The researchers turned their attention to a well-documented event from the Deep Propagating Gravity Wave Experiment, or DEEPWAVE, a field campaign conducted over New Zealand in 2014. On 12 July of that year, a research aircraft flown by the German Aerospace Center traced a rectangular path across the South Island, sampling the atmosphere at two stacked levels: one near 7.9 kilometers altitude and another near 10.9 kilometers, close to the tropopause. The meteorological setup was favorable for wave generation, with moderate northwesterly winds of roughly 10 to 15 meters per second at 5 kilometers altitude blowing perpendicular to the Southern Alps, and about 20 meters per second near the tropopause, with little change in wind direction with height. Those stacked flight legs provided rare, high-frequency in situ measurements of waves and turbulence signatures near the tropopause, exactly the kind of ground truth that model developers crave.

The simulation tool at the heart of the study is the ICON model, the ICOsahedral Nonhydrostatic modeling framework developed jointly by the German Weather Service and the Max Planck Institute for Meteorology. ICON uses an icosahedral-triangular grid built from geodesic Delaunay triangulation, a height-based vertical coordinate and C-grid staggering of its variables. Jagan and Schmidli ran ICON in numerical weather prediction mode at horizontal grid spacings of approximately 2 kilometers, 1 kilometer and 500 meters, all with 137 vertical levels and a model top at 30 kilometers. They then varied the maximum vertical grid spacing in the upper troposphere and lower stratosphere, the critical region spanning roughly 6 to 25 kilometers altitude, among 400, 200 and 100 meters. To anchor the finest end of the comparison, they added an online-nested large-eddy simulation that zoomed down to a remarkable 130 meters over the southern flight segment, using four nested domains stepping from 1 kilometer down to the 130-meter innermost nest.

The turbulence question was addressed by pitting two schemes against each other. The first is the operational turbulent kinetic energy scheme, which predicts a single TKE value and augments the classical shear-driven turbulence budget with empirical extra source terms, one for large-scale horizontal shear and one for breaking subgrid-scale mountains, both originally designed to support aviation turbulence products such as the eddy dissipation rate. The second is the newly developed two-energy scheme, which carries two prognostic turbulence energies, dynamically adjusts its mixing length, avoids prescribed minimum diffusion coefficients and employs an assumed probability density function method for buoyancy production, all intended to better handle the anisotropic, strongly stratified turbulence found aloft rather than the boundary-layer conditions for which most closures were originally tuned.

The reference simulation, run at 1 kilometer resolution with the two-energy scheme, painted a vivid picture of the event. Alternating bands of ascent and descent stretched downstream of the main ridge from the mid-troposphere into the lower stratosphere, strongest in the troposphere and weakening with height. A powerful wave with vertical velocities exceeding 3 meters per second rose above the Mount Cook region, New Zealand’s highest summit, while a weaker wave formed above the Two Thumb Range farther southeast. The simulated wave field proved quasi-stationary across the three-hour flight window, which made the comparison with the aircraft data meaningful. Against the observations, the model agreed reasonably well on the lower flight leg but underestimated wave amplitude on the upper leg, and the large-eddy simulation better captured the wavelength and phase of the waves even though the coarser NWP runs better reproduced their amplitude.

Resolution emerged as a decisive factor. At 2 kilometers, the simulated waves took on an artificially hydrostatic character, their horizontal wavelengths stretching as the model lost the ability to represent shorter scales. At 500 meters the waves became more non-hydrostatic and vertically aligned, and the number of resolved waves multiplied as shorter wavelengths entered the picture, an effect even more pronounced at 130 meters in the large-eddy simulation, where clear examples of trapped lee waves appeared in the stable layer below about 5 kilometers. Vertical resolution mattered just as much: simulations with 400-meter vertical spacing showed clear differences from the reference, particularly in the stratosphere, while 200-meter and 100-meter runs were nearly identical, indicating that near-convergence of the primary wave requires vertical grid spacing of 200 meters or finer in the upper troposphere and lower stratosphere.

The most striking finding concerned the turbulence schemes themselves. Both produced essentially the same resolved wave structures, yet the parameterized turbulent kinetic energy differed enormously between them. The operational TKE scheme generated substantially higher TKE throughout the wave field, with turbulence patches sprinkled across the troposphere and even reaching into the lower stratosphere, whereas the two-energy scheme sensibly confined turbulence to the atmospheric boundary layer and the strongest low-level wave regions. By systematically switching off the empirical source terms, the researchers traced the excess to the horizontal-shear term. This term is a two-dimensional Smagorinsky-type closure that assumes grid-scale deformation feeds an irreversible cascade into unresolved turbulence, but in the simulated mountain-wave case much of the deformation came from large, coherent, largely reversible gravity waves. Unable to tell the difference, the closure converted wave-related deformation into spurious TKE, which is why the excess energy appeared spatially correlated with the wave field rather than with genuinely turbulent regions.

Further confirmation came from bulk diagnostics averaged over a subdomain of the Southern Alps. Area-averaged vertical velocity variance at 500 meters matched the large-eddy simulation closely in the lower troposphere, while the 1 and 2 kilometer runs systematically underestimated it in the trapped-lee-wave layer. Above that layer, the 1 and 0.5 kilometer runs agreed well, signaling near-convergence for the dominant wave. Notably, the momentum flux profiles at 1 kilometer were nearly identical regardless of turbulence scheme, and the spurious horizontal-shear TKE left no corresponding signature in the wave drag, a decoupling the authors interpret as further evidence that the extra turbulence was artificial. The low-level gravity-wave momentum flux, a key target quantity for gravity-wave drag parameterizations, kept increasing as resolution improved from 1 kilometer to 500 meters, meaning the bulk momentum budget had not yet converged even at these very fine grids. Intriguingly, the comparison also showed that the large-eddy simulation is not automatically the more realistic reference: at 130 meters it resolved the shortest waves’ wavelength and phase but damped their amplitude, likely over-dissipating near-grid-scale wave energy, and it underestimated the vertical velocity fluctuations observed in the upper troposphere.

