Atmosphere Watch 2026: Extreme Heat and Transboundary Fires Expose Billions to the 'Ozone-Climate Penalty'

 In the first eight months of 2026, a quiet environmental crisis unfolded as extreme heat and human-caused air pollution converged to expose nearly 40% of the global population to dangerous levels of ground-level ozone. 

Unlike primary pollutants emitted directly from tailpipes or smokestacks, this harmful secondary pollutant formed when precursor emissions—such as nitrogen oxides and volatile organic compounds from transport, industry, domestic solvents, and vegetation—cooked under intense sunlight during severe heatwaves. 

Known as the “ozone-climate penalty,”this feedback loop exacerbated air quality risks across entire continents. Across the planet, only 5% of the world’s population lived in areas that maintained “Good” ozone levels within World Health Organization (WHO) guidelines.

Asia emerged as the hardest-hit continent, where heavy industrial emissions, transport, and natural precursors exposed 60% of the population to extremely high ozone levels, rising to 90% when including "Poor" and "Very Poor" air quality categories. 

In Europe, repeated summer heatwaves drove 50% of the population into poor, very poor, or extremely poor ozone exposure despite long-standing regional air directives. Meanwhile, across North America, extreme wildfires in Canada’s Northwest Territories and Ontario sent massive smoke plumes thousands of kilometers across the continent. 

In July 2026, just days before the FIFA World Cup final on July 19, dense wildfire smoke blanketed major metropolitan areas including New York and New Jersey, prompting urgent public health advisories before rain helped clear the air.The atmospheric impact demonstrated that air pollution respects no international boundaries. 

Canadian wildfire emissions drifted across the Atlantic into Greenland, the Arctic, and Europe, raising fine particulate matter ($\text{PM}_{2.5}$) concentrations thousands of miles from their origin. In Southeast Asia, a powerful El Niño accelerated seasonal fires across Indonesia’s West and Central Kalimantan on the island of Borneo, engulfing millions of people in hazardous pollution by mid-August.

 Concurrently, massive plumes of Saharan desert dust crossed the Atlantic, casting a dense haze over Cape Verde, the Canary Islands, and the Caribbean. Tracking over 50 chemical species via satellite networks and global forecasting models, international researchers demonstrated that local health risks are inextricably linked to transboundary climate events.

Source & Data Attribution: Grounding material derived from the global report Atmosphere Watch (2026), published by the Copernicus Atmosphere Monitoring Service (CAMS) on behalf of the European Commission and operated by the European Centre for Medium-Range Weather Forecasts (ECMWF).

AI Is Sending Billions of Visits to the Web, but Only a Few Sites Are Getting Most of the Traffic

 As artificial intelligence becomes part of how people search for information online, it is also changing where those users go next. An analysis by OneLittleWeb found that AI sources generated 6.3 billion referred visits to 10,171 websites between July 2025 and June 2026, but the traffic was heavily concentrated among a small group of platforms.

OpenAI received the largest share, with 1.8 billion AI-referred visits, equivalent to 28.16% of the traffic measured in the analysis. Google was second with 1.3 billion visits, or 21.25%.

The figures suggest that AI is not simply replacing traditional search engines. It is also sending users back to them. Google recorded the largest increase among the top 10 destinations, gaining 929.6 million AI-referred visits during the period.

YouTube was the third-largest destination, receiving 307.8 million visits, followed by GitHub with 116.8 million. ChatGPT received 114.1 million AI-referred visits.

OneLittleWeb treats OpenAI.com and ChatGPT.com as separate domains. When the two are considered part of the same OpenAI ecosystem, however, they received about 1.91 billion AI-referred visits, or almost 30% of the measured market.

Together, Google, OpenAI and ChatGPT accounted for about half of the AI-referred traffic in the analysis.

The list also shows that AI-driven discovery is reaching beyond search and technology platforms. Facebook received 104.4 million AI-referred visits and WhatsApp 86 million.

One of the more notable destinations was the US National Institutes of Health (NIH), which received 73.8 million AI-referred visits. Wikipedia received 72 million, while Reddit recorded 70.5 million.

