The research, led by Thomas Farges of the French Alternative Energies and Atomic Energy Commission (CEA) together with colleagues at the Université Paris Cité and the Institut de Physique du Globe de Paris, focused on travelling ionospheric disturbances, or TIDs. These are wave-like fluctuations in plasma density that propagate through the ionosphere, the shell of charged particles that begins roughly 60 kilometres above the ground and extends far into space. TIDs come in several flavours: large-scale disturbances race at 300 to 1,000 metres per second with wavelengths of 1,000 to 3,000 kilometres and are usually driven by auroral activity during geomagnetic storms, while medium-scale TIDs move more slowly, at 100 to 300 metres per second, with wavelengths of several hundred kilometres. The latter are typically triggered by events near the Earth’s surface, including volcanic eruptions, earthquakes, tsunamis, rocket launches and, crucially, severe weather.

The physical mechanism linking a thunderstorm to the ionosphere is the atmospheric gravity wave. In the most convective regions of a storm, where updrafts punch through the top of the cloud in what meteorologists call an overshooting plume, air is violently displaced. This generates gravity waves that emerge from the cloud tops and propagate upward, generally at an oblique angle, through the stratosphere and mesosphere before reaching the thermosphere and modulating the plasma there. Because the waves travel obliquely rather than straight up, the ionospheric signature appears hundreds of kilometres away from the storm itself, spreading outward in near-circular ripples, much like the rings around a stone thrown into a pond.

To catch these ripples, the team exploited total electron content, or TEC, a measure of the number of electrons along the line of sight between a GNSS satellite and a ground receiver. By combining signals at two carrier frequencies, the researchers computed TEC for each satellite-receiver pair, discarded low-elevation measurements to avoid tropospheric contamination, and then subtracted a slow polynomial trend and applied a band-pass filter covering periods between roughly 6 and 45 minutes, the window where gravity-wave signatures live. The geographic position of each measurement was mapped to the so-called ionospheric piercing point, the intersection of the signal path with a thin shell at 250 kilometres altitude, allowing thousands of individual time series to be stitched into two-dimensional maps of the perturbed ionosphere.

The five storms analysed were deliberately diverse. They included a derecho that swept from the Mediterranean across northern Italy on 18 August 2022, a stationary Mediterranean storm that stalled over the Cevennes in southern France on 19 September 2020, a sprite-producing mesoscale convective system over the Mediterranean on 21 September 2019, the remnants of ex-tropical storm Kirk that crossed France on 8 October 2024, and the devastating cold-drop storm that flooded the Valencia region of southeastern Spain on 29 October 2024, killing more than 200 people. Storm candidates were preselected using meteorological reports and then filtered with criteria established in earlier work: the storm’s radar reflectivity above 30 dBZ had to cover more than 25,000 square kilometres, and lightning activity, measured by the World-Wide Lightning Location Network, had to exceed three flashes per ten minutes within half-degree grid cells inside the convective zone.

In every case, the ionosphere responded. Quasi-periodic TIDs propagated nearly circularly away from the storm centres, remaining visible in TEC for many hours and reaching horizontal distances of up to 1,000 kilometres. The apparent horizontal phase speeds fell between 100 and 160 metres per second, and the wave periods ranged from 13 to 39 minutes, values fully consistent with earlier observations in the Americas and Asia and with theoretical dispersion relations for gravity waves. For three of the five storms, the temporal evolution of the wave trains could be mapped clearly enough to measure wavelengths of roughly 200 to 300 kilometres, and the radiation pattern proved distinctly non-isotropic: waves were stronger and better defined in some directions than in others, exactly as atmospheric models predict.

The derecho case illustrates the power of the technique. As the storm formed north of the Balearic Islands around 04 UTC, climbed the Gulf of Genoa and slammed into northern Italy, its convective area swelled to about 70,000 square kilometres. Hodochrones, time-distance diagrams built from roughly 130 receivers, showed wavefronts marching away from the source at an average phase speed of 152 metres per second, with local values ranging from 120 to 230 metres per second. Two-dimensional TEC maps revealed circular ripples detectable from 600 to 1,000 kilometres from the storm, with amplitudes of about 0.3 TECU. A two-dimensional Fourier transform of an averaged TEC map yielded a wavelength of 193 kilometres and a preferred propagation azimuth toward the northeast, matching the 180 kilometres measured by eye between successive ripples. Intriguingly, waves travelling southeast carried amplitudes near 0.15 TECU and periods near 30 minutes, while those heading northeast were stronger, around 0.35 TECU, with shorter periods near 20 minutes, a directional asymmetry that mirrors the storm’s own motion.

The asymmetry itself carries scientific weight. Stratospheric and thermospheric winds act as filters, suppressing gravity-wave propagation in the direction the wind blows from, and the direction of the prevailing stratospheric wind reverses with the seasons. Four of the five storms occurred after the autumnal equinox, when the winds at around 65 kilometres altitude turn easterly, and their TIDs propagated preferentially westward. The derecho, the sole summer event, broke the pattern. The researchers argue that this seasonal dependence, long predicted by numerical simulations, is now confirmed observationally over Europe, and that it provides an independent fingerprint tying the ionospheric waves to their tropospheric sources.

By overlaying circles fitted to the observed wavefronts onto radar reflectivity maps, the team also pinned down the waves’ birthplace. The best agreement between the circular patterns and the most convective storm cores, defined by reflectivity above 45 dBZ, occurred when the radar data were taken roughly two to two and a half hours before the ionospheric measurement, a lag consistent with the oblique propagation speed of gravity waves. Because no disturbances were detected directly above the storms, only beyond about 600 kilometres, the observations imply a minimum propagation angle of about 20 degrees from the vertical, a hallmark of atmospheric gravity waves. For the sprite-producing storm and the Cevennes event, only a small arc of the wavefront was visible to the southwest, an asymmetry the authors attribute partly to wind filtering and partly to the geomagnetic field, which impedes the motion of ions and electrons perpendicular to field lines.