The practical implications reach well beyond academic curiosity. Turbulence in the upper troposphere and lower stratosphere is intermittent and localized, shaping stratosphere-troposphere exchange of greenhouse gases and pollutants and posing real hazards on long-haul routes that cross major mountain ranges. The study’s guidance is concrete: reliable simulation of small-scale mountain waves and turbulence at kilometer-scale resolution demands horizontal grid spacings of 1 kilometer or finer and vertical spacings of 200 meters or finer in the UTLS, while trapped lee waves and downstream wave structure remain stubbornly resolution-sensitive even then. Equally important, operational centers running kilometer-scale models over mountainous terrain should reconsider the empirical horizontal-shear and subgrid-orography terms of their TKE closures, which can inflate turbulence where none exists, or adopt alternatives like the two-energy scheme, which behaves stably in strongly stratified conditions without exhibiting the artifact. The authors caution that they do not claim the new scheme is more accurate everywhere, and genuine sub-kilometer turbulence downstream of the secondary ridge remained under-resolved in every configuration, including the large-eddy run. Future work will extend the analysis to additional DEEPWAVE cases and seek to couple the two-energy scheme with gravity-wave parameterizations, bridging the gap between explicitly resolved and parameterized wave-turbulence interactions in global and climate models, a step that could ultimately make turbulence forecasts that pilots and passengers depend on considerably more trustworthy.

Subject of Research: Simulation of mountain waves and turbulence in the upper troposphere and lower stratosphere using the ICON model at varying resolutions and turbulence schemes

Article Title: Impact of model resolution and turbulence scheme on the representation of mountain waves and turbulence

Article References: Impact of model resolution and turbulence scheme on the representation of mountain waves and turbulence. (n.d.). https://doi.org/10.5194/acp-26-13721-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-13721-2026

Keywords: mountain waves, gravity waves, turbulence, ICON model, DEEPWAVE, large-eddy simulation, turbulent kinetic energy, model resolution, upper troposphere, lower stratosphere, aviation safety, numerical weather prediction

The research, led by Gabriel Pascual, Júlia Soliva, Jordi Mazon and David Pino of the Universitat Politècnica de Catalunya in Barcelona, takes an unusual but well-established approach. Rather than tracking individual turbulence events, which last seconds or minutes and cannot be resolved in monthly data, the team analyzed 75 years of monthly ERA5 reanalysis data from 1950 to 2025 to map the slow evolution of the synoptic-scale environments in which clear-air turbulence forms. The method acts like a temporal low-pass filter: it strips away the stochastic noise of individual encounters and reveals the steady, low-frequency variability of the background flow, much as climate scientists use monthly averages of convective available potential energy and wind shear to chart where severe thunderstorms are likely rather than where any single storm struck.

The study domain stretches from 75 degrees north to 20 degrees north and from 35 degrees west to 50 degrees east, covering the Euro-Mediterranean flight corridor. The researchers extracted data at two pressure levels, 200 and 225 hectopascals, which correspond to standard cruise altitudes for long-haul commercial aviation and coincide with the mid-latitude jet stream core and the oscillating boundary of the tropopause, where vertical wind shear and static stability gradients are most pronounced. From the 21 turbulence diagnostics proposed in earlier literature, seven indices were selected for their consistency across previous studies: Horizontal Divergence, the Richardson Number, the Brown Index, the North Carolina State University Index NCSU1, the two Ellrod Indices, and Potential Vorticity.

Each index captures a different facet of turbulence physics. The Richardson Number expresses the dimensionless ratio of buoyancy to shear generation and is the classical measure of whether a flow is dynamically stable; values below the critical threshold indicate that small disturbances can grow into full turbulence. The Ellrod Indices combine vertical wind shear with horizontal deformation, the kinematic precursors identified by Ellrod and Knapp in 1992, and the second variant, which adds a horizontal convergence term, has been verified against pilot reports with a probability of detection between 70 and 84 percent. The NCSU1 index integrates localized rotation, vertical shear and buoyancy-related instability to identify the flow’s capacity to amplify minor perturbations, particularly near jet streams and mountain waves, while Potential Vorticity serves as a tracer for the tropopause itself.

Before searching for trends, the team first tested how the indices relate to one another when computed from time-averaged fields, using the Spearman rank correlation because the data are not normally distributed. The Ellrod Index 1 emerged as the most robust proxy for general turbulence trends, correlating strongly, above 0.70, with EI2, NCSU1 and the Brown Index, all of which share a dependence on vertical wind shear or the deformation tensor. In contrast, Horizontal Divergence and the Richardson Number showed negligible correlation with the deformation-based suite, and Potential Vorticity displayed a consistent moderate negative correlation, reflecting the inverse relationship between PV-defined stability and turbulent kinetic energy. All correlations were statistically significant at the 0.05 level, with their strength varying between the two pressure levels.

The trend analysis then focused on three non-redundant indices: HDG, the negative Richardson Number and NCSU1. The results were strikingly index-dependent. Horizontal Divergence showed no statistically significant long-term trend at either level, with regression p-values of 0.532 at 200 hectopascals and 0.358 at 225 hectopascals, far above the significance threshold. The negative Richardson Number, by contrast, declined significantly over the 75 years, with a p-value of 2.88 times ten to the minus four, indicating a systematic shift toward reduced static stability and enhanced macroscopic vertical wind shear at cruise altitudes. NCSU1 showed a statistically significant positive trend, with a p-value of precisely 0.05, suggesting a long-term expansion of atmospheric configurations prone to generating clear-air turbulence, though the authors caution that this result sits right at the conventional significance threshold and is accompanied by considerable interannual variance.

The seasonal breakdown added further nuance. Winter emerged as the most turbulent season, with Richardson Number values closest to zero, indicating the strongest mechanical turbulence, and with NCSU1 values in some cases twice as large as the annual averages. Summer and autumn were the most quiescent, and their temporal evolutions resembled each other more than they resembled winter or spring. The seasonal trends also diverged in direction: for the negative Richardson Number, every season except spring showed a descending trend consistent with a year-round shift toward a more unstable upper-tropospheric background flow, while spring alone bucked the pattern with a positive trend and very high interannual variability. For Horizontal Divergence, winters actually became more stable over the period, potentially because rising vertical temperature gradients affect lapse rates.