The NIH figure is particularly interesting for health and science communicators because it suggests that authoritative research and health information sources can become destinations after people encounter information through AI.

But the data also point to a concentration problem. The top 10 websites accounted for 64.48% of the AI referral trafficmeasured by OneLittleWeb.

This means that as AI becomes another gateway to information, a relatively small number of large platforms are capturing a substantial share of the traffic generated by that new pathway.

There is also an important distinction between AI-generated answers and AI-referred traffic. The analysis measures visits that arrive at websites from AI sources. It does not measure all the times an AI system mentions a website or uses its information without sending the user to that site.

That distinction matters because one of the biggest questions around AI search is whether websites will continue to receive visitors when users increasingly get their answers directly from AI systems.

The OneLittleWeb analysis provides one indication that AI is currently creating substantial referral traffic, but it also shows that the benefits are not evenly distributed.

For publishers, researchers, health organisations and other information providers, the emerging question is therefore not simply whether people are using AI to search. It is which sources AI sends people to, and what makes those sources worth visiting after an AI-generated answer.

Source: OneLittleWeb, Most Visited Websites in the World analysis, covering July 2025–June 2026. The analysis tracks 10,171 websites and uses estimated traffic data from SEMrush, Ahrefs and Aitools.xyz.

Uranium is back in demand. What does that mean for the future of nuclear power?

 For much of the past decade, uranium mining was not exactly a booming business.

Now the picture is changing.

As countries reconsider nuclear power in the face of concerns about climate change and energy security, the mineral at the heart of conventional nuclear reactors is attracting renewed attention.

A new global assessment by the Nuclear Energy Agency and the International Atomic Energy Agency shows that the world has more uranium than previously estimated. Identified recoverable conventional resources now stand at more than 8.1 million tonnes.

But having uranium underground is not the same as having enough uranium available when nuclear reactors need it.

That is where the story gets interesting.

A small number of countries dominate uranium production

Uranium production is concentrated in a handful of countries.

In 2024, the world produced 61,924 tonnes of uranium from mines.

Kazakhstan was by far the largest producer, accounting for 38% of global production, followed by Canada with 23% and Namibia with 12%. Uzbekistan and Australia each accounted for about 8%.

Together, the five largest producers supplied nearly 90% of the world's uranium production. 

The concentration is even more striking when looking at resources.

Australia holds about 29% of identified recoverable uranium resources at the lower-cost threshold, followed by Kazakhstan with 14% and Canada with 9%. Together, the three countries hold about half of the world's identified resources at that cost level. 

This concentration matters because nuclear reactors require a reliable supply of uranium fuel over many years.

Uranium does not automatically become nuclear fuel

Uranium coming out of a mine cannot simply be put into a nuclear reactor.

It must go through a series of processing and fuel-cycle steps before it can be used.

One important point from the latest assessment is that mine production is already close to, but does not completely cover, reactor requirements.

In 2024, mines produced enough uranium to meet about 96% of global reactor requirements. The remainder came from what are known as secondary supplies. 

These secondary supplies include stockpiles held by governments, utilities and companies, uranium recovered from enrichment processes, recycled material from spent nuclear fuel and, historically, material derived from military inventories.

But these sources cannot necessarily be relied upon indefinitely.

The assessment says secondary supplies are expected to decline as existing inventories are drawn down. Their contribution could fall to around 4% of supply by 2040. 

That puts greater emphasis back on uranium mining.

The nuclear comeback could change the equation

There are currently hundreds of commercial nuclear reactors operating around the world.

As of January 2025, the global fleet stood at 441 operating commercial reactors in 31 countries, with a combined generating capacity of about 398 gigawatts.

Those reactors generated about 2,727 terawatt-hours of electricity in 2024. 

And the nuclear industry could grow significantly.

The assessment projects global installed nuclear capacity could reach between 565 gigawatts and 916 gigawatts by 2050, depending on how quickly nuclear power expands.

