Aerosols, tiny particles suspended in the atmosphere, wield a powerful influence over Earth’s energy balance. They scatter and absorb solar radiation and serve as nuclei for cloud formation, both of which directly affect the planet’s temperature regulation. Traditionally, climate models have simplified these complex interactions by assuming uniform aerosol properties worldwide. However, this new study reveals that such simplifications mask vital regional variations, especially in urban and polluted environments, where aerosol particles exhibit enhanced water uptake.

The research team focused on the parameter κ, a measure of aerosol hygroscopicity, which dictates the extent to which particles absorb moisture. Using an unprecedented dataset encompassing ten geographically diverse sites—ranging from megacities like Delhi and Cairo to pristine regions such as the Atlantic Ocean—they employed explainable machine learning algorithms to analyze size-dependent variations in κ. By integrating chemical composition, particle size distribution, and meteorological data over several years, the model could accurately capture the intricate behavior of aerosol mixtures on a global scale.

What sets this study apart is its methodological innovation. Previous models predominantly assumed idealized internal mixing states, where aerosol particles are treated as homogeneously mixed entities. In contrast, this machine-learning-driven approach accounts for externally mixed particles—those composed of distinct chemical species—which are particularly prevalent in densely populated, polluted urban centers. These findings underscore that aerosols’ chemical heterogeneity plays a pivotal role in governing their hygroscopic growth and, consequently, their climatic effects.

One of the more striking revelations is the observed ‘cooling hole’ phenomenon over the Indian subcontinent. Despite being a region with high pollution levels, India is warming at roughly half the rate of the global average. The enhanced hygroscopic growth of aerosols in urban hotspots such as Delhi suggests that increased water uptake by particles leads to greater reflectivity and stronger cloud formation, which, in turn, could partially offset regional warming. This intricate balance between aerosol chemistry and climatic feedback mechanisms emphasizes the need for more nuanced—and region-specific—parameterizations within climate models.

Further compounding the significance of this research is its public health dimension. Elevated hygroscopic growth is linked to smog formation, with potential respiratory impacts exacerbated in megacities facing notorious air pollution. Drone-based measurements in Delhi verified that such aerosol behavior not only influences climate but also contributes to urban air quality challenges.

The implications of integrating such detailed aerosol hygroscopicity data into climate models are substantial. Prof. Mira Pöhlker, a co-author from the Leibniz Institute for Tropospheric Research (TROPOS), highlights that incorporating regionally resolved parameterizations can adjust estimates of direct radiative forcing by ±0.1 watts per square meter. While seemingly modest, this adjustment is significant on a global scale and could refine the accuracy of future climate projections substantially.

The study also reveals that current global climate models, which often utilize uniform aerosol parameters, may significantly misrepresent the magnitude and even the sign of aerosol-radiation interactions in certain regions. By deploying a data-driven algorithm trained on an extensive suite of observations and leveraging explainable ML techniques, the researchers offer a robust alternative that captures the structural complexity and chemical diversity of atmospheric aerosols more faithfully.

The authors hope that their improved parameterizations will be adopted in the next generation of global climate models, thereby narrowing the uncertainty margins that currently challenge climate science. Such integration could transform projections of regional climate behavior, allowing policymakers and stakeholders to tailor mitigation strategies with greater precision.

In conclusion, this comprehensive study illuminates the intricate link between aerosol hygroscopicity and Earth’s climate system, emphasizing the crucial role of regional variability. Its innovative use of explainable machine learning opens new avenues for advancing atmospheric sciences. As atmospheric aerosol processes remain a critical but elusive component in climate modeling, the findings presented here represent a pivotal step toward resolving long-standing uncertainties in radiative forcing estimates. With growing challenges posed by urban pollution and global climate change, enhancing our grasp of aerosol-cloud interactions is more vital than ever.

Subject of Research: Not applicable

Article Title: Regional aerosol hygroscopicity influences radiative forcing globally.

