical experiments with advanced weather models, and interpret the results alongside previous research. ASPIRE’s team envisions these tools evolving from assistants for time-consuming tasks into systems that help scientists analyze data, design experiments and test hypotheses, while working alongside researchers as a graduate student would with an advisor.

The project’s energy focus is precipitation. According to the announcement, snowpack and rainfall patterns drive hydropower generation, while severe weather creates risks for grid infrastructure. By advancing precipitation prediction more quickly, ASPIRE is intended to provide scientific information relevant to utilities and communities nationwide. Almgren said the collaboration brings AI experts and atmospheric scientists together to translate collected atmospheric data into physical insight, describing the project as “team science” that examines large-scale patterns and fine-scale details. Flores said faculty and students working on ASPIRE would receive hands-on experience with research intended to have an impact across the country.

Boise State Vice President for Research and Economic Development Nancy Glenn said the partnership reflects the university’s commitment to research addressing challenges faced by utilities and critical infrastructure. She said better anticipation of extreme weather could strengthen grid reliability and give communities more time to prepare. The announcement presents ASPIRE as one of a select group of projects supported through the Genesis Mission, one of the Department of Energy’s most prominent AI research initiatives. Additional details about ASPIRE’s full scope and timeline are to be released with Berkeley Lab and the Department of Energy as part of the broader Genesis Mission rollout. The work is supported by the Department of Energy under Award #DE-AC02-05CH11231.

Subject of Research: Agentic AI for precipitation forecasting, hydropower generation and power-grid reliability

Article Title: Boise State joins project funded by the U.S. Department of Energy’s Genesis Mission and led by Lawrence Berkeley National Laboratory (Berkeley Lab)

Article References: Boise State joins project funded by the U.S. Department of Energy’s Genesis Mission and led by Lawrence Berkeley National Laboratory (Berkeley Lab). Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: artificial intelligence, agentic AI, precipitation forecasting, atmospheric science, extreme precipitation, hydropower, power-grid reliability, Department of Energy, Genesis Mission, Berkeley Lab, Boise State University, Boise

The review followed the PRISMA 2020 framework, screening records published between January 2021 and December 2025 in Scopus and Google Scholar. After removing duplicates and applying eligibility criteria focused on climate-driven hydrological change, flood hazards, snowmelt, and extreme precipitation, the team narrowed roughly 600 records down to 42 studies for final synthesis. Each study was then appraised across six dimensions: geographic relevance, hazard relevance, data quality, methodological rigor, future relevance, and decision relevance. This structured appraisal allowed the authors to distinguish conclusions supported by strong, consistent evidence from those requiring more cautious interpretation.

What emerges is a picture of a region already transformed. Central Asia has warmed by approximately 1.2 degrees Celsius over the past 70 years, and snow depth has declined by roughly 20 percent during the same period. Under the high-emission SSP5-8.5 scenario, warming could reach approximately 6.1 degrees Celsius by the end of the 21st century, implying substantial changes to snow storage, runoff formation, and flood-generating conditions across the region’s major river systems. Attribution analyses further indicate that anthropogenic forcing has already increased the likelihood and intensity of extreme temperature and precipitation events, particularly in mountain regions vulnerable to landslides, floods, and glacier-related hydrological change.

Extreme precipitation stands out as one of the most consistently reported and quantitatively supported drivers of flood-related hazards. Observational analyses covering 1936 to 2005 show that annual precipitation from wet days increased by 9.78 millimeters per decade, while the annual maximum one-day precipitation rose by 0.45 millimeters per decade. The proportion of precipitation falling on very wet days also increased, signaling a shift toward heavier rainfall events. Under the RCP4.5 scenario, projections indicate that total wet-day precipitation and maximum one-day precipitation would increase by 4.95 and 5.79 percent per degree of warming, respectively, implying thermodynamically strengthened rainfall extremes under continued warming.

Attribution evidence from vulnerable upstream regions of Kyrgyzstan and Tajikistan is particularly striking. Anthropogenic forcing has increased the frequency of extreme precipitation events by around 20 percent and their intensity by more than 100 percent in these mountain headwaters. Event-based analyses show that high-altitude parts of Central Asia experienced up to 698 extreme precipitation events between 1979 and 2023. These findings suggest that precipitation intensification is not uniform across the region but is becoming more spatially structured and process-dependent, with flood response varying considerably from one basin to another depending on elevation, catchment characteristics, and interactions with snow and glacier processes.

Snowmelt timing emerges as a second critical control on hydrological extremes. A large proportion of regional runoff is seasonally concentrated and influenced by snow accumulation, melt timing, and high-elevation hydro-climatic processes. In the Naryn basin, future warming is associated with a roughly 17 percent decline in maximum snow water equivalent, an advance of peak snow water equivalent timing by about one month, and runoff declines exceeding 40 percent in some upper-basin areas under high-emission scenarios. In the wider Tianshan-Pamir system, modeling suggests that by mid-century the frequency of rain-on-snow events will rise by approximately 49 percent, maximum daily intensity by roughly 43 to 45 percent, and annual total rain-on-snow intensity by approximately 66 percent, while the contribution of snowmelt to runoff may decrease by roughly 11 to 12 percentage points under SSP2-4.5 and SSP5-8.5 conditions.

Seasonal redistribution of runoff is perhaps the most consistent hydro-climatic feature reported across the reviewed literature. In the Sarykan River, summer runoff declined by 5.3 percent while runoff in the remaining months increased by 6.4 percent, indicating clear intra-annual redistribution. In western Kazakhstan, winter low flow increased by up to 5.2 times in parts of the Zhaiyk-Caspian system. Regional attribution analyses confirm that earlier spring streamflow peaks are widespread across Central Asian catchments. In the Kofarnihon River Basin of Tajikistan, climate change accounted for 87.96 percent of runoff change upstream, while anthropogenic activities accounted for 92.47 percent downstream, illustrating how the hydrological signal generated in mountain headwaters can be substantially modified as water moves through intensively managed lower reaches.

The review also highlights compound processes and interacting controls that defy simple attribution. The unprecedented flooding along the Tarim River in August 2022 was not generated uniformly across the basin: high temperatures and heat-related processes dominated some river systems, rainfall dominated others, and a combination of heat and rainfall affected still others. In the Amu Darya River Basin, long-term runoff reductions between 1960 and 2017 were mainly attributed to human activities including reservoir construction, cropland expansion, and irrigation demand, with the estimated contribution of human activities exceeding 100 percent at analyzed stations because anthropogenic effects counteracted an opposing climatic contribution. These findings underscore why future assessments must integrate climatic forcing with land-use change, urbanization, irrigation, and reservoir operation rather than treating them as separate categories.