The most eye-catching finding came from frequency-domain analysis. Applying the Fast Fourier Transform to the annually averaged series, the team found a dominant periodicity of 2.45 years in the Horizontal Divergence and Potential Vorticity indices, with related short-period peaks of about 2.8 years appearing in NCSU1 and EI2. Crucially, no index showed a periodicity corresponding to the familiar 11-year solar cycle. Instead, the 2.45-year signal aligns with historical findings by Clough in 1928, who showed that once the 11-year cycle is removed, variations in solar activity with a most common period of roughly 2.4 years become visible, a periodicity he argued was not accidental because purely stochastic variations would more likely cluster around 1.6 years. When the researchers compared the years in which the 2.45-year periodicity was most prominent with the onset years of the 11-year solar cycles from 1950 to 2025, none coincided, further suggesting the shorter-period solar cycles rather than the primary cycle are at play.

The physical mechanism, the authors propose, runs through the stratosphere. Quasi-biennial oscillations of solar activity have been documented in sunspot records using methods such as Empirical Mode Decomposition and Singular Spectrum Analysis, and the interaction between solar maxima and the easterly phase of the Quasi-Biennial Oscillation is known to precondition the polar vortex for sudden stratospheric warmings. When these warmings occur, the polar vortex breaks down dramatically and disrupts the jet stream, the primary driver of enhanced turbulence over Europe. Recent work by Vokhmyanin and colleagues achieved an 86 percent success rate in predicting winter stratospheric warming occurrences as early as the preceding August by modeling the interplay between solar and geomagnetic activity and the QBO, underscoring that the 2.45-year turbulence cycle is not a random fluctuation but is consistent with solar-driven oscillations modulating the large-scale stratospheric dynamics that dictate upper-tropospheric conditions. The authors are careful to note that a definitive causal link cannot yet be established and that the hypothesis requires further validation with longer time series or higher-resolution solar proxy data.

The study also carries practical weight for aviation. Previous research has shown that a doubling of carbon dioxide concentrations could raise the frequency of moderate-to-severe North Atlantic turbulence by 40 to 170 percent, that severe turbulence exposure there grew by 55 percent between 1979 and 2020, and that anthropogenic warming strengthens vertical wind shear at 250 hectopascals by 27 percent while reducing stratification and the Richardson number by 47 percent. Yet this new analysis shows that over continental Europe the picture is less uniform than over the open ocean: orographic forcing and higher surface roughness disrupt the coherence of vertical wind shear, making regional trends strongly dependent on which diagnostic is used. Distinguishing these macroscopic climate precursors from microscale turbulence events, the authors argue, is essential for aviation safety risk assessments in a changing climate, and the finding that part of Europe’s turbulence-prone environment pulses on a quasi-biennial solar rhythm adds an unexpected, and potentially predictable, dimension to the forecast.

Subject of Research: Long-term climatology of clear-air turbulence-favorable atmospheric conditions over Europe and their modulation by quasi-biennial solar cycles

Article Title: Climatology of large-scale atmospheric conditions favorable to clear-air turbulence over Europe (1950–2025): the role of quasi-biennial solar modulation

Article References: Pascual, G., Soliva, J., Mazon, J., & Pino, D. (2026). Climatology of large-scale atmospheric conditions favorable to clear-air turbulence over Europe (1950–2025): the role of quasi-biennial solar modulation. Climate Dynamics, 64(11), Article 459. https://doi.org/10.1007/s00382-026-08418-3

Image Credits: AI Generated

DOI: 10.1007/s00382-026-08418-3

Keywords: clear-air turbulence, aviation safety, ERA5 reanalysis, Richardson number, Ellrod index, potential vorticity, jet stream, quasi-biennial oscillation, solar activity, stratospheric dynamics, climate change, Europe

The research, led by Isabel H. Smith of the University of Reading, together with Paul D. Williams and Reinhard Schiemann, quantifies changes in moderate-or-greater mountain wave turbulence by 2050 under a high-end warming scenario. The team used a multi-model approach built on the high-resolution tier of the Coupled Model Intercomparison Project, known as HighResMIP, which was designed to close the gap between climate models and numerical weather prediction. After testing seven horizontal grid spacings across three global climate models, the authors settled on three of the finest: HadGEM3-GC3.1-HM at 25 kilometres, EC-Earth-3P-HR at 36 kilometres and MPI-ESM1.2-XR at 34 kilometres. The choice mattered. In their comparison of the 1950s with the 2040s, coarser models with grid spacings of 60 kilometres or more failed to resolve many known turbulence zones altogether, a consequence of smoothed topography and terrain gradient thresholds that erase the steep slopes where mountain waves are born.

Diagnosing turbulence in climate model output is an art of proxies. Because the models cannot simulate turbulent eddies directly at these scales, the researchers applied six established clear-air turbulence indices, including horizontal temperature gradient, frontogenesis function, wind divergence, wind speed, flow deformation and a combined flow deformation and wind speed measure. Each index was then multiplied by a terrain-dependent factor requiring surface heights above 200 metres, terrain gradients steeper than 0.5 kilometres per metre and sufficient low-level wind speed below one kilometre. This construction, validated in earlier work against pilot reports, effectively isolates turbulence generated by mountains. The team defined moderate-or-greater turbulence as the 98th percentile of each index at the 200 hectopascal pressure level, roughly the cruising altitude of jet aircraft, and only accepted trends that passed a 95 percent confidence test in a linear regression over the 101-year period from 1950 to 2050.

The headline result is a world divided. The largest increase in mountain wave turbulence is projected over the Antarctic, where the average trend reaches 0.52 percent per year, equivalent to roughly 52 percent more turbulence over the century, with the strongest seasonal rise of 0.31 percent per year in the Southern Hemisphere summer. Greenland follows closely, a finding with direct consequences for the busy transatlantic corridors that cross its southern slopes. Regions covering Georgia, Azerbaijan and eastern Turkey, along with parts of Chile and Argentina in the southern Andes, also show significant increases. In contrast, the Alps, the Atlas Mountains and the northern and central Andes are projected to become calmer, as are broad swaths of Africa. The study’s sub-continental analysis across 28 regions reveals that even within a single continent the picture can reverse: North America as a whole gains turbulence, yet the zone covering the Rocky Mountains loses it in every season, with a decline of more than 50 percent in spring, the greatest decrease found anywhere in the study.

Seasonality emerges as a critical dimension of the projections. Over North America, winter and autumn carry the strongest increases, consistent with the known climatology in which lowered tropopause heights and peak westerly winds crossing mountain ranges perpendicularly favour wave generation. The region encompassing Greenland, Iceland and northeastern Canada shows increases of 18.2 percent in summer, 16.2 percent in autumn, 11.1 percent in spring and 7.4 percent in winter by 2050. Over Asia, home to the Himalayas and the Tibetan Plateau, the models project more turbulence in winter but less in summer, with trends of 0.023 and minus 0.014 percent per year respectively. Australia displays a particularly sharp seasonal split, with significant declines in the southern summer and autumn but unanimous increases across all models and indices in winter and spring. The Antarctic is most turbulent in Southern Hemisphere winter and least so in summer, yet both seasons show rising trends.