News Publication Date: 7-May-2026

Image Credits: Shravan Deshmukh, TROPOS

Keywords: Aerosol hygroscopicity, radiative forcing, climate models, machine learning, atmospheric aerosols, urban pollution, cloud condensation nuclei, regional climate variability, aerosol chemistry, explainable machine learning

Urban heat islands, a well-documented phenomenon, describe the elevated temperatures in cities compared to their rural surroundings due to factors like urban infrastructure, vegetation loss, and localized heat retention. While urban heat islands have been studied extensively, this study delivers new insights into how climate change could exacerbate these effects in medium-sized cities, which have been overlooked in most global climate assessments that tend to focus on megacities. The research specifically targets 104 cities with populations between 300,000 and one million, offering a deeper understanding of urban warming trends in these critical yet understudied urban centers.

The methodology integrates state-of-the-art global climate projections with advanced machine learning models to forecast changes in daytime land surface temperatures. This approach surpasses conventional climate models that often lack the spatial resolution to accurately capture the complex dynamics of smaller cities. By bridging this gap, the study elucidates that a staggering 81 percent of the examined cities are predicted to experience greater temperature increases compared to their adjacent rural areas. Even more striking is the finding that approximately 16 percent of these cities might witness warming rates 50 to 100 percent higher than their hinterlands under 2°C of warming, a scenario likely within reach by mid to late 21st century.

The implications of these findings are profound, particularly given the demographic importance of medium-sized cities, which collectively outnumber larger urban conglomerations by a factor of more than two and a half. Situated primarily in already warm tropical and subtropical climates, these cities face amplified heat stresses that could significantly impact human health, urban infrastructure, and energy consumption. Elevated temperatures in these urban areas pose heightened risks of heat-related illnesses, strain on healthcare systems, and increased mortality during extreme heat events intensified by climate change.

Dr. Sarah Berk, who led this investigation during her doctoral studies at UEA’s School of Environmental Sciences, emphasizes the dual challenges cities face: increasing regional temperatures coupled with the intensification of urban heat islands. She notes the limitations of global climate models in resolving temperature changes at the scale of medium-sized cities, advocating for complementing these models with machine learning frameworks to capture finer-scale urban warming phenomena. Now based at the University of North Carolina at Chapel Hill, Dr. Berk’s research marks a critical advancement in urban climate science, underscoring the urgency for tailored climate adaptation strategies.

Professor Manoj Joshi of UEA’s Climatic Research Unit, a co-author of the study, highlights that many tropical and subtropical urban areas are projected to exceed the warming rates of their surrounding environments, thereby intensifying urban heat stress. This discrepancy between urban and rural warming is particularly notable in densely populated regions of North-East China and northern India, where certain cities may experience temperature rises of around 3°C, while Earth System Models project only 1.5 to 2°C warming for their rural hinterlands.

The study’s focus on medium-sized cities is a deliberate and crucial aspect, as these urban centers often lack the resources and infrastructure available to megacities to mitigate extreme heat impacts. Enhanced urban warming driven by climate change means that these cities could encounter unprecedented heat challenges that demand urgent, localized mitigation and adaptation measures. The integration of machine learning enables the dissection of complex climate-urban interactions that traditional modeling may overlook, aiding policymakers and urban planners in creating resilient urban environments.

To ensure the robustness of their analysis, the research team deliberately excluded cities located in mountain and coastal areas, where topography, proximity to large water bodies, and other localized factors can confound temperature readings. This methodological rigor ensures that the observed warming trends are strongly attributable to the physical processes related to regional climate and urbanization effects rather than extraneous geographical influences.

Among the five largest urban centers analyzed, Jalandhar in India, Fuyang in China, and Kirkuk in Iraq demonstrate the most pronounced additional warming — about 0.7 to 0.8°C higher compared to their rural surroundings. Conversely, Marrakech in Morocco and Campo Grande in Brazil did not exhibit significant disparities, suggesting variability in urban heat island amplification due to differing regional and urban characteristics. However, other cities like Asyut in Egypt, Patiala in India, and Shangqui in China may experience increases in urban warming ranging from 1.5 to 2°C above rural counterparts, essentially doubling their heat exposure relative to surrounding areas.

This research not only spotlights the pressing reality of enhanced urban heat exposure in tropical and subtropical cities but also underscores the limitations of relying solely on conventional Earth System Models for urban climate projections. By integrating artificial intelligence methodologies with climate science, the study paves the way for more nuanced and actionable climate risk assessments tailored to urban centers around the globe. This integrated approach is vital for understanding and addressing the mounting threats to human health and urban sustainability posed by climate change.