The study also sounds a note of caution for anyone hoping to use the ionosphere as a storm detector. During the Valencia flood, background TIDs of unrelated origin obscured the storm’s signature during the first peak of lightning activity, and the signal only became clear during the second peak. Geomagnetic storms, such as the one in its recovery phase when ex-Kirk arrived, can generate TIDs of their own that must be carefully separated from thunderstorm waves by analysing both speed and direction. The solar terminator, the moving boundary between day and night, launches wave packets that can mimic storm-generated disturbances; during the Cevennes event, terminator-driven waves showed amplitudes and speeds uncomfortably similar to gravity waves, and the authors note that distinguishing the two requires checking whether the propagation direction deviates by more than about 30 degrees from the terminator line or whether the speed falls below 200 metres per second. Solar flares, by contrast, produce an almost instantaneous, dayside-wide TEC jump that is easy to tell apart. The researchers conclude that, at this stage, the type of thunderstorm cannot be identified from TID characteristics alone, and they warn that intense storm activity could even complicate the ionospheric detection of tsunamis, which generate gravity waves with strikingly similar parameters. Still, with Europe’s GNSS networks now dense enough to image these ripples in real time, thunderstorms have been added to the growing list of surface events, from earthquakes to rocket launches, whose fingerprints can be read in the electrically charged sky above.

Subject of Research: Thunderstorm-generated travelling ionospheric disturbances observed with dense GNSS networks over Western Europe

Article Title: Characterization of thunderstorm-generated travelling ionospheric disturbances by dense GNSS sensors network over Western Europe

Article References: Farges, T., Princelle, E., & Astafyeva, E. (2026). Characterization of thunderstorm-generated travelling ionospheric disturbances by dense GNSS sensors network over Western Europe. Frontiers in Astronomy and Space Sciences, 13, Article 1893553. https://doi.org/10.3389/fspas.2026.1893553

Image Credits: AI Generated

DOI: 10.3389/fspas.2026.1893553

Keywords: travelling ionospheric disturbances, thunderstorms, atmospheric gravity waves, GNSS, total electron content, ionosphere, derecho, sprites, lightning, Western Europe, space weather, solar terminator

The research draws on data from ATDnet, the Met Office’s very low frequency lightning location system, which detects the electromagnetic sferics produced mainly by cloud-to-ground return strokes as they propagate through the Earth-ionosphere waveguide. Rather than counting flashes, ATDnet records individual strokes, the discrete current pulses that make up a complete flash, and this distinction matters throughout the analysis. The team mined a dataset covering the region between 13 degrees West and 5 degrees East and 49 to 61 degrees North from 2008 to 2016, and selected two days of exceptional lightning activity as case studies: 28 June 2012, when three supercell thunderstorms swept across the English Midlands, and 1 July 2015, when three structurally distinct severe storms crossed northern England.

Assigning individual strokes to individual storms is far from trivial, and the researchers leaned heavily on published synoptic analyses of both days. For the 2012 event, they used Doppler radar snapshots from the Clee Hill and Ingham stations, along with observations of a tornado near Sleaford and accompanying hail, to carve out the strokes belonging to each of the three supercells. The 2015 case relied on radar stations at Hameldon Hill, High Moorsley and Ingham. The result was six distinct stroke datasets: the 2012 storms produced 7,955, 11,988 and 5,655 strokes respectively, while the 2015 storms yielded 4,218, 455 and 1,926. Five of the six storms were classified as supercells, with the remaining one a single-cell thunderstorm, giving the team a rare opportunity to compare lightning behaviour across storm types.

With the strokes assigned, the researchers characterised three physical variables for each storm. The first is the inter-event time, the gap between successive strokes, which ranged across four orders of magnitude, from below 0.01 seconds to more than 100 seconds. Crucially, the distributions turned out to be bimodal, with a shorter peak around 0.2 seconds and a longer peak between 1 and 10 seconds. The team interprets this structure physically: the short peak reflects inter-stroke intervals, the time between successive current pulses within a single flash, while the longer peak corresponds to inter-flash periods tied to the charge-replenishment processes inside the storm. The short-peak values align well with high-speed camera studies from Brazil, the eastern Mediterranean and the United States, which found mean inter-stroke intervals of roughly 83 to 98 milliseconds.

The second variable is the movement speed of the stroke cluster, which serves as a proxy for the speed of the thunderstorm itself. Here the researchers applied two complementary least-squares techniques. The first fits a plane directly to the spatio-temporal coordinates of the strokes, with the slope of that plane encoding the storm’s velocity. The second transforms the data into spatio-temporal lag space, where the spatial and temporal separations between every pair of strokes are compared, and fits a line through the origin. Both methods agreed closely: the 2012 supercells moved at 13 to 17 metres per second, while the faster 2015 storms reached 18 to 31 metres per second, equivalent to up to about 108 kilometres per hour. These values sit comfortably within the ranges reported by radar-based thunderstorm climatologies over Central Europe and North Dakota.

The lag-space analysis also exposed hidden complexity within the storms. Bandings of pixels in the two-dimensional count plots revealed sub-clusters of strokes moving at similar speeds but offset from the main cluster, and other bandings moving at different speeds altogether, evidence that a single thunderstorm can behave as a collection of smaller cells travelling together. The authors treat this internal structure as a limitation on the precision of their bulk speed estimates rather than a flaw in the method, noting that the plane-fit deliberately returns a single representative speed for the storm’s dominant motion, which is exactly what their downstream model requires.

The third variable is spatial spread: how far strokes land from the storm’s track. To estimate the track, the team applied a uniform linear filter to the easting and northing coordinates of the strokes ordered by occurrence, then calculated the distances of each stroke from that smoothed path. Typical spreads reached up to 80 kilometres in either the easting or northing dimension, with one outlier at 226 kilometres. The distribution of these distances proved non-trivial, elongated diagonally in a way that reflects the orientation of each storm’s track relative to the geographic coordinate axes. Fitting a bivariate Gaussian distribution to the data, using the median and interquartile range to blunt the influence of outliers, captured the overall extent well, though it smoothed away the finer structure of the empirical distribution.