The mechanism behind these shifts appears to lie substantially in the winds near the surface. Mountain waves draw their energy from low-level airflow, so any change in wind speed over terrain directly modulates wave amplitude and the likelihood of breaking. The study found a positive correlation between projected changes in turbulence and changes in 10-metre wind speed, examining both average and maximum winds to capture the general trend and the occurrence of extreme gusts. In the Antarctic, increases in turbulence in both summer and winter were linked to strengthening average and maximum low-level winds, a finding that echoes observational work by the British Antarctic Survey showing surface winds there have strengthened by about 15 percent since 1980. Over Canada and Alaska, rising turbulence tracked increasing winds, while the decline over the Rockies in autumn was associated with weakening maximum wind speeds, and the drop over India and Sri Lanka followed a decrease in average low-level winds.

These results complicate a simpler narrative from earlier research. Previous studies of clear-air turbulence, which is driven largely by shifting and strengthening jet streams, have consistently projected global increases under warming, and a 2023 analysis of mountain wave turbulence using coarser models found increases over roughly two-thirds of the globe, concentrated in the mid-to-high latitudes. The new high-resolution work partially confirms that latitudinal pattern but adds important exceptions, notably the declines over the Rockies and the northern and central Andes, which sit outside the tropical band where weakening was expected. The authors attribute some of the divergence to terrestrial stilling, the observed and projected slowing of surface winds across many Northern Hemisphere mid-latitude regions, which suppresses the raw energy available to generate mountain waves even as upper-level dynamics change.

The study is candid about its limitations. Climate model topography remains smoother than reality, so some known turbulence hotspots in the western United States are not fully captured, and mountain wave turbulence often develops downstream of ranges in ways that fixed terrain-gradient thresholds may miss. All projections assume the high-end SSP8.5 emissions scenario with little mitigation, so the changes described represent an upper bound rather than a central expectation. Index choice also matters: over South America, one index placed turbulence changes in a range of 0 to 50 percent while another suggested 180 to 250 percent, a spread the authors view as complementary rather than contradictory, since different diagnostics capture different generation mechanisms. Averaging across models and indices can also mask disagreement, as occurred over the Himalayas, where one model’s significant increase outweighed two smaller declines.

For the aviation industry, the practical message is one of targeted vigilance rather than uniform worsening. Turbulence already costs airlines millions of dollars a year in delays, injuries and structural damage, and pilot reports, the traditional observational record, carry median position uncertainties of around 50 kilometres. The new projections suggest airlines serving polar and subpolar routes, particularly those crossing Greenland or the few routes over the Antarctic, should anticipate a measurably rougher atmosphere by mid-century, while carriers over the Alps, the Atlas and the northern Andes may face somewhat reduced exposure. As high-resolution climate models continue to close the gap with weather forecasting systems, studies of this kind are turning an invisible, radar-proof hazard into something that can at least be mapped in advance, giving flight planners and meteorologists a decades-long head start on the turbulence of tomorrow.

Subject of Research: Projected global changes in mountain wave aviation turbulence under climate warming

Article Title: Global shifts in mountain wave turbulence within high resolution climate models

Article References: Smith, I. H., Williams, P. D., & Schiemann, R. (2026). Global shifts in mountain wave turbulence within high resolution climate models. Weather and Climate Dynamics, 7(3), 1733-1758. https://doi.org/10.5194/wcd-7-1733-2026

Image Credits: AI Generated

DOI: 10.5194/wcd-7-1733-2026

Keywords: mountain wave turbulence, aviation safety, climate change, HighResMIP, CMIP6, clear-air turbulence, gravity waves, Greenland, Antarctic, Rocky Mountains, terrestrial stilling, climate models

The stakes are easy to underestimate. In conventional aviation, the final barrier against midair collision is the human pilot, guided by regulations such as the Federal Aviation Administration’s Title 14 CFR Part 91 and advisory guidance on the pilot’s role in collision avoidance, and by the International Civil Aviation Organization’s rules of the air. A remote pilot, however, suffers from limited situational awareness, communication latency, and degraded visual conditions. Decades of research into sense-and-avoid technologies, from early flight trials by Korn and Edinger in 2008 through comprehensive reviews by Yu and Zhang and by Fasano and colleagues, have produced a rich toolbox of radar, electro-optical, and acoustic sensors and a zoo of tracking and avoidance algorithms. What has been missing, the Korean team argues, is a complete, standards-compliant system design that ties those pieces together on hardware small enough to fly on a small unmanned platform.

That is where international standards come in. The team anchored their design to RTCA DO-365B, the Minimum Operational Performance Standards for Detect and Avoid Systems, and to ASTM F3442/F3442M, which specifies performance requirements for DAA equipment. These documents define a reference architecture built around three core functions: Track Processing, Track Alerting, and Guidance Processing. Track Processing fuses raw sensor detections into stable, continuously updated tracks of surrounding aircraft. Track Alerting monitors those tracks against well-clear criteria and issues alerts when another aircraft threatens to come too close. Guidance Processing then computes and guides an appropriate maneuver to restore safe separation. Rather than inventing a parallel architecture, the researchers took this standardized skeleton and filled in every functional detail, specifying inputs, outputs, algorithms, and interfaces so that the design could be implemented directly in flight hardware.

The engineering behind Track Processing is where much of the difficulty lies. Sensors such as radar, lidar, and ADS-B receivers, which listen to the cooperative transponder broadcasts that Garmin’s GDL 90 data interface specification standardizes, rarely agree perfectly on where an intruder is or what it is doing. The team therefore had to solve the classic multitarget tracking problems documented by Bar-Shalom and Li, Fortmann and colleagues’ joint probabilistic data association, and Blackman’s multiple hypothesis tracking: deciding which sensor returns belong to which aircraft, filtering noisy measurements into smooth state estimates, and rejecting spurious detections. Their implementation maintains track quality metrics so that uncertain or fading tracks do not trigger false alarms, drawing on track-quality-based association methods such as those of Sinha and colleagues. The result is a stream of reliable track states, including position, velocity, and confidence, handed to the alerting stage.