That would push annual uranium requirements to between about 84,800 tonnes and 143,900 tonnes a year. 

In other words, the amount of uranium the world needs could rise substantially.

The problem is not simply whether uranium exists

The world may have enough uranium underground, but bringing that uranium into production takes time.

Developing a new uranium mine can take 15 to 20 years when planning, permitting and construction are taken into account.

The assessment warns that without new mine approvals and expansions, existing and committed mines could face supply shortfalls under high-demand scenarios. 

That creates a race of sorts.

Countries planning to expand nuclear power need to think about fuel supply well before a reactor begins producing electricity.

Not all uranium mines are the same

The way uranium is extracted also matters.

One of the most widely used methods is in-situ leaching, or ISL. Instead of removing large quantities of rock, a solution is injected underground to dissolve uranium, which is then pumped to the surface.

ISL accounted for about 55% of global uranium production in 2024, according to the assessment.

Underground mining accounted for about 26%. It is particularly important for some of the world's very high-grade deposits, including Canada's McArthur River and Cigar Lake mines. 

ISL can cause less visible surface disturbance and produces less waste rock than conventional mining. But it comes with its own environmental questions, particularly around groundwater monitoring and restoration of aquifers after mining. 

So the uranium story is also an environmental story.

Africa is already part of the uranium story

Africa is not on the sidelines of this global market.

Namibia was the world's third-largest uranium producer in 2024, supplying about 12% of global mine production. Niger and South Africa also have significant identified uranium resources. 

That raises questions for African countries with uranium deposits.

What happens to communities living near uranium mines?

Who benefits from the mineral?

What happens to the land and water after mining ends?

And as demand for nuclear fuel grows, how will countries balance the economic opportunities from uranium with environmental protection and the long-term management of mining sites?

The Red Book notes that the uranium industry is also dealing with the environmental and social aspects of mining, including the remediation of legacy sites. 

So, is there enough uranium for the nuclear future?

For now, the answer is broadly reassuring.

The assessment says identified resources are sufficient to meet projected global requirements through 2050, even under its high-growth scenario. But that does not mean supply will automatically meet demand.

New mines have to be explored, financed, permitted and built.

Existing mines have to remain productive.

Processing and enrichment capacity have to keep pace.

And countries will have to decide how much of their future energy systems they want nuclear power to provide.

Beyond 2050, the picture becomes more complicated. Under a conventional once-through fuel cycle, continued growth in nuclear power could eventually place greater pressure on known uranium resources. The report points to continued exploration, unconventional uranium resources and technologies such as fast reactors and closed fuel cycles as possible parts of the longer-term picture. 

The important point is that the nuclear conversation is no longer only about reactors.

It is also about what happens before the reactor is switched on.

What if AI could finally speak your language?

 A coalition of 60 organisations from across the artificial intelligence ecosystem has committed to helping an estimated 3.4 billion people use AI in the languages they speak and in their own voices over the next five years.

The initiative brings together AI companies, researchers, governments, philanthropic organisations, technology developers and community groups to address what the coalition describes as one of the biggest gaps in access to AI — language.

The world's roughly 7,000 languages are not equally represented in the data, tools and benchmarks used to develop and test AI systems. Only a small proportion are considered sufficiently well-resourced to support strong AI capabilities.

For speakers of languages that are poorly represented in AI systems, this can mean tools that are less accurate, less useful or less able to understand local ways of communicating.

The problem is particularly important in sectors such as health, education, agriculture, financial services and public services, where inaccurate information or poor interpretation can have consequences beyond inconvenience.

The coalition says the challenge is not simply one of translating English or other widely supported languages into local languages.

Dialect, slang, idioms and cultural context can alter meaning. Voice is also important, particularly for people who may find typing or text-based interfaces difficult or where literacy and access to written information are barriers.

The coalition's five-year goal is for an estimated 3.4 billion people who speak languages currently underrepresented in AI models to be able to use AI tools in their own language and voice.