These characterisations feed directly into the study’s centrepiece: a synthetic lightning stroke model. The model treats a thunderstorm as a moving point source and generates point events along its path. A user sets an initiation point, a constant movement speed and a direction, then samples inter-event times from an empirical cumulative distribution and spatial offsets from the fitted bivariate Gaussian. A sorting and re-indexing step, validated by autocorrelation analysis, ensures the synthetic sequence reproduces the temporal structure of real storms, which the study showed to be significantly autocorrelated across all six cases, from short to long range. Demonstrations include a single run mimicking one 2012 supercell and a full synthetic three-storm system replicating the entire first case study, with parameters adjustable to explore how spatial spread alone reshapes the pattern of strikes.

The framework is deliberately simple, and the authors are candid about its limits. Real thunderstorms accelerate, decelerate and change direction, while the model assumes straight-line propagation at constant speed, an assumption that introduces growing positional error over long runs. The inter-event time data convolve two physically distinct processes, and the combined supercell sampling distribution is dominated by the larger 2012 dataset, giving synthetic storms a heavier tail that lets them persist longer than their real-world counterparts. Even so, the model’s transparency and controllability are its strengths: every parameter has a clear empirical basis and a clear procedure, making the framework replicable and extensible.

The implications reach beyond lightning. The authors frame their approach as a template for any natural hazard that can be characterised as point events produced by a moving source, from tornado touchdown locations to wildfire ignitions, and they identify storm-relative coordinate systems, non-linear track modelling and additional storm categories as priorities for future work. For hazard assessment, the ability to generate statistically faithful synthetic lightning datasets offers something observational records cannot: unlimited, fully controlled scenarios for testing forecasting algorithms, validating clustering methods and stress-testing infrastructure risk models. In an era when severe convective storms are a growing concern across Europe, turning the chaotic flicker of lightning into rigorous statistics may prove one of the most practical ways to see a storm coming.

Subject of Research: Statistical characterisation and spatio-temporal modelling of lightning strokes from individual UK thunderstorms

Article Title: Characterisation and modelling of lightning strokes as point events in time and space

Article References: Zandovskis, U., Pigoli, D., & Malamud, B. D. (2026). Characterisation and modelling of lightning strokes as point events in time and space. Natural Hazards and Earth System Sciences, 26(10), 4997-5033. https://doi.org/10.5194/nhess-26-4997-2026

Image Credits: AI Generated

DOI: 10.5194/nhess-26-4997-2026

Keywords: lightning, thunderstorms, supercells, ATDnet, spatio-temporal point processes, inter-event times, storm movement speed, spatial spread, synthetic modelling, natural hazards, UK Met Office, severe weather

The work, carried out by Diego Orozco-Gomez of Institución Universitaria Pascual Bravo together with Elisa Rodriguez and Freddy Bolanos of Universidad Nacional de Colombia in Medellín, proposes two distinct methods for synthesizing lightning channels based on a geometric metric called tortuosity. Tortuosity, in essence, quantifies how winding and contorted a path is, and it is precisely this quality that gives lightning its characteristic jagged, branching appearance in photographs. By controlling the tortuosity of generated paths, the researchers can produce synthetic images whose channels statistically resemble real lightning, providing artificial intelligence algorithms and heuristic systems with a rich, controllable supply of training examples without waiting for the skies to cooperate.

The first of the two proposed methods relies on a hyper-heuristic algorithm. Hyper-heuristics are a class of automated decision-making approaches that, rather than applying a single fixed problem-solving procedure, select and combine lower-level heuristics dynamically as the search unfolds. In this application, the hyper-heuristic explores different strategies for building a lightning-like path segment by segment, adjusting its behavior to steer the resulting curve toward a target tortuosity. This makes the approach flexible: because the high-level controller learns which low-level heuristics work best in which situations, it can adapt to a wide range of target path shapes without being re-engineered by hand for each new case.

The second method is a more direct heuristic algorithm designed specifically for the task of drawing tortuous lightning paths. Instead of delegating decisions to an adaptive controller, this algorithm encodes domain-informed rules about how a lightning channel should grow, producing paths that match a desired tortuosity in a single, streamlined pass. The trade-off is typical of heuristic design: less generality, but far greater speed and simplicity. When the two approaches were put head-to-head, the specialized heuristic algorithm demonstrated superior performance, achieving a mean accuracy of approximately 99 percent in hitting its tortuosity targets and an average execution time of just 7 milliseconds per generated path.

That execution time deserves emphasis. Seven milliseconds per path means the algorithm can generate hundreds of synthetic lightning channels in the time it takes a photographer to capture a single storm frame, making it practical to construct large, diverse training sets on demand. For deep learning pipelines, where models such as convolutional networks and YOLO-style object detectors typically require thousands of annotated examples to generalize well, the ability to mass-produce realistic, correctly labeled synthetic images could dramatically lower the barrier to building reliable lightning recognition systems. Synthetic data also allows researchers to systematically vary properties such as path complexity, branching, and orientation, covering corners of the parameter space that real observations rarely sample.

Crucially, the team did not stop at generating plausible-looking curves; they validated their synthetic paths against reality. The paths produced by the heuristic algorithm were evaluated against 54 real lightning images, yielding an average Recall of 87.03 percent. Recall, in this context, measures how well the synthetic paths capture the features of genuine lightning channels that a recognition system would need to detect. A recall above 87 percent against real photographic evidence suggests that the synthetic images are not merely visually convincing but statistically and structurally close enough to authentic strikes to serve as meaningful training material, which is the acid test for any synthetic data pipeline.

The study builds on the authors’ earlier research into lightning image classification, including a prior model for classifying lightning images without branches based on the tortuosity metric. That lineage matters, because it grounds the new synthesis methods in a quantitative understanding of what real lightning channels look like in digital images. It also connects to a broader research landscape: recent work has applied deep learning to high-speed lightning footage for semantic segmentation, classified single versus multiple ground strike points from video, and developed knowledge distillation algorithms for lightning detection networks. Meanwhile, simulation studies of lightning discharges have explored how environmental conditions influence ionization and spark behavior, and computational models of electromagnetic fields from cloud-to-ground and cloud-to-cloud channels continue to inform lightning protection engineering.

The motivation for all of this automation is sobering. Lightning is a documented trigger of industrial accidents, including fires and explosions at facilities handling hazardous materials, and it remains a significant cause of weather-related fatalities worldwide, with national fatality studies tracking its human toll across decades. Three-dimensional direct strike models for substations, risk assessment frameworks for lightning protection systems, and damage probability models for process equipment all depend on understanding where and how strikes occur. Automatic visual detection of lightning, powered by well-trained machine learning models, adds a complementary layer of capability, one that could support early warning, infrastructure monitoring, and post-storm damage assessment in ways that ground-based sensor networks alone cannot.