Despite growing research attention, the evidence base remains geographically and thematically imbalanced. Region-wide Central Asian studies and Kazakhstan-specific analyses together account for more than half of the reviewed sample, while other single-country studies represent only 9.5 percent. Turkmenistan is particularly underrepresented, and conclusions for that setting should be regarded as especially cautious and provisional. Methodologically, observational trend and attribution analysis dominates at 54.8 percent of studies, followed by climate scenario modeling at 33.3 percent, while remote sensing, terrain-based hazard mapping, and dataset evaluation remain less developed. The literature is strongest in diagnosing hydro-climatic change but comparatively weaker in translating that understanding into operational flood risk tools, infrastructure criteria, or decision-ready adaptation frameworks.

The review identifies several priorities for future research: integrated multi-driver assessments that explicitly separate climatic and anthropogenic contributions; improved hydro-climatic observations and validation in mountain headwaters and transboundary basins; multidimensional flood characterization that accounts for frequency, duration, timing, and antecedent wetness; better integration of catchment and regulation effects including land-use change and urbanization; stronger country-specific and cross-basin synthesis; enhanced integration of transboundary systems into climate-risk research; and a decisive shift from hydro-climatic diagnosis toward adaptation-relevant risk assessment. Machine learning approaches, including sequence-based deep learning methods, could help capture non-linear dynamics among precipitation, snowpack evolution, temperature anomalies, and runoff timing in data-sparse basins. As Central Asia’s water authorities, river basin managers, and policymakers confront intensifying flood hazards alongside growing water stress, the review provides both a roadmap for future research and a structured evidence base for climate adaptation decisions across the region.

Subject of Research: Hydrological extremes and flood-related hazards in Central Asia under climate change

Article Title: Hydrological extremes and flood-related hazards across Central Asia: a systematic review and evidence-based synthesis of drivers, regional patterns, and research gaps

Article References: Sabitova, S., Broomandi, P., Bagheri, M., Arkhangelsky, E., Leuchner, M., & Kim, J. R. (2026). Hydrological extremes and flood-related hazards across Central Asia: a systematic review and evidence-based synthesis of drivers, regional patterns, and research gaps. Frontiers in Water, 8, Article 1890812. https://doi.org/10.3389/frwa.2026.1890812

Image Credits: AI Generated

DOI: 10.3389/frwa.2026.1890812

Keywords: Central Asia, flood hazards, hydrological extremes, snowmelt, rain-on-snow, extreme precipitation, climate change, runoff redistribution, transboundary basins, systematic review, water stress, climate adaptation

The stakes could hardly be higher. Mountain catchments supply fresh water to billions of people downstream, host extraordinary biodiversity, and sit squarely among the landforms most sensitive to a warming climate. When extreme rainfall strikes a steep Himalayan valley or a tropical Andean slope, the consequences can include flash floods, landslides and catastrophic debris flows. The four studies highlighted in the collection tackle these problems from different angles, spanning hydrological modelling, extreme event characterisation and environmental risk mapping, and together they demonstrate the breadth of what modern remote sensing can achieve in terrain that resists conventional monitoring.

One of the most striking findings concerns a deceptively simple question: how often does a satellite need to look at a watershed to capture its most violent moments? Sharma and colleagues assessed temporal sampling uncertainty in hydrological modelling using observations from the Global Precipitation Measurement mission and the TROPICS constellation of small satellites. Their analysis focused on data-scarce watersheds prone to hydrological extremes, where the timing of satellite overpasses can determine whether a flash-flood-producing storm is recorded at all or missed entirely. The conclusion was remarkably concrete: for reliable rainfall-runoff simulation and effective hydrological assessment in such watersheds, a minimum of three high-frequency CubeSats is likely necessary to achieve revisit intervals short enough to analyse the extreme events that trigger flash floods.

This finding matters because it converts an abstract modelling concern into an actionable engineering requirement. Precipitation products differ in their spatial resolution, and uncertainties propagate through hydrological models in ways that can distort runoff simulations precisely when accuracy matters most, during extremes. By delineating the impact of temporal sampling uncertainty on modelled runoff, the study provides satellite mission planners with a quantified target. As smallsat constellations become cheaper to deploy, the research suggests that a modest fleet of radar-equipped CubeSats could transform flood forecasting in mountain basins that have never had a ground-based gauge.

Half a world away in the tropical Andes, another team confronted the same challenge of catching extreme rain in the act. Urgilés and colleagues employed high-resolution X-band weather radar to characterise 67 critical extreme rain events, mapping hotspots based on parameters such as peak intensity, spatial area and the cohesion of storm core cells. Weather radar of this class can resolve the internal structure of a storm as it evolves minute by minute, revealing how concentrated the rainfall becomes within small, intense cells that satellite precipitation products with coarser footprints can smear or miss. The work provides critical insights for improving localised flood forecasting in a region where steep topography funnels intense rainfall into narrow valleys with little warning.

The remaining two studies turned to the diverse mountain regions of India, from the Sikkim Himalayas to the Western Ghats of Kerala. In the largely ungauged Lachen Basin of the upper Teesta catchment, Mukherjee and Krishna investigated a coupled temperature-index based snowmelt runoff model. They combined the traditional temperature-index routing approach with a linear reservoir routing scheme, driving the model with remote sensing and gridded datasets rather than ground observations. The approach led to improved water resource forecasts in this data-scarce Himalayan basin, demonstrating that snowmelt-driven river flows can be modelled credibly even where stream gauges are absent, a capability with direct implications for hydropower operations and downstream water security.

In Kerala’s Iduki district, Drisiya and Singh illustrated the value of multi-source remote sensing in assessing the cumulative pressures on tropical mountainous landscapes. Their integrated eco-environmental risk mapping revealed a significant overlap between extreme precipitation events and landslide-affected areas, identifying high-risk zones across the region. Crucially, they observed that risk was heightened by increasing development corresponding to higher tourism and the forest alteration that accompanies it in steep topography. The findings serve as a stark reminder that human activity is transforming land use in natural mountainous landscapes and increasing their environmental sensitivity, compounding the hazards posed by a changing climate.