Track Alerting then applies the well-clear concept that regulators use to define when another aircraft is dangerously close, a notion central to NASA’s DAIDALUS detect-and-avoid alerting logic developed by Munoz, Narkawicz, and Hagen. The alerting function must be tuned carefully: alert too late and the aircraft cannot maneuver in time; alert too early and the system cries wolf, eroding trust and triggering unnecessary deviations. The researchers implemented alerting thresholds consistent with the standards’ guidance, so that the system signals a threat with enough lead time for the guidance function to act, mirroring the philosophy behind TCAS II resolution advisories that have protected crewed airliners for decades.

The heart of the paper’s technical contribution, however, sits in Guidance Processing. Once an alert fires, the aircraft must choose a maneuver, typically a vertical or horizontal deviation, that maximizes the closest point of approach with the intruder while remaining flyable and predictable. Collision cone approaches, such as the classic method of Chakravarthy and Ghose, and reactive avoidance algorithms for multiple unmanned aircraft, like the one Choi and colleagues published in 2013, offer geometric ways to reason about which directions of flight lead toward danger and which lead away. The team designed a collision avoidance method in this family and then benchmarked it quantitatively against an existing method used in DAA systems, including comparisons informed by NASA’s ICAROUS integrated configurable algorithms for reliable operations of unmanned systems. The outcome was striking: the proposed method achieved competitive collision avoidance performance while demanding substantially less computation, a decisive advantage when the processor must also run tracking, alerting, flight control, and communications on a small aircraft’s power and weight budget.

Computational efficiency is not a cosmetic metric in this field. The ACAS XU program, which applies advanced decision-making and even deep neural network compression, as explored by Julian, Kochenderfer, and Owen, to aircraft collision avoidance, has shown how expensive it can be to certify and run sophisticated avoidance logic. Studies by Wu and Lee on the impact of airborne radar uncertainties, and by Abramson and colleagues on sensor uncertainty mitigation, underline that real-world DAA performance depends as much on robust, fast processing as on raw algorithmic elegance. By demonstrating that their guidance method delivers comparable safety margins at lower computational cost, the Korean team has addressed precisely the bottleneck that keeps full DAA capability off most small platforms. Their work also complements flight-test efforts such as Kotegawa’s proof-of-concept ACAS XU airborne system and NASA’s AiRanger program, which evaluated small form-factor lidar and radar sensors for DAA applications.

Equally important is how the system was proven. The researchers validated their design through hierarchical test procedures, a layered approach in which individual functions are tested in isolation, then integrated subsystems are exercised together, and finally the complete hardware is evaluated end to end against simulated encounters. This mirrors the certification mindset of aerospace engineering, where a function that works in simulation means little until it survives hardware-in-the-loop testing with realistic sensor noise, timing jitter, and message latencies. The quantitative evaluation of the collision avoidance method was embedded in this hierarchy, allowing the team to compare miss distances, alerting behavior, and processing load against the baseline method under identical encounter conditions. Such structured validation is exactly what regulators will demand before small autonomous aircraft are allowed to operate beyond visual line of sight in civil airspace.

The broader significance extends well beyond one laboratory prototype. Airspace integration for unmanned aircraft is widely recognized as a prerequisite for the drone economy’s next phase, and DAA is the enabling technology for that integration. By showing a complete path from standardized architecture to implemented hardware to validated performance, the study offers a template that other developers can follow and that standards bodies can reference. The work was supported by a grant funded by the Korea AeroSpace Administration, reflecting national investment in autonomous aviation capability. If small aircraft are to share the sky safely with airliners, helicopters, and each other, they will need exactly this kind of disciplined engineering: sensors that see, trackers that understand, alerters that warn, and guidance that acts, all fast enough to fit on a flying computer the size of a shoebox. This study moves that vision measurably closer to reality.

Subject of Research: Airborne detect-and-avoid and collision avoidance systems for small autonomous unmanned aircraft

Article Title: Development of An Airborne Detect-and-Avoid System for Small Autonomous Aircraft

Article References: Choi, H., Kim, J., Bae, J., Kim, T. S., Kim, D., Koo, B., Lee, J., Suk, J., Yang, S.-K., & Choi, H. K. (2026). Development of An Airborne Detect-and-Avoid System for Small Autonomous Aircraft. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01262-z

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01262-z

Keywords: detect-and-avoid, collision avoidance, unmanned aircraft systems, drones, airspace integration, track processing, well clear, RTCA DO-365B, guidance processing, sensor fusion, autonomous flight, aviation safety

The research, published as an open-access paper in the International Journal of Aeronautical and Space Sciences, simulated a 50.8 millimeter hailstone, the classic two-inch benchmark of aviation hail threat models, slamming into a radome modeled on the Airbus A320neo at a combined velocity of roughly 240 meters per second, the speed of a typical cruise condition. The team, led by Anand Pai with co-authors including P. K. Sowmya, Khushhali Chaudhary, Suhas Yeshwant Nayak, Chandrakant R. Kini and B. Satish Shenoy, compared two industry-standard fiber-reinforced polymer systems: glass fiber-reinforced polymer, or GFRP, and Kevlar, or p-aramid, fiber-reinforced polymer, known as KFRP. The verdict was unambiguous. The KFRP radome limited peak deformation to 8.1 millimeters, an 18 percent reduction compared with the 9.88 millimeters suffered by the GFRP variant, while weighing about 19 percent less.

The stakes are far from academic. The researchers compiled a survey of documented hailstone strikes on commercial aircraft over the past decade, and the record shows a rising frequency of encounters, a trend the authors attribute to the growing incidence of severe convective weather and the continued expansion of global air traffic. Hailstones routinely strike the nose cone, fuselage, wing skins, leading edges, control surfaces, engine nacelles and fan blades. The radome is particularly exposed because of its forward-facing geometry, and the minimum skin thickness requirements imposed to protect it carry a real weight penalty, one that matters acutely for smaller jets where every kilogram affects performance and fuel burn.