What makes the Colombian team’s contribution particularly elegant is its conceptual reframing of the problem. Rather than treating lightning image generation as a pure computer graphics challenge or a pure machine learning challenge, they treat it as a path planning and optimization problem, borrowing tools from fields such as robotics and operations research. The bibliography of the paper spans classic algorithmic work such as Bresenham’s 1965 line-drawing algorithm, random walk methods rooted in the theory of Brownian motion, sampling-based path planners like RRT variants, and modern metaheuristics including differential evolution, particle swarm optimization, and simulated annealing. By channeling this diverse algorithmic arsenal toward a meteorological imaging problem, the authors demonstrate how cross-disciplinary method transfer can solve practical data scarcity problems in earth science.

The practical implications extend well beyond lightning. The same tortuosity-controlled synthesis framework could plausibly be adapted to generate training data for other crack-like, branching, or filamentary structures, from electrical treeing in insulating materials to fracture patterns in materials science. And because the researchers have made the data generated in the work publicly available through Science Data Bank, other groups can immediately build on the synthetic dataset, benchmark their own generation algorithms against the reported 99 percent accuracy and 7 millisecond generation time, and test whether models trained on these synthetic bolts transfer to real-world storm footage. As deep learning continues to penetrate atmospheric observation, studies like this one highlight an increasingly important truth: sometimes the fastest way to teach a machine about nature is to manufacture convincing imitations of it, one jagged, seven-millisecond lightning path at a time.

Subject of Research: Synthetic generation of tortuous lightning path images to train automatic lightning recognition models

Article Title: Generating synthetic lightning path images for training lightning recognition models

Article References: Orozco-Gomez, D., Rodriguez, E., & Bolanos, F. (2026). Generating synthetic lightning path images for training lightning recognition models. Earth Science Informatics, 19(12), Article 212. https://doi.org/10.1007/s12145-026-02267-8

Image Credits: AI Generated

DOI: 10.1007/s12145-026-02267-8

Keywords: lightning, synthetic images, machine learning, heuristic algorithm, hyper-heuristic, tortuosity, deep learning, path planning, training data, lightning detection, Earth Science Informatics, Colombia

A new study published in Atmospheric Chemistry and Physics by Francisco J. Pérez-Invernón of the Instituto de Astrofísica de Andalucía in Granada and an international team of collaborators has now tackled that problem with an unusually powerful combination of instruments. The researchers focused on the Ebro Valley in northeastern Spain, a region of notably high lightning activity where airflows shaped by the Mediterranean Sea, the Atlantic Ocean, the Pyrenees and the Iberian thermal low frequently converge to spark thunderstorms. By pairing optical lightning observations from the Meteosat 12 Lightning Imager, the first lightning detector of Europe’s Meteosat Third Generation fleet in geostationary orbit, with ground-based radio measurements from a 24-station Lightning Mapping Array, the team was able to characterize individual flashes in unprecedented detail and connect their properties to the amount of NOx they generate.

The technical challenge is formidable. Satellites such as the TROPOspheric Monitoring Instrument, TROPOMI, flying on the Sentinel-5 Precursor spacecraft, can measure nitrogen dioxide columns from orbit with a horizontal resolution of about 3.5 by 5.5 kilometers at nadir. But the NOx they see over a storm is a mixture: some of it was freshly injected by lightning, some was lofted from the boundary layer, some arrived from distant sources, and some is stratospheric. Disentangling the lightning contribution requires knowing both the vertical distribution of the lightning-produced gases and the background against which they appear. The team addressed this by running high-resolution simulations with the WRF-Chem atmospheric model, nudged toward the actual lightning flashes observed by Meteosat 12, so that the model reproduced where and when the storms deposited their chemical payload. Comparing paired simulations with and without fresh lightning emissions allowed the researchers to calculate the fraction of the measured column attributable to lightning in each pixel, as well as the air mass factors needed to convert measured slant columns into vertical column densities.

The lightning observations themselves came from two complementary systems. The Lightning Imager aboard Meteosat 12, launched in December 2022 and declared operational in late 2024, stares continuously at the full Earth disk from geostationary orbit, detecting the 777-nanometer optical emission of atomic oxygen in lightning discharges with a CMOS detector sampling at one kilohertz. It records for each flash a suite of properties: the illuminated footprint, the total radiance, the duration, and the temporal evolution of the optical signal. On the ground, the recently upgraded eXtended Catalonia Lightning Mapping Array, XCALMA, tracks the very high frequency radio impulses emitted by stepped leaders as they carve lightning channels through the cloud, reconstructing the three-dimensional geometry of each flash without any distortion from intervening cloud layers. From XCALMA data the team extracted total flash channel length, flash volume and flash duration, quantities that no space-based optical sensor can measure directly.

Between July 2024 and May 2025, the researchers identified eighteen thunderstorm days that satisfied strict quality criteria, including at least 650 flashes detected by the Lightning Imager within the well-covered region of the mapping array and a mean flash age of less than 3.5 hours before the TROPOMI overpass, which occurs at roughly 13:30 local solar time. For each storm they computed the lightning NOx production efficiency, defined as the moles of NOx produced per flash, using a retrieval that accounts for the detection efficiency of each lightning system and the exponential decay of NOx during the hours between flash occurrence and satellite observation. The results were striking in their internal consistency: using Meteosat 12 flashes with an assumed five-hour NOx lifetime yielded 37 plus or minus 25 moles of NOx per flash, while using the ground-based mapping array yielded 31 plus or minus 24 moles per flash, with slightly higher values of 41 and 38 moles when a three-hour lifetime was assumed.

The heart of the study lies in how these production efficiencies relate to the physical characteristics of the flashes themselves. Across the eighteen storms, the team found consistent positive correlations between lightning-produced NOx per flash and nearly every flash property they examined. Longer flash channels produced more NOx, as expected from laboratory and modeling work showing that production scales with the volume of air heated by the discharge. Larger flash volumes, longer flash durations, brighter optical radiances and bigger illuminated footprints all correlated positively with per-flash production. The strongest linear relationships emerged when the ground-based channel length, volume and duration measurements were used, with the mapping array metrics proving to be better proxies for NOx production than the satellite optical properties, likely because cloud cover attenuates and distorts the light that reaches geostationary orbit.