Despite these advances, the editorial is candid about the constraints that still bind mountain remote sensing. Open-source Digital Elevation Models often represent topography inadequately at the resolutions these applications demand, and the modelling of snow and glacier land cover in hydro-fluvial and geomorphological models remains imprecise. Validation poses a further challenge in data-sparse regions where physical accessibility is a critical issue, which in turn constrains the development of machine learning frameworks and physically based process models that require reliable ground truth. The spatio-temporal resolution of available remote sensing data itself inhibits the broader application of models across mountain catchments, where conditions can change dramatically over distances of a few hundred metres.

Looking ahead, the editorial identifies clear opportunities. The next generation of satellite missions is expected to offer higher spatial and temporal resolution, enhanced spectral coverage and significant innovation in sensing modalities, expanding the observational envelope for mountain science. Closer coupling of remote sensing data with in-situ measurements and crowd-sourced information is expected to improve Earth observation and bio-physical products in mountainous regions, blending the strengths of orbiting sensors with the knowledge of communities who live in these landscapes. The collection contributes to a growing body of evidence for remote sensing as a foundational element of mountain science and sustainability, and the editors encourage the broader scientific community to continue advancing both the science and application of Earth observation in mountainous environments.

Subject of Research: Editorial: Remote sensing applications in mountainous regions

Article Title: Editorial: Remote sensing applications in mountainous regions

Article References: Varade, D., Shukla, D. P., Mandal, D., & Das, N. (2026). Editorial: Remote sensing applications in mountainous regions. Frontiers in Remote Sensing, 7, Article 1981032. https://doi.org/10.3389/frsen.2026.1981032

Image Credits: AI Generated

DOI: 10.3389/frsen.2026.1981032

Keywords: remote sensing, mountains, hydrology, extreme precipitation, landslides, satellites, snowmelt, flood forecasting, Editorial, Remote, sensing, applications

The study employed two complementary data sources: ground-based measurements from weather stations across Bulgaria and the ERA5-Land reanalysis product, which provides gridded hourly data from 1950 to the present. The researchers found that the two datasets show good agreement in terms of the frequency of extreme precipitation events, but not in terms of their average values. Notably, the ERA5-Land reanalysis systematically underestimates heavy precipitation events, a finding consistent with previous evaluations of ERA5 precipitation over Spain and other regions.

To detect trends, the team applied a suite of statistical methods including Sen’s slope estimator, the Mann-Kendall test, and correlation analysis. Sen’s slope estimator provides a robust nonparametric measure of trend magnitude, while the Mann-Kendall test assesses the statistical significance of monotonic trends in time series data. These methods are particularly well-suited to precipitation data, which often violate the distributional assumptions of parametric tests.

The results show that the frequency of extreme precipitation events—the number of cases per year—increases in almost all studied stations and grid cells during the period 1979–2025, according to both instrumental observations and ERA5-Land reanalysis. However, when the analysis is extended to the longer period 1950–2025 using only the reanalysis data, no trend is detected in frequency. This discrepancy suggests that the intensification of extreme precipitation is a relatively recent phenomenon, likely linked to the accelerated warming of recent decades.

The average values of daily extreme precipitation also show predominantly positive trends for the period 1979–2025 according to both datasets, and weak positive or no trends for the longer 1950–2025 period. In some stations and grid cells, these trends are statistically significant. The most pronounced increases in extreme precipitation are observed in southern and, especially, southeastern Bulgaria, a region that appears to be experiencing a disproportionate intensification of heavy rainfall.

One of the most striking findings of the study is an increase in the 21st century in the duration of extreme precipitation periods that extend beyond a single day. Multi-day extreme precipitation episodes are particularly dangerous, as they can saturate soils and overwhelm drainage infrastructure, leading to severe flooding. This finding aligns with broader research across the Balkans, where studies in Serbia, Montenegro, Bosnia and Herzegovina, Romania, and Greece have documented similar trends toward more intense and persistent heavy rainfall under a warming climate.

The physical mechanisms underlying these trends are consistent with the expectations of climate science. As the atmosphere warms, the Clausius-Clapeyron relation dictates that its water-holding capacity increases by approximately 7% per degree Celsius of warming. This means that when precipitation systems develop, they have more moisture available to convert into rainfall, increasing the likelihood of extreme daily totals. Changes in atmospheric circulation patterns over the Mediterranean and the Balkans, which Nojarov has documented in previous work, may also contribute to the observed intensification.

The spatial heterogeneity of the trends is noteworthy. While positive trends dominate across most of the country, the concentration of the most significant increases in southern and southeastern Bulgaria suggests that regional factors—including proximity to the Aegean Sea, orographic effects, and the influence of Mediterranean cyclone tracks—play a crucial role in modulating how global warming translates into local extreme precipitation. This finding has important implications for water management, flood preparedness, and infrastructure planning in the region.

The methodological rigor of the study deserves emphasis. By comparing ground-based observations with reanalysis data over the same period, the researchers were able to validate the reliability of their trend detections and identify the limitations of the reanalysis product. The finding that ERA5-Land underestimates heavy precipitation events is particularly valuable for the research community, as it cautions against relying solely on reanalysis data for studies of precipitation extremes—a caution echoed by other evaluations of ERA5 and ERA5-Land precipitation in Europe.

As Bulgaria and its neighbors continue to experience the effects of a warming climate, the findings of this study provide crucial evidence for understanding how extreme precipitation is evolving in southeastern Europe. The predominance of positive trends in both the frequency and intensity of extreme precipitation events during recent decades, combined with the emergence of longer-duration extreme precipitation periods in the 21st century, suggests that the region faces increasing risks of flooding and water-related hazards. Continued monitoring, improved high-resolution modeling, and adaptive water management strategies will be essential for building resilience in the face of these changing precipitation patterns.