The findings illuminate a clear message: the next generation of urban planning, public health strategies, and climate adaptation policies must incorporate advanced predictive tools such as machine learning to grasp the true extent of urban heat risks. With climate-driven extreme heat events projected to increase in frequency and severity, these insights offer a critical foundation for resilient and climate-sensitive urban development, especially in regions where populations are rapidly urbanizing under already high thermal stress.

Supported by the Natural Environment Research Council and the ARIES Doctoral Training Partnership, and with collaboration extending to researchers now at the Karlsruhe Institute of Technology, this pioneering work elevates our comprehension of urban climate dynamics amid global warming. The study titled “Amplified warming in tropical and subtropical cities at 2°C climate change” was formally published on February 3, 2026, and marks a significant milestone in urban climatology research, offering a clarion call for urgent, informed responses to the emerging urban climate crisis.

Subject of Research: Amplified urban warming and urban heat island intensification in medium-sized tropical and subtropical cities under projected 2°C global warming.

Article Title: Amplified warming in tropical and subtropical cities at 2°C climate change

News Publication Date: 3-Feb-2026

Web References: https://www.pnas.org/doi/10.1073/pnas.2502873123

References: Berk, S., Joshi, M., Nowack, P., & Goodess, C. (2026). Amplified warming in tropical and subtropical cities at 2°C climate change. Proceedings of the National Academy of Sciences.

Keywords: urban heat island, climate change, tropical cities, subtropical cities, machine learning, climate projections, urban warming, medium-sized cities, heat stress, global warming, adaptation, urban climatology

The Coupled Model Intercomparison Project Phase 6 (CMIP6) serves as the cornerstone for this research. CMIP6 provides a standardized framework for the evaluation of climate models and their projections, contributing to our understanding of climate dynamics. With numerous General Circulation Models (GCMs) included in this framework, assessing their relative performance regarding precipitation forecasting is vital. The significance of this ranking process is underscored by the vital role precipitation plays in water resource management, especially in a country like India, where agriculture heavily relies on monsoon patterns.

Machine learning stands out as a transformative tool in this study, significantly enhancing the capacity to analyze and interpret complex datasets. By integrating machine learning algorithms, the study benefits from heightened predictive accuracy and efficiency. For the Ujjani Dam catchment, the researchers were able to create more reliable models that mirror historical precipitation patterns while accounting for future climatic variations. This innovative approach opens new avenues for hydrological forecasting and risk management, particularly in disaster-prone regions.

Their work also highlights the difference between the SSP245 and SSP585 scenarios. The Shared Socioeconomic Pathways (SSPs) provide narrative frameworks for understanding future socio-economic developments and their impacts on greenhouse gas (GHG) emissions. SSP245 reflects a world where efforts are made to mitigate climate change, while SSP585 represents a scenario with high emissions and limited intervention. Understanding the implications of these scenarios is crucial for policymakers when formulating climate resilience strategies.

One of the remarkable aspects of this research is the formulation of multi-model ensembles. By combining outputs from various GCMs, the researchers enhance the robustness of their precipitation projections. This ensemble approach captures the uncertainty inherent in climate models, offering a more comprehensive perspective than relying on a single model. The various models bring different strengths and weaknesses, thus creating a balanced view of future precipitation trends in the Ujjani Dam catchment.

The findings from this research are not only scientifically significant but also have practical implications. For countries like India, which are highly vulnerable to climate variability, understanding precipitation trends is key to ensuring water security and food production. The analysis performed by Venkatesh and Kale could provide critical insights for irrigation planning, agricultural adaptation, and disaster risk management. Such analysis can empower stakeholders, including government authorities, farmers, and local communities, to make informed decisions based on recent projections.

Additionally, the ranking of CMIP6 GCMs provides groundwork for future research endeavors. By identifying the most reliable models for precipitation forecasting, subsequent studies can focus on refining projections and exploring additional environmental impacts. This research underlines the necessity of continuous evaluation and enhancement of climate models, which are indispensable for understanding climate change and facilitating adaptation strategies.

The implications of climate change extend far beyond precipitation. Research such as this correlates various climate indicators and considers how they interact. This holistic understanding is precisely what is needed to combat climate challenges effectively. As climate science evolves, so too must the methodologies employed by researchers, blending traditional climate analysis with innovative technologies like machine learning.