Modeling a hailstone is deceptively hard. Natural hail is polycrystalline ice, a layered, anisotropic material riddled with air and water pockets, and its mechanical behavior swings dramatically with strain rate. At the high strain rates of 100 to 300 per second typical of these impacts, ice transitions from ductile to brittle failure, yet high-speed imaging shows it retains residual strength through fragmentation and axial splitting even after damage initiates. Its compressive strength and hardness can climb by more than 300 percent as temperature falls from minus 15 to minus 50 degrees Celsius. The team represented the projectile using Smoothed Particle Hydrodynamics, a mesh-free method that sidesteps the severe mesh distortion that cripples traditional Lagrangian finite element approaches during extreme deformation. The hailstone was discretized into 67,731 particles of one millimeter each, governed by a bilinear isotropic constitutive model that captures the critical moment when the ice’s deviatoric stress vanishes at the failure threshold and the projectile begins to behave like a fluid.

The radome itself was built with equal care. The 7.5 millimeter skin was assembled from thirty discrete plies of 0.25 millimeters each in a symmetric [0/90]15 stacking sequence of 2 by 2 twill weave fabric. Rather than plugging in generic material cards, the researchers derived the effective orthotropic properties through a multi-scale homogenization workflow: representative volume elements of the unidirectional yarn architecture and the bidirectional woven architecture were simulated in ANSYS Material Designer, using manufacturer data for Kevlar 2200 aramid fibers with a Young’s modulus of 108 gigapascals and EC11 204 E-glass fibers at 72.4 gigapascals, both embedded in an epoxy matrix. The resulting laminate definitions were then exported into ANSYS Explicit Dynamics, where the coupled SPH-finite element framework ran the impact scenarios over a 100 microsecond window.

The simulation reproduced a two-stage impact response well known from experimental literature on flat composite panels: a rapid loading stage in which stress and deformation climb to a peak, followed by a rebound stage as the projectile fragments and the stress decays. On the curved radome geometry, the two materials diverged after about 40 microseconds. Up to that point their deformation histories were nearly indistinguishable, but the GFRP shell then deformed more steeply toward its larger final value, while the KFRP laminate held a more stable late-stage profile. Contact stress told a complementary story. The KFRP radome reached a higher peak stress of roughly 1581 megapascals at about 60 microseconds, versus about 890 megapascals at 70 microseconds for GFRP, indicating a stiffer, more localized response that accelerates hailstone fragmentation and spreads the debris tangentially along the radome’s curvature, a phenomenon the authors call geometric shedding.

Energy accounting reinforced the picture. Both simulations began with an identical projectile kinetic energy of 1925 joules, and total energy was conserved throughout, with hourglass energy held below 2 percent of peak internal energy, satisfying standard explicit dynamics stability criteria. But the KFRP radome accumulated more internal energy and showed a sharper decay of the projectile’s kinetic energy, signatures of enhanced momentum dissipation through multi-axial matrix shearing and plastic work. In other words, the aramid architecture does not merely deflect the blow; it absorbs and dissipates the impact energy more effectively across the convex surface of the nose cone.

Perhaps the most technically interesting contribution is the ply-by-ply failure analysis. Because global mechanical metrics poorly capture the directional nature of composite damage, the team implemented a sequential explicit-to-implicit mapping: deformation states extracted at 40, 60, 80 and 100 microseconds from the explicit solver were transferred into a static implicit environment, where the anisotropic Tsai-Wu failure criterion yielded an Inverse Reserve Factor for individual plies. An IRF at or above 1.0 signals the onset of material degradation. The results exposed a fundamental difference in damage character. In the GFRP radome, elevated failure indices spread diffusely, with the rear ply reaching IRF values up to 3.30 across wide, unconfined contours, a pattern that raises the risk of widespread micro-cracking, shear failure and interlaminar delamination. The KFRP radome, by contrast, confined its most severe indices, which climbed as high as 11.79, to a narrow zone at the impact apex, while the mid-plane plies stayed comparatively protected. Damage containment, not merely damage resistance, emerged as the Kevlar laminate’s decisive advantage, preserving the load-bearing capacity of the surrounding shell and the radar instruments behind it.

To bridge the gap between idealized single impacts and operational reality, the researchers also ran a cluster scenario of five concurrent 50.8 millimeter hailstones spaced 60 millimeters apart, extending the simulation to 5 milliseconds to capture staggered strikes and shockwave interactions. The result was a wider concave indentation with surface cratering across the central radome region, a damage envelope that closely resembles photographic evidence of real hail damage on an operational Interjet A320 nose cone. That qualitative agreement, the authors argue, validates the framework’s ability to capture the momentum transfer required to produce macroscopic cratering. The team is candid about limitations: the bilinear ice model omits thermal effects and microstructural heterogeneity, and future work will add stochastic multi-impact scenarios and cohesive zone modeling to probe the interplay between intralaminar softening and delamination. Still, the message for aircraft designers is striking. A radome built from Kevlar rather than glass fiber could shed roughly 19 percent of its mass, about 114 kilograms in the modeled geometry, while deforming less, absorbing more energy and keeping damage tightly contained where it matters least to the aircraft’s survival.

Subject of Research: Computational simulation of hailstone impact response of composite aircraft radomes

Article Title: Dynamic Impact Response of Composite Aircraft Radomes Subjected to Hailstone Impact: A Computational Study

Article References: Pai, A., Sowmya, P. K., Chaudhary, K., Nayak, S. Y., Kini, C. R., & Satish Shenoy, B. (2026). Dynamic Impact Response of Composite Aircraft Radomes Subjected to Hailstone Impact: A Computational Study. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01308-2

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01308-2

Keywords: aircraft radome, hailstone impact, Kevlar, GFRP, composite laminates, SPH, finite element analysis, Tsai-Wu failure criterion, aviation safety, explicit dynamics, impact damage, A320neo

The starting point is the so-called W-shaped development process, proposed by the European Union Aviation Safety Agency together with the company Daedalean in their Concepts of Design Assurance for Neural Networks reports. The classic V-model, which underpins standards such as DO-178C in avionics, ISO 26262 in automotive engineering, and the ECSS framework in aerospace, arranges development as two arms: requirements are refined into code on one side, and hierarchical testing verifies them on the other. The W-shaped process extends this with a second V dedicated to what EASA calls learning assurance—verifying that a learned model generalizes to unseen operational data and behaves robustly within its Operational Design Domain, the envelope of conditions in which it is meant to operate. For supervised learning, where datasets with labels exist before training begins, the process works. But reinforcement learning is different, and that difference matters.