Subject of Research: Extreme precipitation trends in Bulgaria in conditions of a warming climate

Article Title: Extreme precipitation trends in Bulgaria in conditions of a warming climate

Article References: Nojarov, P., Svetozarevic, J., Parvanova, D., & Belev, G. (2026). Extreme precipitation trends in Bulgaria in conditions of a warming climate. Theoretical and Applied Climatology, 157(11), Article 727. https://doi.org/10.1007/s00704-026-06655-4

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06655-4

Keywords: extreme precipitation, Bulgaria, climate change, ERA5-Land, Mann-Kendall test, Sen's slope estimator, Theoretical and Applied Climatology, Bulgarian Academy of Sciences, southeastern Europe, hydrology, Extreme, precipitation

The study is notable for the dataset it relies on. The team employed the High Asia Refined Analysis version 2, or HARv2, a 10-kilometer-resolution atmospheric product generated by dynamically downscaling the global ERA5 reanalysis with the Weather Research and Forecasting model. Because the Himalayan terrain is notoriously under-observed, with sparse gauge networks at high elevations, the researchers validated HARv2 against two independent benchmarks: the India Meteorological Department gridded precipitation dataset at 25 kilometers, the only ground-based observational reference available for the region, and the 12-kilometer IMDAA reanalysis developed by the National Centre for Medium Range Weather Forecasting. Five statistical metrics, including correlation coefficient, root mean square error, and Kling-Gupta efficiency, confirmed reasonable agreement across all three sub-regions, with correlations rising from 0.73 in Jammu & Kashmir to 0.91 in Uttarakhand against IMD.

That north-to-south improvement in model performance is itself physically meaningful. The Indian Summer Monsoon is a large-scale, organized circulation system that a WRF-based downscaling captures well, whereas Jammu & Kashmir depends heavily on western disturbances, the eastward-moving synoptic weather systems that deliver most winter and spring precipitation to the western Himalaya. These systems are far more variable and harder to simulate accurately, which explains the lower correlation values in the northernmost sub-region.

To quantify precipitation extremes, the team applied ten indices defined by the Expert Team on Climate Change Detection and Indices, a standardized framework of the World Meteorological Organization used worldwide for detecting climate extremes. Trends were assessed with the Mann-Kendall non-parametric test and Sen’s slope estimator, the most robust statistical tools for climatological trend detection. The results for Jammu & Kashmir were striking. Total annual wet-day precipitation, PRCPTOT, declined by 95.41 millimeters per decade in the IMD dataset and 74.27 millimeters per decade in HARv2, both statistically significant. The Simple Daily Intensity Index fell by roughly 0.35 millimeters per day per decade in both datasets, and the number of days exceeding 10 millimeters and 20 millimeters dropped by about 3 to 4 days and 1 day per decade respectively. The contribution of very heavy rainfall days, R95p, also declined significantly.

Equally concerning is the behavior of dry spells. The consecutive dry days index, CDD, showed a significant annual increase of 4.1 days per decade over Himachal Pradesh, supported by significant pentadal trends, while Jammu & Kashmir exhibited significant CDD increases at the five-year averaging scale even though the annual trend remained non-significant. Taken together, the declining totals, falling rainfall intensity, fewer moderate-to-heavy events, and lengthening dry spells point toward a transition to a drier and more intermittent precipitation regime across the northern Northwest Himalaya, with direct implications for snowpack accumulation, glacier mass balance, agriculture, and water security for the millions of people downstream who depend on Himalayan freshwater.

Uttarakhand tells a completely different story. None of the ten indices showed statistically significant trends over the 40-year record, though several, including R95p, R99p, and consecutive wet days, displayed weak increasing tendencies. The researchers attribute this to the dominant role of orographic processes: the region’s steep terrain efficiently converts even modest variations in incoming moisture into precipitation through strong terrain forcing, buffering the local climate against large-scale circulation changes.

The most technically ambitious part of the study involved atmospheric diagnostics designed to answer a deceptively simple question: is moisture supply the controlling factor behind Himalayan precipitation extremes? The team computed vertically integrated moisture transport, VIMT, by integrating the product of specific humidity and horizontal wind components from 1000 to 300 hectopascals in both HARv2 and ERA5. Over Jammu & Kashmir, the two datasets agreed moderately, with a correlation of 0.58 and a mean bias of only 15.67 kilograms per meter per second. But the agreement deteriorated progressively southward, falling to 0.29 over Himachal Pradesh and 0.22 over Uttarakhand, while the mean bias ballooned to 57.52 and 92.52 kilograms per meter per second respectively. This divergence reveals the growing importance of terrain channelling, localized moisture convergence, and precipitation-driven moisture depletion, processes that the fine-resolution HARv2 resolves but the coarser ERA5 smooths away.

Correlations between HARv2 diagnostic variables, including precipitable water, cloud liquid water, cloud ice water, cloud cover, and vertical velocity at 500 and 300 hectopascals, and the ETCCDI indices exposed a clear regional gradient in precipitation-generating mechanisms. Over Jammu & Kashmir, significant correlations were limited and mostly confined to wet spell duration, suggesting extremes there depend more on synoptic disturbances and terrain-triggered uplift than on moisture availability alone. Over Himachal Pradesh, a transitional regime, the relationships broadened to include total precipitation, intensity, and multi-day extremes linked with cloud hydrometeors and vertical motion. Over Uttarakhand, the coupling was strongest of all, with the highest correlation observed between PRCPTOT and vertical motion at 500 hectopascals, at 0.71. The conclusion is unambiguous: moisture alone does not explain precipitation extremes in the Northwest Himalaya, and orographic uplift and synoptic-scale disturbances are equally important drivers.

The El Niño-Southern Oscillation adds another layer of complexity. Classifying years by the Niño 3.4 sea surface temperature anomaly index, the team found that over Jammu & Kashmir the index correlated significantly and negatively with VIMT from both HARv2 and ERA5, confirming that El Niño phases weaken the Walker circulation and suppress moisture inflow while La Niña phases enhance it. This teleconnection weakened over Himachal Pradesh and Uttarakhand, where local terrain-driven processes increasingly modulate the moisture field. ENSO composites showed stronger and more widespread moisture penetration into higher elevations during La Niña years, with significant positive departures across nearly all diagnostic variables over Uttarakhand.

The authors acknowledge limitations, including the known wet bias of HARv2 over complex terrain, which affects magnitude-sensitive indices such as R95p and R99p, and the sparse high-elevation gauge coverage in the IMD benchmark. The 20-year sub-period split, chosen to separate the pre-satellite and satellite eras, was not formally tested with a changepoint algorithm, so inter-period differences should be read as multi-decadal variability rather than definitive regime shifts. Still, the central message stands: the Northwest Himalaya must be treated as a mosaic of climatically distinct sub-regions, each demanding targeted adaptation strategies, improved observational coverage, and continued high-resolution process-based monitoring to protect a region where drying in the north and terrain-amplified extremes in the south will shape hydrological risk for decades to come.