In summary, the research conducted by Venkatesh and Kale provides a crucial contribution to our understanding of climate variability and its implications, particularly in relation to precipitation projections in India’s Ujjani Dam catchment. Navigating the complexities of climate models through a rigorous, data-driven methodology unveils critical insights essential for future planning in regions facing the repercussions of climate uncertainty. By prioritizing transparency in model ranking and leveraging multiple forecasting techniques, the transformative potential of this research can be realized in real-world applications.

As more studies emerge in this field, it becomes increasingly evident that interdisciplinary cooperation will drive future advancements in climate science. Integration of diverse perspectives, including meteorology, data science, and environmental policy, is essential to address the multifaceted challenges presented by climate change. The path forward resides in embracing innovative research paradigms and navigating the intricate web of climate systems, with the ultimate goal of achieving a sustainable future for all.

As we continue to confront the realities of a changing climate, the importance of research like that conducted by Venkatesh and Kale cannot be overstated. Their efforts contribute a vital layer of understanding that is not only of academic relevance but also of immense practical importance in a world increasingly impacted by climatic shifts. As we move deeper into the 21st century, the intersection of machine learning with climate research holds promising potential that could redefine our approaches to environmental sustainability and resilience.

Through this lens, the fight against climate change can be reframed not as a daunting challenge but as an opportunity for innovation and collaborative action. The findings of this research are a clarion call for continued investment in climate science and adaptive strategies that prioritize both environmental integrity and human livelihoods.

Subject of Research: Precipitation projections under SSP245 and SSP585 scenarios in Ujjani Dam catchment, India.

Article Title: Ranking of CMIP6 GCMs and formulation of multi-model ensembles for precipitation projection under SSP245 and SSP585 scenarios over the Ujjani Dam catchment in India by using machine learning and conventional methods.

Article References:

Venkatesh, J., Kale, G.D. Ranking of CMIP6 GCMs and formulation of multi-model ensembles for precipitation projection under SSP245 and SSP585 scenarios over the Ujjani Dam catchment in India by using machine learning and conventional methods.
Environ Sci Pollut Res (2025). https://doi.org/10.1007/s11356-025-36853-y

Image Credits: AI Generated

DOI:

Keywords: Climate Change, Machine Learning, CMIP6, Precipitation Projections, SSP245, SSP585, Ujjani Dam, Water Resource Management, Multi-Model Ensembles.

The research, led by Eun Jeong Cha and Chi-Ying Lin of the University of Illinois, employs advanced machine learning models to simulate the evolving impacts of hurricanes under various climate futures. By focusing on wooden single-family homes reinforced with concrete masonry across eight states—Texas, Louisiana, Mississippi, Alabama, Florida, Georgia, South Carolina, and North Carolina—the team crafted a finely tuned assessment of vulnerability and potential economic loss. Their methodology integrates atmospheric physics, material degradation models, and socioeconomic data, enabling a nuanced understanding of how intensifying winds and heavier precipitation increase structural risk to residential properties.

Much of the recent destructive hurricane activity in the Southeast has foreshadowed the projected trends. Catastrophic storms like Hurricane Irma in 2017 and the recent Hurricane Helene in 2024 inflicted widespread devastation. Helene’s event stands out, especially, with total estimated damages exceeding $78 billion—a figure encompassing both wind destruction and unprecedented flooding in western North Carolina. The severity of such events underscores the urgent need to recalibrate disaster preparedness and insurance frameworks to account for evolving climatic stressors.

The study focuses on what climatologists term the "worst-case scenario," characterized by the Intergovernmental Panel on Climate Change’s (IPCC) RCP8.5 pathway. This high-emission model assumes continued reliance on fossil fuels and negligible global climate policy intervention, leading to significant increases in atmospheric greenhouse gases. Under RCP8.5, global mean surface temperatures are projected to rise approximately 2.0° Celsius by 2065 and upwards of 3.7° Celsius by 2100 when compared to the baseline period of 1986-2005. This severe warming is expected to escalate hurricane intensity through enhanced ocean heat content and atmospheric moisture, factors that jointly fuel storm strength and precipitation extremes.

Cha explains that using RCP8.5 as a stress test scenario provides critical insights into how worst-case climate futures might reshape hurricane risk dynamics. “Our models project that the combination of stronger winds and increased rain ingress will substantially amplify residential property losses,” she notes. Their simulations show a range of 49 to 76 percent increase in combined wind and rain damage losses by 2060; by 2100, this range widens to a staggering 71 to 102 percent. Notably, these projections signal not only amplified storm intensity but also compounded vulnerabilities arising from aging infrastructure and demographic shifts.