Among the initial signatories are Amazon, Anthropic, Google, Microsoft, NVIDIA, OpenAI Foundation, Mistral, Mozilla Data Collective, UNICEF, the World Bank Group, Gates Foundation, the UK Foreign, Commonwealth and Development Office, AI Singapore, AI4Bharat, Masakhane, Data Science Nigeria and the Ministry of Telecommunications and Digital Affairs of Senegal.

The coalition says no single company, government or foundation can address the language gap alone.

Its work will focus on four areas.

The first is building an open language layer, including shared and responsibly collected language data that developers can use under open licences.

The second is developing assessments and benchmarks to measure whether AI systems are actually improving in underrepresented languages.

The third is turning language data into models and applications that can be used by AI developers, including those without the resources of the largest technology companies.

The fourth is ensuring that the expansion of AI language capabilities is carried out responsibly, with attention to privacy, consent and data sovereignty.

The coalition says local communities should play a central role in building these systems, including through local research, capacity building and ownership.

That could be particularly important in Africa, where thousands of languages and dialects are spoken but many have limited digital data available for AI development.

The statement also acknowledges that language alone will not determine whether people benefit from AI. Connectivity, electricity and access to affordable devices remain barriers to using digital technologies in many communities.

Over the coming year, the organisations say they will work together to develop the coalition's structure, governance and specific workstreams.

The initiative invites additional community organisations, researchers, technology companies, governments and funders to join.

Its stated ambition is not simply to make AI capable of translating more languages, but to make it possible for people to interact with AI in the language and voice they use in everyday life.

For health workers, farmers, teachers and communities that rely on local-language communication, the potential implications are significant. But whether the initiative can deliver useful and accurate AI in thousands of underrepresented languages will depend on the quality of the data, the involvement of communities and whether the resulting tools become available beyond the organisations building them.

August 2026 ties record as global temperatures reach 1.65°C above pre-industrial levels

 August 2026 was the joint warmest month ever recorded, tying with July 2023, as global temperatures reached 1.65°C above the estimated pre-industrial level, according to the Copernicus Climate Change Service (C3S).

The month was the first since November 2025 in which the global average surface air temperature exceeded 1.5°C above pre-industrial levels.

The new figures also point to exceptional warming of the oceans. Average sea surface temperatures over the extra-polar oceans reached a record high for August, tying with March 2024. The highest daily global sea surface temperature ever recorded was also reached in August, at 21.11°C.

C3S, which is implemented by the European Centre for Medium-Range Weather Forecasts (ECMWF), attributed the warming to human-driven climate change caused by the continued burning of fossil fuels, combined with rapidly rising sea temperatures in the equatorial Pacific linked to strong El Niño conditions.

The warming Pacific is already influencing conditions on land. In Indonesia, the combination of hot and dry conditions has contributed to an environment favourable for the intensification of seasonal wildfires, according to the Copernicus Atmosphere Monitoring Service.

Europe also experienced an exceptional summer.

Western Europe recorded its warmest summer on record, surpassing the previous record set in 2003. Heatwaves continued into August after an extended period of extreme heat that began in May.

Dry conditions affected large parts of western, central and eastern Europe, with severe drought reported in France, the United Kingdom, Hungary, Romania and Serbia. Several major rivers, including the Rhine, Danube, Southern Bug and Dnieper, recorded exceptionally low flows.

Samantha Burgess, Strategic Lead for Climate at ECMWF, said the latest figures show how climate change is increasingly driving extremes across the atmosphere and oceans.

“August 2026 was a remarkable month in the global climate record,” Burgess said. She noted that the combination of record global temperatures, exceptionally warm oceans and western Europe’s warmest summer on record was increasingly affecting communities, economies and ecosystems.

The August figures come as scientists continue to monitor the developing El Niño conditions in the tropical Pacific and their potential effects on temperatures and extreme weather around the world.

Source: Copernicus Climate Change Service (C3S), implemented by the European Centre for Medium-Range Weather Forecasts (ECMWF), September 2026. Data source: ERA5.

When AI Can See Your Brand but Still Won’t Recommend It

AI is changing the way people use search engines to discover information and businesses. Instead of simply typing a query into Google and choosing from a list of links, people are increasingly asking AI systems for direct answers and recommendations.