Subject of Research: Climate Sciences

Article Title: An observational perspective on extreme precipitation indices over the mountainous regions of Northwest Himalaya

Article References: Nag, K., Pattanayak, S., Venkatramanan, V., Sharma, V., & Shekhar, M. S. (2026). An observational perspective on extreme precipitation indices over the mountainous regions of Northwest Himalaya. Discover Geoscience, 4(1), Article 404. https://doi.org/10.1007/s44288-026-00775-0

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00775-0

Keywords: Northwest Himalaya, extreme precipitation, ETCCDI, HARv2, climate change, ENSO, moisture transport, drought, Jammu & Kashmir, Himachal Pradesh, Uttarakhand, observational

Scientists led by Carla de Souza Farias of Brazil’s National Institute for Amazonian Research set out to answer a deceptively simple question: does the mere presence of a city in the central Amazon induce more extreme precipitation? Previous work on Manaus had examined single storm events, leaving open whether the urban effect was a one-off curiosity or a robust, repeatable pattern. To settle the matter, the team assembled a decade of observations from 2010 to 2020 and identified 45 extreme precipitation events that appeared in two independent datasets: the IMERG satellite rainfall product and the ERA5 atmospheric reanalysis produced by the European Centre for Medium-Range Weather Forecasts.

The screening was deliberately strict. Of the roughly 177 extreme events flagged by IMERG and 176 by ERA5, only 45 were coincident in both products. This filtering matters because satellite grids and reanalyses often disagree on the exact timing and location of localized convective storms in the Amazon, and the researchers wanted simulations initialized with physically consistent conditions. Extreme days were defined as those exceeding the 97.5th percentile of daily rainfall, and the events clustered in the rainy season from November to May, with February and June showing the highest monthly counts.

For each of the 45 events, the team ran paired simulations with the Weather Research and Forecasting model, or WRF, coupled to the Single Layer Urban Canopy Model. In the urban experiment, Manaus was represented with three district classes: industrial and commercial zones, low-intensity residential areas, and high-intensity residential neighborhoods, with parameters tuned using local street widths, roof dimensions, and census-based population density. In the control experiment, the entire urban footprint was replaced with forest. Ninety simulations in total, each 60 hours long, allowed the researchers to compare the same storms over the same 427 square kilometers with and without the city.

The verdict was striking. In 65 percent of the events, total accumulated rainfall was higher with the city present, and in the urban center the average increase reached 16.8 percent. Hourly rainfall intensity told an even sharper story: about 79.2 percent of hourly values in the urban run exceeded those of the forested control, and the maximum hourly precipitation intensity rose by up to 50 percent. The urban scenario also produced longer-lived storms, with the 90th percentile of event duration reaching 19 hours compared with 16.6 hours without the city, suggesting that urban forcing helps sustain convection once it begins.

The mechanism unfolds in three linked steps. First, the city heats up. Asphalt, concrete, and other high-thermal-inertia materials raised the simulated surface temperature by as much as 2.9 degrees Celsius at midday relative to the forested scenario, with air temperature at two meters rising by about 0.6 to 0.7 degrees Celsius. This warming acts as a thermal trigger: the anomalous surface heating peaks around noon, and the largest precipitation intensification follows in the afternoon between 14:00 and 17:00 local time, exactly the window in which the urban heat island of Manaus has been observed to peak in field measurements.

Second, the heated boundary layer drives the atmosphere upward. The researchers found a statistically significant positive correlation between the precipitation difference and vertical wind speed at the 800 hectopascal level, in the middle troposphere, with a correlation coefficient of 0.39 and a p-value of 0.008. Stronger updrafts over the city lift warm, unstable air toward the level where deep convective clouds form. Third, and perhaps most surprising given the setting, the city imports its own moisture. Although near-surface humidity was actually lower over the urban area than over the forest, because trees pump water into the air through evapotranspiration, specific humidity at 800 hectopascals showed a significant positive correlation with the precipitation increase. The interpretation is that urban heating and the rough, friction-generating canopy converge moist air laterally from the surrounding forest and rivers, drawing it into the urban core where it fuels towering convective cells.

The spatial fingerprint of this process is unmistakable. When the domain was divided into nine geographic sectors, the central quadrant coinciding with the urban area recorded the highest accumulated rainfall, 1,136.1 millimeters across the 45 events, and captured the maximum in 11 cases, or 24.4 percent of the total. The southwest sector followed with 1,101.3 millimeters and nine events, consistent with the prevailing northeast-to-southwest flow that advects storms initiated over the city downwind. In the forested control run, by contrast, the precipitation maxima shifted to the peripheral southeast and east sectors, and the central sector dropped to 972.6 millimeters with only five winning cases. In effect, the city traps and concentrates rainfall that would otherwise be distributed over forest and rivers.

The findings place Manaus within a growing global picture. A 2024 analysis of more than 1,000 cities found that over 60 percent of urban areas act as so-called urban wet islands, where precipitation exceeds that of surrounding rural areas, a phenomenon that has roughly doubled in magnitude over the past two decades. Yet the Amazonian case is distinctive because of its setting: the city sits between the Negro and Solimões rivers, embedded in an atmosphere so humid that near-surface specific humidity exceeds 18 grams per kilogram even during extremes. That abundance of moisture means the urban heat island does not dry out storms, as it can in arid cities such as Phoenix, but instead supercharges them, converting a modest thermal anomaly into a deep convective engine fed by the surrounding rainforest.

The practical implications are sobering. Manaus already suffers recurring flash floods and landslides, aggravated by rapid and often unplanned growth on highly impermeable ground. If urbanization reliably intensifies the heaviest rainfall by double-digit percentages, then the city’s own expansion is compounding its flood risk at precisely the moment climate change is loading the atmosphere with more water vapor. The authors note that their controlled experiments deliberately excluded anthropogenic heat emissions and urban aerosols, both of which could add further complexity, and they call for multi-layer canopy models and coupled chemistry simulations in future work. For now, the message for Amazonian cities is clear: as the forest gives way to pavement, the storms above it grow heavier, and the water that falls has nowhere to go but the streets.

Subject of Research: Urbanization effects on extreme precipitation in the central Amazon city of Manaus

Article Title: Does the urban presence of a city in the central Amazon induce more extreme precipitation?