One of the most intriguing findings concerns continuing currents, the slow transfer of charge that can persist for tens to hundreds of milliseconds after a cloud-to-ground stroke. Earlier work on rocket-triggered lightning suggested that these slow discharge processes account for roughly eighty percent of the total NOx produced per flash, but confirming this in natural storms has been nearly impossible because continuing currents hide below cloud and are poorly detected by optical sensors from space. The team devised new variables from the Lightning Imager data, counting the number of time-contiguous optical groups per flash and the fraction of flashes showing at least five milliseconds of uninterrupted optical signal. These continuity measures correlated positively, though weakly, with NOx production, and less strongly than total flash duration measured on the ground. The gap between the two correlations hints that significant lightning activity below the cloud, invisible to the satellite, contributes to the chemical budget, and that radio-frequency mapping remains essential for a complete picture.

The study also clarified how the detection efficiency of the new geostationary instrument behaves. The mapping array detected between 1.3 and 6.5 times more flashes than Meteosat 12 over the same storms, and the ratio of the two counts grew for more active thunderstorms, confirming that the satellite’s detection efficiency drops as flash rates rise. The satellite performed relatively better for brighter flashes, particularly cloud-to-ground flashes, which the analysis showed tend to be longer, larger and more luminous than intracloud flashes. Flashes initiating at lower altitudes, where cloud-to-ground activity predominates, were also more extensive and more energetic. These calibration insights matter beyond Spain: as the global constellation of geostationary lightning mappers expands, understanding their biases is essential for anyone using their data to estimate lightning’s chemical or climatic influence.

The broader implications reach into climate modeling and air quality forecasting. Lightning-produced NOx is roughly six times more efficient at driving ozone production than anthropogenic NOx, generating about a hundred ozone molecules per NOx molecule emitted, and it influences the tropospheric budgets of methane and carbon monoxide. Because climate change is expected to alter thunderstorm frequency and intensity, models need accurate, flash-property-aware parameterizations of lightning emissions. This study moves the field toward that goal by demonstrating that flash geometry, duration and brightness, measurable from space, carry real information about per-flash chemistry. The team estimates that their refinements, including lightning-nudged simulations and the upgraded mapping array, reduce the total uncertainty of satellite-based production estimates from the previously reported range of twenty to fifty percent down to roughly twelve to forty percent. With the upcoming Sentinel-4 spectrometer and the Meteosat Third Generation Infrared Sounder set to provide coincident air composition and humidity profiles over Europe, the Ebro Valley is poised to remain a natural laboratory where the chemistry of thunderstorms is finally being written down, flash by flash.

Subject of Research: Lightning-produced nitrogen oxide emissions measured with satellite and ground-based lightning observations over the Ebro Valley, Spain

Article Title: Characterizing lightning NOx production in the Ebro Valley with meteosat 12 LI and LMA observations

Article References: Pérez-Invernón, F. J., Gordillo-Vázquez, F. J., van der Velde, O., Montanyà, J., López Trujillo, J. A., Pineda, N., Loyola, D., Seo, S., Knote, C., Fierro, A. O., & Huntrieser, H. (2026). Characterizing lightning NO x production in the Ebro Valley with meteosat 12 LI and LMA observations. Atmospheric Chemistry and Physics, 26(19), 14277-14301. https://doi.org/10.5194/acp-26-14277-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-14277-2026

Keywords: lightning, nitrogen oxides, NOx, Meteosat 12, Lightning Imager, Lightning Mapping Array, TROPOMI, WRF-Chem, thunderstorms, atmospheric chemistry, ozone, Ebro Valley

The research team assembled an unusually rich observational record for a single storm. They drew on ERA5 reanalysis fields from the European Centre for Medium-Range Weather Forecasts, half-hourly imagery from the INSAT-3DR satellite, radiosonde profiles launched from Patna, surface winds from 54 automatic weather stations quality-controlled with range and spike checks, cloud-to-ground and intracloud flash records from the Indian Lightning Location Network, and 24-hour rainfall totals from 472 rain gauges interpolated by ordinary kriging. Casualty data from Bihar’s disaster management department were validated through field surveys within 48 hours of the event, with lightning fatalities attributed using a strict proximity criterion of 500 metres and 30 minutes against the lightning network’s flash records. This multi-platform approach allowed the authors to trace the storm’s full lifecycle from initiation to dissipation with a precision rarely possible in this region.

The satellite record reveals a dramatic two-act structure. A first convective cell formed over north-central Uttar Pradesh at around 0300 UTC, its longevity apparently sustained by orographic influences from Uttarakhand’s hills, strong daytime surface heating and a steady influx of moisture from the Bay of Bengal. A second cell developed independently over the northern Chhattisgarh–Jharkhand border at roughly 0700 UTC, in an environment of intense surface heating, low-level convergence and orographic lifting along the plateau margins. By 0715 UTC the two convective leading edges were converging over west-central Bihar, and at approximately 0915 UTC they merged over the state’s central districts, coinciding with a sharp intensification of lightning across Bhojpur, Arwal, Jehanabad, Patna, Nalanda and Sheikhpura.

After the merger, the system took on an elongated, arc-like form that the authors describe as bow-echo-like, a configuration classically associated with rear-inflow jets, embedded vortices and damaging straight-line winds. The evidence is a sharp horizontal gradient in cloud-top brightness temperature, reaching 15 to 20 kelvin per pixel between 0945 and 1045 UTC, precisely when cloud-to-ground flash counts peaked at 2615 to 2871 strikes per 30-minute window and automatic weather stations recorded gusts exceeding 85 kilometres per hour. Cloud tops climbed to 14 to 17 kilometres, with cloud-top pressure plunging below 300 hectopascals over the Nalanda–Patna–Bhojpur corridor. The authors are careful to note that the Patna S-band Doppler radar was non-operational during the event, so the bow-echo classification rests on satellite and lightning proxies alone and cannot be kinematically verified. Even so, the coincidence of the arcing thermal boundary, overshooting tops and concentrated wind damage near the gradient apex over Nalanda is striking.