Among the eight states analyzed, Texas emerges as the region expected to face the highest upward shift in hurricane wind speeds, with average increases predicted at 14 percent by the 2050s relative to current levels. This uptick in wind intensity translates directly into heightened structural risk and financial losses. Following Texas, the tri-state region encompassing Louisiana, Mississippi, and Alabama exhibits similarly large increases in expected damages, reflecting their extensive coastal exposure and historical hurricane frequency.

Interesting insights also arise at smaller geographic scales within states. The analysis highlights that some inland counties, traditionally less exposed to storm damage, could experience disproportionately large percentage increases in hurricane risk. Charleston County, South Carolina, serves as a prime example where projected climate scenarios suggest a notable spike in vulnerability. Factors such as limited past exposure combined with infrastructure not designed for the rapidly intensifying wind and rain conditions contribute to this heightened future risk. These localized findings challenge conventional understanding that mainly coastal regions face the gravest hazard, prompting a reassessment of regional preparedness and resilience planning.

This spatial variability in risk underscores the importance of detailed, county-level assessments for effective resource allocation and policy formulation. Cha emphasizes, “Climate change does not impact all areas uniformly, even within the same state. Our findings reveal the necessity of region-specific, granular risk evaluations to inform federal and state mitigation strategies, especially as populations grow in hurricane-prone zones.” Such refined data is critical for prioritizing structural retrofits, enhancing building codes, and directing emergency management efforts to vulnerable communities.

An important addition to this study is its incorporation of rain ingress as a significant damage factor, alongside wind-driven losses. While wind damage has historically garnered more attention in storm risk models, the increasing prevalence of heavy rainfall events associated with hurricanes demands explicit consideration. Water intrusion through compromised building envelopes during intense storms contributes substantially to repair costs, mold growth, and long-term structural degradation, often compounding the overall impact beyond wind damage alone. Cha remarks that many existing insurance models underestimate these effects, leading to potential underpricing and insufficient risk coverage.

Complementary research by Yue Shi, a PhD candidate at the Norwegian School of Economics, aligns closely with these findings. Shi’s recent study published in Risk Analysis investigates the correlation between extreme rainfall and insurance claims, finding that claims surge with increasing precipitation intensity and vary widely across geographies. Both studies converge on the conclusion that as climate change progresses toward wetter storm conditions, insurance industries must adapt their models to integrate evolving risk profiles realistically and equitably.

The implications of these insights for urban planners, policymakers, and insurers are profound. Accurate quantification of future hurricane hazards and their economic repercussions is vital for devising effective mitigation strategies and ensuring community resilience. The research presented by Cha and colleagues enriches the scientific foundation necessary to guide policy decisions, improve building codes, and develop innovative insurance products better calibrated to a transforming climate regime.

The Southeast’s ongoing demographic expansion and urbanization amplify these challenges, as increasing numbers of residents and assets become exposed to intensifying hurricane hazards. Coupled with aging infrastructure and limited historical precedence for such extreme conditions, communities face compounded vulnerabilities. Investments in resilient construction practices, early-warning systems, and emergency response capabilities will be indispensable to reduce future losses and protect public well-being.

As climate change propels storms toward unprecedented intensities, this study’s machine learning-driven approach marks a significant advance in risk assessment. By melding climatology, engineering, and economic analysis, it sets a new benchmark for understanding and anticipating the multifaceted threats posed by future hurricanes. These developments herald a critical juncture for science-informed decision-making aimed at safeguarding the coastal populations most at risk.

The ongoing dialogue between climate scientists, engineers, and insurance firms cultivated through such research endeavors will be crucial for fostering adaptive resilience strategies. Only through collaborative effort and informed policy can the escalating threat of hurricanes in a warming world be effectively managed, reducing human and economic suffering in the decades ahead.

Subject of Research: Impact of climate change on hurricane wind risk and associated economic losses in the Southeastern United States.

Article Title: Evaluating the impact of climate change on hurricane wind risk: A machine learning approach

News Publication Date: 21-May-2025

Web References:

• Risk Analysis Journal Article