Article References: de Souza Farias, C., Candido, L. A., de Melo Monte, M. J., Buscioli Capistrano, V., Andreoli, R. V., Holanda Hall, D., de Jesus Machado, J., & Quaresmas Dias-Júnior, C. (2026). Does the urban presence of a city in the central Amazon induce more extreme precipitation?. Theoretical and Applied Climatology, 157(11), Article 722. https://doi.org/10.1007/s00704-026-06646-5

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06646-5

Keywords: Manaus, Amazon, extreme precipitation, urban heat island, WRF model, convection, urbanization, flooding, tropical meteorology, moisture convergence, land-use change, climate modeling

The team turned to the WegenerNet, a climate station network in the Feldbach region of southeastern Austria, where 155 gauges have operated since 2007 within an area of roughly 330 square kilometres. After removing twelve stations that were relocated during the study period, the researchers analysed 143 stations recording precipitation every five minutes from 2007 to 2022. That combination of spatial density and temporal resolution is extraordinarily rare: most national gauge networks are far sparser and log data only hourly or daily, which means short-lived convective storms, the very events that matter most for flash flooding, routinely slip between the gauges unnoticed. In Feldbach, a region where thunderstorms, hail, and severe wind gusts are among the most frequent in Austria, hardly any storm escapes detection.

The methodological logic is straightforward in principle. A point rain gauge records the rainfall at one spot, but a catchment integrates rainfall over its whole surface, and because rain rarely falls uniformly, the average depth over an area is almost always less than the peak depth at the storm’s core. The ARF captures this ratio: multiplying a point design value by the factor yields the areal equivalent. The researchers applied a fixed-area approach, constructing square windows around each station ranging from 1 to 225 square kilometres, computing areal precipitation as the mean of the 200-by-200-metre grid cells within each window, and then running an extreme value analysis on both point and areal series. Because the 15-year record is short, they used a peak-over-threshold sampling scheme with roughly forty extreme events per series, fitting an exponential distribution whose parameters were smoothed across sixteen durations spanning five minutes to twenty-four hours.

The results confirm several long-standing expectations with unprecedented clarity. Areal reduction factors increase as duration lengthens, reflecting the fact that long-lasting storms tend to cover wide areas, and they decrease as the area grows, because averaging over a larger surface increasingly dilutes the localised peak. The five-minute duration produced the lowest factors of all, with a minimum mean value of 0.595 for a one-year return period, meaning that the areal average over the largest areas was barely sixty percent of the point extreme. Seasonal analysis revealed a striking split: the smallest factors occurred in spring and summer, when short, intense convective thunderstorms dominate and rainfall is spatially patchy, while winter, governed by broad, advective weather systems, produced consistently higher factors, sometimes with a substantial fraction of values exceeding one.

But the study’s most consequential finding challenges a comfortable assumption embedded in engineering guidelines. Many national standards, including the widely used German DWA-M 531, allow point precipitation to stand in for areal precipitation on catchments up to 25 square kilometres, treating the deliberate overestimation as a safety margin. The Austrian data show this shortcut is far from harmless. Even for areas of just one square kilometre, the mean ARF fell below one in every season and return period examined, with the largest reduction, 7.2 percent, occurring in spring at a one-year return period. Applying no reduction to a 25-square-kilometre area at a five-minute duration would inflate the areal design rainfall by 26.5 percent, an overestimation the authors describe as excessive when stacked on top of the other conservative margins already built into design practice. Oversized drainage infrastructure is not merely wasteful; it can alter flow regimes and complicate maintenance decisions for decades.

The second surprise concerns return periods, the statistical yardsticks that describe how rare an event is. The conventional wisdom, supported by studies from the United States and elsewhere, holds that ARFs shrink as events become rarer, because the most extreme storms tend to be the most localised. For durations between one and twelve hours and areas above nine square kilometres, the Austrian data do show this decline. For sub-hourly durations, however, the pattern reverses: the factors partially increase with the return period, apparently because the spatial variability of very short storms diminishes as their intensity grows. The authors report that no comparable behaviour could be found in the existing literature, and they attribute the discovery directly to the quality of their data, noting that the WegenerNet appears to be the first station-based data set of this kind ever used to estimate areal reduction factors. Whether the pattern holds elsewhere, or reflects a regional or observational-scale effect, remains an open question.

Topography, by contrast, turned out to matter far less than some earlier studies suggested. The Feldbach landscape, with its elongated hills of the East Styrian foothills rising from about 250 to 598 metres above sea level, might plausibly imprint itself on storm structure. Yet Spearman correlations between the reduction factors and station elevation never exceeded 0.2 in magnitude, and for a ten-year return period the relationship even turned slightly negative. The authors caution that the interpolation scheme used to build the regionalised precipitation grids did not explicitly account for elevation, a defensible choice given the modest relief of the network, so the weak signal should be interpreted with care. Still, the finding contrasts with work in Shenzhen, China, where mountainous terrain produced markedly lower factors than flat coastal zones, underscoring that the geography of these factors is genuinely contested.

That geographic debate gives the study practical urgency. Some analyses have argued that factors derived in one country transfer readily to another, while a nationwide assessment across the United States found high regional variability and warned against arbitrary transposition. The Austrian results sit squarely in the cautionary camp: because the factors depend on duration, area, return period, and season in ways that interact non-trivially, applying a single constant factor across broad regions or treating point and pixel-based estimates, such as radar or satellite retrievals, as interchangeable with gauge measurements risks systematic error. The authors also observed an irregularity at the longest durations, where eighteen- and twenty-four-hour factors occasionally dipped below those of six and twelve hours, possibly because regional-scale storms extended beyond the network’s footprint and were incompletely captured.

The broader lesson is about what it takes to see the atmosphere clearly. Radar offers superb spatial and temporal coverage but performs poorly for rare events because its archives are short; gridded gauge products lack the temporal resolution for sub-daily analysis; sparse gauge networks miss storm centres entirely. Only a dense, high-resolution, long-running gauge network like the WegenerNet can resolve the fine texture of convective rainfall that determines how a point measurement translates to a landscape. The Braunschweig team recommends that comparable networks elsewhere be exploited to test whether the sub-hourly reversal with return period is a universal feature of storm physics or a fingerprint of the Styrian climate, and that alternative fixed-area calculation methods be applied to the same data. For now, the message for practitioners is concrete: even a single square kilometre is not a point, and the humble reduction factor deserves a place in every design calculation, no matter how small the catchment.