The thermodynamic setup was equally decisive. Hovmöller diagnostics along the storm-affected corridor show that between roughly 0600 and 1000 UTC all three formulations of convective available potential energy—surface-based, mixed-layer and most-unstable—intensified dramatically, exceeding 1000 joules per kilogram with a concentrated core above 1500 joules per kilogram near 24.5 to 25 degrees north. Crucially, the difference between skin temperature and two-metre air temperature was strongly positive during the morning, indicating that the land surface was warming well above the overlying air. This radiative heating eroded the convective inhibition that had capped the boundary layer, and by 0800 to 1000 UTC strong upward motion at 850 hectopascals, approaching minus 1.35 to minus 1.80 pascals per second, signalled organised low-level convergence feeding the developing convective column. Before 0600 UTC the vertical velocity field had been near-neutral or weakly subsident, underscoring that the available instability alone was not enough—surface heating was the trigger.

The Patna radiosonde profiles tell the same story in three acts. On 9 April the sounding showed substantial potential with surface-based CAPE of 2523.5 joules per kilogram, most-unstable CAPE of 3342.0 joules per kilogram and moderate capping of minus 209.1 joules per kilogram, together with a downdraft CAPE of 1579.7 joules per kilogram, well above the threshold for strong evaporatively driven downdrafts. On 10 April, the day of the storm, surface-based CAPE rose to its three-day maximum of 2780.9 joules per kilogram while convective inhibition fell sharply to minus 91.0 joules per kilogram, and the Bulk Richardson Number of 41.75 fell within the range conventionally associated with organised multicellular convection. By 11 April the atmosphere had stabilised almost completely, with surface-based CAPE collapsing to 126.8 joules per kilogram and a strongly capped profile, while the wind profile backed rather than veered, consistent with post-convective cold-pool outflow.

Synoptic-scale features provided the supporting scaffolding. A low-pressure area over the west-central Bay of Bengal, an upper-air cyclonic circulation over north-east Bihar, smaller circulations over north-west Madhya Pradesh and south Assam, and a western disturbance near 68 degrees east together steered moisture-laden south-easterly winds into convergence with inland westerlies. Deep-layer wind shear of 15 to 20 metres per second between 850 and 500 hectopascals displaced precipitation-driven downdrafts from their updrafts, allowing convection to persist rather than being undercut by rain-cooled outflow. At 200 hectopascals, strong anticyclonic outflow and upper-level divergence vented the storms’ exhaust, while mid-tropospheric temperatures of minus 8 to minus 12 degrees Celsius overlying a warm, moist boundary layer steepened lapse rates. Negative omega values of minus 0.3 to minus 0.4 pascals per second at 700 hectopascals confirmed large-scale ascent during peak activity.

The human toll mapped onto these physical signatures with sobering clarity. Wind fatalities concentrated heavily along the merged system’s track, with Nalanda alone accounting for 24 deaths, followed by Bhojpur with four, Patna with three, and Jehanabad and Gaya with two each. Lightning deaths, by contrast, were dispersed across the state, with Siwan and Jamui recording three apiece and a dozen other districts reporting one or two. Two lightning hotspots emerged in the flash-density maps: one over the eastern districts of Purnea, Madhepura and Katihar near the convergence of Bay of Bengal moisture and the Himalayan foothills, and another over the southern districts of Nawada, Sheikhpura and Nalanda near the Chhotanagpur plateau margin. Rainfall, though locally heavy—74.2 millimetres at Bhawanipur in Purnea and more than 55 millimetres at several other stations—proved a mixed blessing, waterlogging some harvest-ready crops while replenishing soil moisture for standing maize, mung and orchards.

The authors emphasise that socioeconomic vulnerability amplified the meteorology into a disaster. The storm struck in the late afternoon, when thousands of agricultural labourers were in open fields harvesting wheat and pulses, maximising exposure to both lightning and destructive winds. Bihar’s population density of over 1,100 people per square kilometre, its reliance on fragile mud-and-thatch dwellings, and limited access to sturdy shelter in rural areas meant that buildings were simply not engineered to withstand winds of this magnitude. Generic district-level alerts proved insufficient given the sharply differing hazard footprints of wind and lightning, and the authors argue for hazard-specific, impact-disaggregated warnings that incorporate population density, agricultural calendars and shelter availability, delivered through mobile networks, community radio and trained local volunteers.

The study is candid about its limits: without radar, the internal structure of the merged system remains inferred; ERA5 fields represent grid-scale averages that may miss storm-scale processes; and the single Patna sounding may not represent conditions across all affected districts. Yet the diagnostic chain it documents—moisture transport, CAPE build-up, inhibition erosion, cell merger and bow-echo-like organisation—offers a replicable template for nowcasting severe pre-monsoon convection across the Indo-Gangetic Plain. With denser surface and radar networks, real-time lightning assimilation into numerical models, and machine-learning approaches trained on both atmospheric and impact data, the authors argue that the atmospheric precursors of such disasters could be translated into probabilistic, life-saving warnings for South Asia’s most convectively active and densely populated regions.

Subject of Research: Dynamical and thermodynamic mechanisms of the severe April 2025 thunderstorm over eastern India

Article Title: Dynamical and thermodynamic mechanisms driving the April 2025 severe thunderstorm over eastern India

Article References: Shankar, A., Kant, S., & Kamal, L. (2026). Dynamical and thermodynamic mechanisms driving the April 2025 severe thunderstorm over eastern India. Discover Geoscience, 4(1), Article 394. https://doi.org/10.1007/s44288-026-00764-3

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00764-3

Keywords: severe thunderstorm, Bihar, bow echo, CAPE, lightning, mesoscale convective system, cell merger, ERA5 reanalysis, INSAT-3DR, wind shear, impact-based early warning, Nor'wester

The story begins with the physics of cosmic rays themselves. Most primary cosmic rays are protons, electrically charged particles whose trajectories are bent by Earth’s magnetic field before they ever reach the atmosphere. When they arrive, they collide with atomic nuclei in the air and generate cascades of secondary particles. Among these secondaries are muons, short-lived charged particles produced at high energies that travel at nearly the speed of light and penetrate deep into the atmosphere. As Sunil Gupta of the Tata Institute of Fundamental Research (TIFR) in Mumbai, one of the authors of the study, explains, the cosmic rays detected at ground level consist predominantly of these highly penetrating muons. The GRAPES-3 facility includes a 560-square-metre muon telescope that records around four billion muons every day, providing an extraordinarily dense stream of data.