In reinforcement learning, an agent learns by interacting with its environment: it observes a state, takes an action, and receives a reward that reflects how good the action was for achieving its goal. There are no ground-truth labels to check outputs against, and—crucially—there is no dataset before training. Training data is generated by the very interactions that produce learning. The researchers, led by Ángel-Grover Pérez-Muñoz of Universidad Politécnica de Madrid and colleagues, found that this single characteristic fractures the W-shaped process at its foundation. The process assumes data management precedes learning; in reinforcement learning the two are inseparable. EASA itself has deferred guidance on reinforcement learning to 2028, which makes an empirical test of the framework both timely and rare.

To probe the limits, the team designed a use case that is as much a safety exercise as an AI challenge. A defender drone, controlled entirely by a neural-network reinforcement learning agent, must follow a predefined sequence of waypoints through a simulated urban landscape built in Microsoft’s AirSim simulator, while an adversary drone tries to ram it. Three attacker strategies escalate in difficulty: a deterministic pursuer that simply chases the defender’s last known position; a predictive pursuer that estimates the defender’s position one second ahead with random scaling to mimic sensor error; and a still harder variant that randomizes its prediction horizon between one and two seconds. The defender must reach each waypoint within three meters, maintain at least 1.25 meters of separation from the attacker, and repeat this under waypoint altitudes it never saw during training.

The technical machinery underneath is careful and deliberate. The agent’s state is a twelve-dimensional vector built entirely from relative quantities—distance to the attacker, distance to the next waypoint, relative velocity, and the defender’s own velocity—so that the learned policy is invariant to where the drone happens to be in the world. Actions are continuous velocity commands on three axes, bounded at seven meters per second. The reward function stacks a cubic distance penalty, large bonuses for reaching waypoints and completing the course, and steep penalties for collisions, proximity violations, and wasted time. Training used the Soft Actor-Critic algorithm with four-layer actor and critic networks of 512 neurons each; notably, the team found SAC outperformed Proximal Policy Optimization because it prioritized minimizing penalties over chasing rewards—an intuitive fit for a system whose primary duty is not crashing.

The results carry a cautionary lesson about overtraining. An agent trained for 400,000 steps against the simple deterministic attacker achieved success rates between roughly 75 and 81 percent against all three attackers, including strategies it had never encountered, evaluated over 1,100 episodes across eleven random seeds. Its sibling trained for a full million steps on the same attacker fared dramatically worse, dropping as low as 27 percent against the predictive pursuers—a statistically significant collapse, confirmed with bootstrap tests of ten million resamples. Extended training, it appears, caused the agent to overspecialize on the deterministic chase pattern. Agents trained directly against the harder attackers did no better, with one achieving near-zero success. The lesson for safety engineers is stark: more compute and more training do not guarantee more general, more trustworthy behavior, and the certification process must be able to catch this.

What elevates the study beyond a simulation exercise is what happened next. The actor network—the part of the agent that turns observations into actions—was automatically translated from Python into C code compliant with MISRA-C, the conservative subset of C used in safety-critical software, which forbids dynamic memory allocation and thereby eliminates memory leaks, heap fragmentation, and unpredictable timing. The code was deployed on a Zynq UltraScale+ MPSoC board under the XtratuM hypervisor, which partitions the hardware in both space and time so that a fault in the AI component cannot propagate to other subsystems. Because the deployed agent must never learn in operation, only the frozen policy flies; the critic networks stay behind.

The embedded system passed its examinations. Running at a required control period of 200 milliseconds with a 100-millisecond deadline, static worst-case execution time analysis using the OTAWA tool bounded a single inference at 38 milliseconds, while 1,500 measured inferences averaged 23.62 milliseconds with a maximum of 71.18 milliseconds—comfortably inside the deadline even with interference from other tasks. Numerical comparison between the Python original and the C implementation, necessarily reduced from 64-bit to 32-bit floating point, showed errors on the order of one hundred-thousandth of a meter per second, negligible against an action range of fourteen meters per second. Success rates on the hardware—79.1, 81.4, and 79.6 percent against the three attackers—fell within the confidence intervals of the original model, demonstrating that the certified translation did not silently degrade the learned behavior.

The broader significance lies in the mapping the researchers drew between reinforcement learning’s needs and the W-shaped process’s stages. They found that the entire development-assurance half of the process—implementation, integration, and verification on real hardware—is agnostic to the learning technique and applied cleanly. The learning-assurance half required reinvention: instead of predefined datasets, the team defined scenarios derived from the Operational Design Domain to govern data generation; data quality requirements such as representativeness and completeness, which normally apply to whole datasets, were reinterpreted as properties of those scenarios; and a new step was inserted to define the agent-environment interface—the states, actions, and reward function—before training could begin.

This is, by the authors’ account, the first application of EASA’s W-shaped process to a reinforcement learning system, and it arrives as EUROCAE and SAE finalize the ARP6983/ED-324 standard for aeronautical AI. The work suggests a realistic path forward: certification frameworks built for supervised learning can stretch to accommodate learning agents that discover their own strategies, provided the emphasis shifts from curating data to curating the scenarios that generate it. The researchers caution that their results live in simulation, and closing the sim-to-real gap—sensor noise, communication delays, environmental perturbations—remains future work, as does confronting learning-based attackers. But for a field waiting on regulators until 2028, the demonstration that a reinforcement learning drone can be trained, certified, and embedded within real-time safety constraints is a milestone worth more than its 71-millisecond inference time might suggest.

Subject of Research: Applying the EASA W-shaped development process to reinforcement learning for safe drone navigation and collision avoidance

Article Title: Application of the W-shaped process for a Reinforcement Learning use case on drone navigation

Article References: Pérez-Muñoz, Á.-G., García-Quijano, H., López-García, G., Alonso, A., & Pérez, M. S. (2026). Application of the W-shaped process for a Reinforcement Learning use case on drone navigation. Machine Learning with Applications, 26, Article 101009. https://doi.org/10.1016/j.mlwa.2026.101009

Image Credits: AI Generated

DOI: 10.1016/j.mlwa.2026.101009

Keywords: reinforcement learning, drone navigation, aviation safety, EASA, W-shaped process, machine learning certification, neural networks, embedded systems, collision avoidance, Soft Actor-Critic, operational design domain, safety-critical systems

Most previous research in this area has concentrated on identifying single landing coordinates for small unmanned aerial vehicles. A quadcopter losing a motor needs a spot of ground measured in meters; a general-purpose aircraft needs a corridor, an extended strip whose length, width and straightness must all be verified simultaneously. The Moroccan team identified this as a fundamental gap: no widely adopted method existed to move beyond isolated landing points and automatically detect and outline full viable landing strips from imagery alone. Their answer was to reformulate the entire problem as one of geometric optimization, layered on top of a modern deep learning segmentation engine.