This creates a new challenge for businesses: being visible online is no longer enough — AI needs to understand a brand, its expertise and credibility well enough to recommend it. The big unanswered issue is measurement, because businesses may influence a customer’s decision through an AI-generated answer without knowing that it happened.

1. Is AI changing how people discover brands? AI-generated answers are becoming part of the early research process. People may ask AI for recommendations before they ever use Google, visit a website or click an advert. 

2. Is being visible enough?
No. A brand may have plenty of content but still be left out when AI produces a shortlist or recommendation. 

3. What is the “dark funnel”? It is the part of the customer journey that businesses cannot easily see or attribute. Someone can receive an AI recommendation, consider a company and even make a decision without creating the usual trail of clicks, website visits or form submissions. 

4. Can this influence be measured?
This is one of the biggest unresolved questions. How can businesses measure whether AI mentioned them, recommended them or influenced a customer. Marketers may have to look at patterns and trends rather than expect perfect attribution. 

5. What makes AI recommend one brand over another?
 Authority, credibility, clear positioning, useful content, third-party references, reviews, case studies and other trust signals. Being understood and trusted may matter as much as being ranked. 

6. Does this make PR more important?
If AI systems draw on information from across the web, credible media coverage and other independent mentions could become part of how a brand establishes authority. 

7. Does AI mean businesses should simply produce more content?
Not necessarily. Original, human content may become more valuable as generic AI-generated material fills the internet. Expertise, experience and distinctive perspectives can give content something that automated material may lack. 

8. Is AI search really a completely new game?
AI visibility is being overstated because AI answers still depend on content quality and distribution. In that view, the fundamentals of good marketing have not disappeared. 

9. What happens when AI gets it wrong?
Accuracy is another concern. AI can misunderstand or misrepresent a company, so appearing in an AI-generated answer is not automatically beneficial. The issue is not just visibility but how accurately a brand is represented. 

10. What should marketers do now?
There should be a shift from thinking only about rankings and clicks to building clear, credible and consistent authority across a brand's digital presence. 

I Wrote to Check on Him, and His Reply Reminded Me Why I Do Science Journalism

Every so often, I find myself going through my email, deleting old messages and thinking about people I have not spoken to in a while. Sometimes I stop and wonder how they are doing and, if I can, I reach out.

This time, I found myself writing to my former director, Boyce Rensberger, who led the Knight Science Journalism Fellowship when I was at MIT in 2008. It has been many years, and he is one of the many people who played a part in shaping my journey in science journalism. I simply wanted to know how he and his wife, Linda, were doing.

I wasn't expecting anything profound in his reply. But then he wrote:

“I think of you often and marvel that you are still engaged in the great work of health journalism, and in one of my favorite parts of the world.”

And then came the words that really stayed with me:

“You are one of the best investments the Knight Fellowship ever made.”

I honestly had to sit with that for a while.

Science/health journalism is not always an easy road, particularly here. There are many moments when you question yourself, wonder whether anyone is paying attention, or ask whether it is worth continuing to push for evidence, clarity and good journalism when so many other forces seem to be working against it.

I have been doing this for a long time now. Since my time at MIT in 2008, I have continued to report on health and science, train journalists, build networks (e.g www.hejnu.ug), and create spaces where scientists and journalists can talk to each other and make science understandable to the people who need it  “so that my grandmother in the village can understand it.” That is also why we started the media science cafés in 2014: to bring scientists and journalists together, break down the jargon, and make scientific knowledge more accessible to the public.

Sometimes I forget how far that journey has been. So Boyce's words came at a very good time.

A reminder that the work we do can leave a mark, even when we don't see it immediately. A reminder that someone is watching. A reminder that perhaps that investment all those years ago was not wasted. Boyce reminded me why I keep going.

Thank you, Boyce. ❤️

And thank you to the Knight Science Journalism community for believing in me back in 2008.

Still here. Still asking questions. Still trying to make science matter to people.