Crucially, those muons are not evenly balanced in electric charge. Because the primary cosmic rays are mostly positively charged, positive muons are more abundant in the atmosphere than negative ones. Physicists quantify this imbalance with the muon charge ratio, the number of positive muons divided by the number of negative ones, which normally exceeds one. That small excess is what allows GRAPES-3 to do something it was never originally designed to do. Inside a thundercloud, electric charges become separated, producing intense electric fields and large potential differences. When muons cross these regions, the field acts on positive and negative particles in opposite ways, generally slowing positive muons while accelerating negative ones. If the two charges were present in equal numbers, these effects would cancel almost perfectly, leaving no detectable change in the total muon flux.

Hari Haran Balakrishnan of TIFR, first author of the new study, puts the point starkly: if nature provided equal numbers of positive and negative muons throughout the telescope’s field of view, the thunderstorm phenomenon simply could not be observed with the current setup. Because positive muons dominate, however, each thunderstorm leaves a small but measurable imprint on the total muon count, and from that imprint researchers can infer the electrical potential inside the cloud. Gupta describes muons as an ideal gift for these kinds of studies, likening them to an electric current flowing through the atmosphere. In an earlier analysis, the team reported a striking result: a potential of 1.3 gigavolts measured in a single thunderstorm, a figure that dwarfs anything obtained by conventional means.

The idea that thunderstorms might generate potentials on that scale is not new. As early as the 1920s, the physicist Charles Thomson Rees Wilson, who would later win the 1927 Nobel Prize in Physics, suggested that thunderclouds could reach such enormous voltages. Direct measurements, however, had previously topped out at only around 130 million volts, more than an order of magnitude lower. The difficulty is practical as much as conceptual: sampling the interior of a large, rapidly evolving storm generally requires instruments carried by aircraft or weather balloons into or near active weather, a hazardous and spatially limited approach. Muons invert the problem entirely. They pass through the atmosphere continuously, they can be monitored from the safety of the ground, and they sample the whole volume of the cloud along their paths.

But the earlier GRAPES-3 observations carried an unexpected puzzle alongside their headline result. Between April 2011 and December 2020, the telescope detected 487 thunderstorm events, and their distribution across the sky was strikingly lopsided. Fully 81.5 percent of the events appeared in the eastern part of the instrument’s field of view, while only 13.7 percent appeared in the west, a difference of almost six to one. The obvious suspicion was local weather: perhaps storms in this part of India genuinely favour the east. An independent array of instruments that monitors the atmospheric electric field said otherwise. Actual thunderstorms, measured directly, showed no comparable preference for the eastern sky, which meant the asymmetry had to originate in the detector rather than in the atmosphere.

The answer lies in an effect that atmospheric and cosmic-ray physicists have known about for nearly a century: the east-west effect. Because cosmic-ray protons are charged, Earth’s magnetic field filters their arrival directions. Lower-energy positive particles arriving from the east are screened out more strongly than those arriving from the west, so the mix of cosmic rays, and therefore of secondary muons, depends on direction. This directional threshold, known as the geomagnetic cutoff, varies significantly across GRAPES-3’s field of view. The consequence is that the muon charge ratio, the very quantity that makes thunderstorm sensing possible in the first place, is itself direction-dependent, and the telescope is therefore more sensitive to thunderstorms in some parts of the sky than in others.

To test whether this could account for the observed lopsidedness, the researchers turned to computer simulations that model the production and propagation of muons through the atmosphere. The simulations showed that the balance between positive and negative muons changes markedly with direction: the muon charge ratio is about 1.37 in the east but only 1.14 in the west. That difference, they found, is more than enough to explain the event counts. As Gupta emphasizes, it is not that the voltage is different; thunderstorms to the east are not necessarily more electrically powerful than those to the west. Rather, the larger imbalance between positive and negative muons arriving from that direction makes the telescope more responsive to their electrical effects. The simulations confirmed the mechanism in both directions: when the direction-dependent charge ratio was included, a strong east-west asymmetry emerged, closely resembling the observations, and when the same ratio was imposed in both directions the asymmetry disappeared entirely.

The resolution of the puzzle matters well beyond bookkeeping for one detector. Understanding this directional sensitivity is essential if GRAPES-3 is to serve as a reliable, calibrated probe of thunderstorm electricity, because any inferred potential must account for how efficiently the instrument responds along each line of sight. With that correction in hand, the technique offers something no balloon campaign can match: continuous, ground-based monitoring of the largest electric fields in nature, drawn from a particle flux that arrives whether or not anyone is watching. The broader significance is considerable, since thunderstorm electrification is linked to lightning initiation, to terrestrial gamma-ray flashes, and to the still-uncertain role of storms in the global atmospheric electrical circuit. What began as an instrument for studying particles from the cosmos has ended up illuminating one of the most familiar and, remarkably, still one of the most mysterious phenomena on our own planet. The same muons that carry messages about distant astrophysical sources now carry, quite literally, the voltage of the storm overhead.

Subject of Research: Using cosmic-ray muons to probe electric fields and potentials inside thunderclouds

Article Title: From the cosmos to thunderstorms

Article References: From the cosmos to thunderstorms. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: cosmic rays, muons, thunderstorms, GRAPES-3, electric fields, geomagnetic field, muon charge ratio, atmospheric electricity, lightning, particle physics, TIFR, JCAP

The central question is deceptively simple. When the ground is hot and dry, does that help storms form, or does it suppress them? The answer, it turns out, depends heavily on the time of year and on which link in the chain you examine. The study finds strong and seasonally varying coupling between land surface temperature and atmospheric moisture, but the relationships between individual thermodynamic or dynamical variables and precipitation are generally weak and nonlinear. In other words, no single number measured at the surface or in the boundary layer can reliably predict rainfall on its own. The atmosphere over India behaves less like a machine with a single trigger and more like a system of conditional gates, several of which must open at once before convection fires.