The first stage of the pipeline tackles the question of what terrain is safe to land on at all. The researchers built a hybrid deep learning model that fuses two complementary architectures: U-Net, a convolutional neural network originally developed for biomedical image segmentation, and SegFormer, a Transformer-based model that captures long-range context across an image. Rather than forcing a choice between the two, the team combined them through an ensemble method, blending their outputs so that the local, detail-sensitive reasoning of the convolutional network reinforces the global scene understanding of the Transformer, and vice versa. The result is a robust binary segmentation mask that classifies every pixel of an aerial image as either safe-to-land terrain or unsafe ground.

Technically, the pairing is well motivated. U-Net’s encoder-decoder structure excels at preserving fine spatial boundaries, while SegFormer’s self-attention mechanism allows it to reason about relationships between distant regions of a scene, such as whether a seemingly clear field is bounded by trees, power lines or buildings. This architecture hybridization reflects a broader trend in semantic segmentation research, where convolutional and attention-based approaches are increasingly merged to capture both fine texture and scene-level semantics. On a custom dataset of aerial imagery of Moroccan terrain, previously assembled and pixel-wise labeled by the same group for exactly this application, the ensemble achieved a Mean Intersection-over-Union of 80.84 percent for the segmentation task, a strong score indicating substantial agreement between the model’s safe-terrain masks and ground-truth annotations.

The segmentation mask, however, is only raw material. A safe region shaped like an amoeba is useless to a descending aircraft; what matters is the largest rectangle of contiguous usable ground that can be inscribed within it. That is the second and arguably most original contribution of the study: a novel, efficient algorithm that finds the Longest Inscribed Rectangle, which the authors abbreviate as LNIR, within any binary mask. The rectangle that solves this geometric optimization problem corresponds directly to the optimal landing strip, because it captures the maximal continuous run of terrain that satisfies the shape constraint an actual aircraft approach demands.

Geometric optimization problems of this kind are notoriously difficult. Finding the largest rectangle inside an arbitrary polygonal region is a classic problem in computational geometry, and naive approaches scale poorly with image size. The LNIR algorithm is designed for efficiency, operating directly on the segmentation output and extracting the best-fit strip without exhaustive search. Crucially, the team also generalized the method to handle strips of arbitrary orientation. A viable field does not care about the axes of the image grid, so the algorithm iteratively rotates the search space, re-evaluating candidate rectangles at each angle until the orientation yielding the longest inscribed strip is found. This rotation strategy ensures the method works equally well for a runway aligned north-south, east-west or anywhere in between.

Real-world deployment demands more than accuracy in ideal conditions. Emergency landings do not wait for clear skies, so the researchers stress-tested the framework under simulated adverse weather, applying fog and rain degradations to their imagery and re-running the full pipeline. The system maintained high performance despite the visual corruption, demonstrating a robustness that is essential for any safety-critical application. Fog and rain reduce contrast, blur edges and shift color distributions, conditions that routinely break computer vision systems trained on clean data; the fact that this pipeline survives them suggests the ensemble segmentation approach learns features that are genuinely structural rather than merely textural artifacts of fair weather.

Equally important is speed. The entire pipeline, from raw aerial image to delineated landing strip, demonstrates real-time capability, meaning it can in principle keep pace with the continuously updating view from an aircraft-mounted camera or an unmanned scout vehicle. The researchers frame the work as a powerful automated tool for a critical aviation safety application, and they have made the ingredients available to the community: the dataset of pixel-wise labeled Moroccan emergency landing sites is hosted on Mendeley Data, and the code implementing the LNIR method is publicly accessible on GitHub, lowering the barrier for other groups to reproduce, benchmark and extend the approach.

The implications extend beyond the immediate use case. The authors position the LNIR algorithm as a versatile geometric optimization tool with potential for broader feature extraction from remote sensing data anywhere that elongated rectangular structures matter. Road and railway segment extraction, agricultural strip monitoring, solar farm siting, vegetation corridor analysis and pipeline inspection all reduce, at some level, to finding the best inscribed or aligned rectangle within a classified region, and an efficient, rotation-invariant solver for that problem is a reusable piece of scientific infrastructure. The work also arrives at a moment when hybrid CNN-Transformer segmentation models are proliferating across remote sensing, from urban scene parsing to geological structure detection, and the Moroccan study offers a concrete demonstration that such hybrids can be pushed all the way through to actionable geometric outputs rather than stopping at pixel labels.

There remain, of course, the usual caveats separating an academic pipeline from certified flight hardware. The evaluation relied on a custom dataset drawn from Moroccan terrain, and generalization to deserts, forests, snowfields or dense urban environments would require further validation. The segmentation accuracy of roughly 81 percent IoU, while impressive, leaves room for error in exactly the boundary regions that determine whether a strip is long enough. And any autonomous emergency system would eventually need to integrate additional sensing modalities such as LiDAR depth information, account for slope and surface bearing strength, and satisfy the exacting certification standards of aviation regulators. Still, the conceptual leap is clear and consequential: the study shows that the full chain from aerial photograph to quantified, oriented, optimal landing strip can be automated end to end, in real time, in bad weather. For a pilot gliding toward unfamiliar ground with failing systems, that chain could one day mean the difference between guessing and knowing where the aircraft can safely come to rest.

Subject of Research: Automated detection and geometric delineation of emergency landing strips in aerial imagery using hybrid deep learning segmentation and a longest inscribed rectangle optimization algorithm

Article Title: Automated Delineation of Viable Emergency Landing Strips in Aerial Imagery Using a Novel Geometric Optimization Algorithm

Article References: Illi, A., Bouzaachane, K., El Hadaj, S., & El Guarmah, E. M. (2026). Automated Delineation of Viable Emergency Landing Strips in Aerial Imagery Using a Novel Geometric Optimization Algorithm. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01278-5

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01278-5

Keywords: emergency landing, aerial imagery, deep learning, semantic segmentation, U-Net, SegFormer, computational geometry, geometric optimization, inscribed rectangle, remote sensing, aviation safety, UAV