Subject of Research: Empirical estimation of areal reduction factors for extreme precipitation using a high-density rain gauge network in Austria

Article Title: Areal reduction factors from high-resolution rain gauges in Austria

Article References: Goshtasbpour, G., Alpers, J., Schröter, K., & Müller-Thomy, H. (2026). Areal reduction factors from high-resolution rain gauges in Austria. Natural Hazards and Earth System Sciences, 26(10), 5057-5071. https://doi.org/10.5194/nhess-26-5057-2026

Image Credits: AI Generated

DOI: 10.5194/nhess-26-5057-2026

Keywords: areal reduction factors, extreme precipitation, rain gauges, WegenerNet, Austria, hydrology, design storms, intensity-duration-frequency, convective storms, flood risk, extreme value analysis, urban drainage

The research forms part of the Mediterranean Extreme Events Experiment, known as M3E, an international initiative co-coordinated by OGS, the Istituto Nazionale di Oceanografia e di Geofisica Sperimentale, and the CMCC Foundation, the Euro-Mediterranean Center on Climate Change. The experiment combines targeted in-situ observations with numerical modelling to improve forecasts of extreme weather and to deepen scientific understanding of ocean–atmosphere interactions during storms. Its central tool is deceptively simple: drifting buoys equipped with barometers and sea surface temperature sensors, deployed across parts of the Mediterranean where conventional atmospheric measurements have long been scarce. The instruments transmit their readings in near real time, making them available to operational forecasting systems where they can be assimilated into numerical weather prediction models.

The motivation for filling this observational gap is rooted in the Mediterranean’s peculiar geography and meteorology. The sea is relatively small, yet it is capable of generating extraordinarily powerful storms, including small, rapidly evolving low-pressure systems that are notoriously difficult to capture accurately in forecast models. Emanuela Clementi, senior scientist at CMCC and co-author of the study, emphasised the importance of collecting observations directly at sea, noting that the Ionian Sea and the central Mediterranean have historically had limited coverage of in-situ atmospheric measurements despite being areas where intense storms and persistent cyclones can develop. Without measurements from these waters, forecast models must rely on sparse data, satellite retrievals and assumptions, leaving room for significant errors in the initial conditions from which predictions are generated.

The study’s findings demonstrate that atmospheric pressure observations from the M3E buoys had a beneficial impact on forecasts during Storm Harry. In particular, data from a buoy positioned in the Ionian Sea contributed to improving the representation of atmospheric circulation as the storm crossed the region. This matters because numerical weather prediction is fundamentally an initial-value problem: the model’s forecast evolves from a snapshot of the current state of the atmosphere, and even small errors in that snapshot can grow rapidly, especially in the presence of unstable, fast-developing weather systems. Pressure measurements from the surface of a data-sparse sea directly constrain that snapshot where it is weakest.

To quantify the value of the buoy network, the researchers conducted an additional experiment comparing forecasts that included the available Mediterranean drifting-buoy observations with forecasts in which those observations were deliberately withheld. This observational denial approach is a standard technique in forecast sensitivity studies, allowing scientists to isolate the contribution of a particular data stream. The results revealed significant differences in forecast sea-level pressure near the Balearic Islands, close to the small low-pressure systems associated with the storm. In other words, removing the buoy data degraded the model’s depiction of the very features that made Storm Harry dangerous, providing concrete evidence that the network improves forecasts precisely where it matters most.

The forecasting performance during the event was mixed in instructive ways. Forecasts produced by the European Centre for Medium-Range Weather Forecasts, ECMWF, successfully predicted the heavy rainfall in southern Italy approximately seven days ahead of the event, a remarkable lead time for an extreme precipitation event. However, forecasting precipitation over North Africa proved considerably more challenging, partly because of the small, rapidly evolving low-pressure systems that are difficult to represent accurately in models. This asymmetry highlights a persistent challenge in Mediterranean meteorology: mesoscale systems that develop quickly over warm waters can slip through the observational and modelling net, and their impacts can be severe for coastal communities on the southern shores of the basin.

The implications extend beyond atmospheric forecasting alone. CMCC contributes to M3E through its coordination role and its expertise in ocean forecasting, cyclone dynamics, extreme precipitation and observational systems. The atmospheric forecasts produced by ECMWF provide essential forcing conditions for the CMCC Mediterranean Forecasting System, MedFS, which operates within the Copernicus Marine Service. Clementi described the connection as a natural one for CMCC, explaining that a better atmospheric forecast also means a better forecast of the sea. During extreme storms, accurate wind and pressure fields drive wave height, storm surge and currents, all of which matter for maritime safety, coastal protection and emergency response.

Because observations are essential for understanding and forecasting these processes, CMCC also contributes directly to expanding the observational network by providing and deploying additional drifting buoys, including in the Tyrrhenian Sea, to improve spatial coverage across the Mediterranean. The effort is not purely operational. The buoy observations offer an opportunity to better understand how the ocean and atmosphere interact during extreme storms, a scientific question with implications for how Mediterranean cyclogenesis is represented in models. By bringing together in-situ observations, weather forecasting and ocean modelling, M3E promotes an integrated approach supported by a partnership of European and US institutions and endorsed by the World Weather Research Programme, WWRP.

The research identifies small-scale low-pressure systems as a priority for future observational campaigns. Better coverage of these features, the authors argue, could help reduce forecast uncertainty and improve understanding of the local impacts of Mediterranean storms. This is a significant conclusion for the region, where compact but intense systems can deliver destructive rainfall and winds to limited areas with little warning. The Mediterranean’s vulnerability is compounded by its dense coastal population, its role as a migration corridor and its exposure to landslides and flooding in steep terrain, all of which were evident in the impacts of Storm Harry.

Building on these initial results, plans are already underway to expand the observing network from autumn 2026. Seven drifting buoys acquired by CMCC are set to be deployed in the Tyrrhenian Sea, alongside additional deployments in the Ionian Sea, strengthening coverage in areas where improved data are particularly important for forecasting extreme events. For a sea that can generate very powerful storms while the countries around it pay a high price, as Clementi put it, the message from Storm Harry is clear: filling the observational gap exactly where these storms form is one of the most direct paths to earlier, more reliable warnings, and the first evidence that the buoy data deliver on that promise has now been formally established.