To untangle these relationships, the research integrates an unusually broad set of observational streams. Land surface temperature came from the MODIS MOD11A2 product, evapotranspiration from MOD16A2, both distributed by NASA. Atmospheric variables, including dew point temperature, convective available potential energy (CAPE), convective inhibition (CIN), and wind fields, were drawn from the ERA5 reanalysis produced by the European Centre for Medium-Range Weather Forecasts. Precipitation estimates came from the TRMM 3B43 and GPM IMERG missions, lightning observations provided an independent proxy for actual convective activity, and land-use and land-cover information came from the India WRIS database. Nearly all of the processing was carried out on the Google Earth Engine platform, allowing pixel-level analysis across the entire subcontinent.

Two diagnostic frameworks sit at the heart of the analysis. The first is a normalized CAPE–CIN instability index, or NCII, which combines the fuel available to rising air parcels (CAPE) with the energetic barrier they must overcome (CIN) into a single measure of thermodynamic readiness. The second is the Convective Triggering Potential–Humidity Index framework, or CTP–HI, which classifies the atmospheric profile above a given location according to whether the boundary layer and free troposphere are configured in a way that favors or suppresses feedbacks between the land surface and convective initiation. Together, these tools allow the study to separate the question of whether the atmosphere is primed for storms from the question of whether storms actually occur.

The NCII results deliver one of the study’s most important cautionary messages. The index shows only weak-to-moderate correspondence with observed lightning activity, which serves as a direct tracer of deep convection. Thermodynamic readiness, in other words, does not necessarily coincide with convective occurrence. A region can be loaded with CAPE and still produce no storms if CIN is too strong, if moisture convergence is absent, or if no dynamical lifting mechanism nudges air parcels past their level of free convection. This finding echoes a long-standing theme in convective meteorology, rooted in ingredients-based methodologies for flash flood forecasting: instability is necessary but never sufficient. For forecasters, the implication is that instability indices alone will systematically misidentify both where storms will form and where they will not.

The CTP–HI analysis adds a spatial and seasonal dimension to that caution. The study finds substantial seasonal variation in the atmospheric thermodynamic environment, with conditions favoring a wet-soil advantage, in which moist land surfaces enhance the likelihood of afternoon convection, covering 20.52 percent of the valid analysis area during the monsoon months of June through September and 14.03 percent during the post-monsoon period of October and November. Strikingly, most of the domain remains outside the defined feedback regimes altogether, meaning that for the majority of India, the classical soil-moisture-precipitation feedback loop described in earlier benchmark studies simply does not apply most of the time. The land-atmosphere feedback story over India is therefore not a single narrative but a patchwork that shifts with the seasons.

Perhaps the most vivid result concerns evapotranspiration. When the study examines pixel-level correlations between evapotranspiration and the NCII instability index, the strength of the relationship swings dramatically across the year. It is strongest during October and November, with a correlation coefficient of 0.63, followed by the winter months of December through February at 0.47 and the pre-monsoon spring at 0.28. During the peak monsoon itself, the relationship actually flips sign, showing a weak negative correlation of minus 0.15. This seasonal modulation is physically interpretable: during the monsoon, when the large-scale circulation and abundant moisture supply dominate, the local land surface’s contribution of water vapor to the column becomes relatively unimportant, and heavy rainfall itself suppresses surface heating. In the transition seasons, by contrast, when large-scale forcing is weaker, evapotranspiration from vegetated, moisture-rich land emerges as a meaningful contributor to the spatial organization of instability.

That spatial organization is consistently tied to what covers the ground. The study finds that patterns of NCII are reliably associated with land-use and land-cover characteristics, with vegetated and moisture-rich regions generally exhibiting higher instability than arid or barren areas. Forests, irrigated agriculture, and wetlands effectively pump moisture into the boundary layer, lowering the lifting condensation level and sharpening local gradients in equivalent potential temperature that can focus convective initiation. This connection between vegetation and instability has implications that extend beyond academic interest, because land-use change across India, from deforestation to agricultural expansion to urbanization, may be quietly reshaping where the country’s storms prefer to form.

The broader significance of the work lies in the framework it endorses: a conditional, scale-dependent view of land-atmosphere interaction. Rather than asking whether soil moisture causes rainfall in some universal sense, the study argues that land-surface processes are associated primarily with the spatial organization of thermodynamic instability, while atmospheric moisture, stability, and dynamical forcing jointly determine whether convection and precipitation actually materialize. This division of labor resolves a persistent tension in the literature, where some studies have reported strong soil moisture-precipitation feedbacks and others have found them elusive. Both camps may be right, just in different seasons, regions, and atmospheric regimes.

For a country whose economy remains deeply sensitive to rainfall variability, the practical stakes are considerable. Understanding that the land surface matters most during the post-monsoon and winter transition seasons, and least during the height of the monsoon, could help refine the interpretation of high-resolution model forecasts and guide where improvements to land-surface initialization would yield the greatest benefit. The study also demonstrates the power of open, cloud-based Earth observation: by stitching together NASA satellites, European reanalysis, precipitation missions, and lightning networks within a single analytical environment, it shows how the pieces needed to decode one of the world’s most consequential weather systems are now freely available to any researcher with a laptop and a good question. As climate change continues to warm the troposphere and alter land-cover patterns across South Asia, the seasonal choreography of land and atmosphere documented here will not stay still, making this kind of process-based monitoring an essential tool for the decades ahead.

Subject of Research: Seasonal land–atmosphere coupling and convective instability over India

Article Title: Seasonal modulation of land–atmosphere coupling and convective instability over India: process-based insights into evapotranspiration controls and conditional convection

Article References: Dutta, D. (2026). Seasonal modulation of land–atmosphere coupling and convective instability over India: process-based insights into evapotranspiration controls and conditional convection. Climate Dynamics, 64(11), Article 442. https://doi.org/10.1007/s00382-026-08411-w

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08411-w

Keywords: land-atmosphere coupling, evapotranspiration, convective instability, CAPE, CIN, CTP-HI, Indian monsoon, lightning, MODIS, ERA5, soil moisture feedback, Climate Dynamics