Subject of Research: Impact of drifting buoy atmospheric observations on forecasts of the Mediterranean storm Harry in January 2026

Article Title: Storm Harry: Drifting buoys help improve early forecasts of extreme events in the Mediterranean Sea

Article References: Storm Harry: Drifting buoys help improve early forecasts of extreme events in the Mediterranean Sea. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: Storm Harry, Mediterranean Sea, drifting buoys, M3E experiment, weather forecasting, CMCC, OGS, ECMWF, data assimilation, extreme precipitation, low-pressure systems, Quarterly Journal of the Royal Meteorological Society

The research team, led by Yuanyuan Ding of East China University of Technology together with Wenbo Chen and Zhuozhao Chen, set out to solve a persistent puzzle in wetland ecology. Scientists have long known that precipitation strongly influences the structure and stability of wetland plant communities, but the mechanisms linking extreme rainfall to diversity patterns have remained murky, particularly because those links appear to change depending on the spatial scale at which they are observed. Understanding these mechanisms matters far beyond academic curiosity: Poyang Lake’s famous seasonal grasslands provide critical foraging habitat for rare and endangered birds, including Siberian cranes, so shifts in vegetation diversity ripple straight up the food web.

By integrating satellite observations with machine learning approaches, including partial least squares structural equation modelling, the researchers were able to disentangle the direct and indirect pathways through which extreme precipitation acts on the lake’s vegetation. The first major finding is that extreme rainfall events have generated pronounced spatial heterogeneity and non-stationarity in both grassland diversity and the environmental factors that govern it. In practical terms, the wetland no longer responds as a uniform system: neighboring patches of grassland can behave very differently after the same storm, and the statistical relationships between rainfall and diversity shift across the landscape rather than holding to a single rule.

That non-stationarity is central to the study’s significance. Traditional ecological models often assume that a driver such as precipitation has a consistent, transferable effect across a region. The Poyang Lake analysis shows instead that the influence of extreme precipitation is contingent on local context, varying with position in the floodplain, exposure to inundation and the underlying environmental template. For conservation managers, this means that a one-size-fits-all restoration prescription is unlikely to work; interventions tuned to local conditions, informed by where and when extremes strike, will be essential.

The second core result concerns scale. Soil, topographic and hydrologic factors all influence vegetation diversity, but their relative importance diverges sharply depending on whether diversity is measured at the landscape scale or at the community scale. At the landscape level, water level emerged as a key mediator, translating extreme rainfall into changes in landscape diversity, with a negative effect: higher water levels were associated with lower landscape-scale diversity. At the community scale, by contrast, soil organic carbon played the mediating role, positively influencing the importance values of plant species. In other words, the same extreme rainfall event can depress diversity across broad swathes of the lake while simultaneously altering the competitive balance among species within individual plant communities, through entirely different environmental channels.

This scale dependence has a broader resonance in ecology. Landscape-scale diversity reflects the number and arrangement of habitat types across the floodplain, which is governed largely by how water sculpts the terrain, while community-scale diversity reflects which species coexist in a given patch, which depends heavily on soil conditions. The study demonstrates that extreme precipitation feeds into both of these arenas, but through distinct causal chains, and that analyses conducted at a single scale risk missing critical pathways. For Poyang Lake, where the annual flood pulse creates a mosaic of exposed grasslands, mudflats and open water, this dual mediation by water level and soil carbon captures the essence of how a flood-driven ecosystem is organized.

Perhaps the most striking finding is temporal. Vegetation diversity patterns did not respond immediately to extreme precipitation; they lagged behind it. And the lags themselves were scale-dependent: community-scale responses occurred later than landscape-scale responses. Landscape structure, shaped by water levels, adjusts relatively quickly after major rainfall, whereas the composition of plant communities, filtered through soil processes such as organic carbon dynamics, takes longer to reorganize. The timing and extent of these delayed responses were governed primarily by annual precipitation intensity, frequency and monthly precipitation totals, meaning that wetter years with more frequent extremes push both faster and larger shifts in the vegetation.

Ecological time lags are increasingly recognized as a defining feature of how ecosystems respond to climate change, and the Poyang Lake results add an important wetland case. If plant diversity responds to rainfall with a delay, then today’s diversity is partly a memory of past extremes, and management decisions based on current conditions alone will systematically miss the coming changes. The study suggests that the seasonality and intensity of precipitation, rather than simply the total annual amount, act as the pacing mechanisms for wetland vegetation, which gives forecasters a potential handle on anticipating future vegetation shifts.

Across every analysis, one factor kept asserting itself: extreme precipitation remained the core driver of vegetation diversity patterns, with its influence expressed through the mediating roles of water level at the landscape scale and soil organic carbon at the community scale. This consistency, set against the backdrop of intensifying hydrological extremes in the Poyang Lake Basin documented by previous research, paints a coherent picture of a wetland whose biodiversity is being continuously reorganized by its water regime. The authors note that their findings offer valuable insights for the conservation and restoration of wetland vegetation diversity and are expected to provide a scientific foundation for the sustainable management of the lake ecosystem.

For the broader community of wetland scientists and restoration practitioners, the study is a demonstration of how combining multi-source remote sensing with machine learning and structural equation modelling can illuminate mechanisms that field surveys alone cannot easily resolve across a vast and dynamic floodplain. As climate change intensifies both floods and droughts around the world, the lessons from Poyang Lake are likely to travel: protect the hydrological regimes that sustain habitat mosaics, monitor soil conditions that mediate community change, and plan for the delays, because the vegetation of tomorrow’s wetlands is already being written by today’s weather.

Subject of Research: Multi-scale effects of extreme precipitation on wetland vegetation diversity at Poyang Lake, China

Article Title: Multi-scale effects of extreme precipitation on the vegetation diversity patterns in Poyang Lake, China

Article References: Ding, Y., Chen, W., & Chen, Z. (2026). Multi-scale effects of extreme precipitation on the vegetation diversity patterns in Poyang Lake, China. Plant and Soil. https://doi.org/10.1007/s11104-026-09009-1

Image Credits: AI Generated

DOI: 10.1007/s11104-026-09009-1

Keywords: extreme precipitation, Poyang Lake, wetland vegetation, biodiversity, machine learning, remote sensing, time-lag effects, soil organic carbon, water level, floodplain grasslands, PLS-SEM, ecological restoration