Periodic Reporting for period 1 - HEAT (Hybrid dry–hot Extremes prediction and AdapTation)
Período documentado: 2023-05-01 hasta 2025-10-31
Half a million people die due to heat stress every year. These numbers keep rising as the climate continues to change. Heatwaves are becoming more frequent and severe, and increasingly synchronised with droughts. Droughts reduce the ability of the land surface to cool down via evaporation, further enhancing heatwave temperatures.
How these compound drought–heatwave events spatially propagate, and how deadly they will be in the future, remains unclear. Counterintuitive findings now indicate that drought can even dampen heatwave deadliness by reducing air humidity.
As a result, our ability to forecast dry–hot events and their impacts on mortality remains limited. Subseasonal timescales, between two weeks and two months, have traditionally been a blind spot: conventional weather forecast models are not tailored to these scales. However, the adoption of Artificial Intelligence (AI) may hold the key to filling this gap and reliably predicting the occurrence of heat stress episodes weeks in advance. This would bring enormous societal benefits by enabling early warnings and emergency planning.
In this project, we explore an innovative way to generate subseasonal forecasts of droughts and heatwaves, and their resulting human heat stress. A hybrid approach will be adopted, combining physics-based models with AI algorithms. Building on this framework, we will deepen our understanding of the climatic drivers of heat stress and evaluate the potential of land-based adaptation strategies to reduce its impacts. These strategies include afforestation, crop selection, and large-scale irrigation.
Altogether, HEAT will foster our preparedness and resilience to future heat stress episodes by improving their prediction, investigating the mechanisms that trigger them globally, and identifying realistic and effective land adaptation strategies to mitigate them.
How these compound drought–heatwave events spatially propagate, and how deadly they will be in the future, remains unclear. Counterintuitive findings now indicate that drought can even dampen heatwave deadliness by reducing air humidity.
As a result, our ability to forecast dry–hot events and their impacts on mortality remains limited. Subseasonal timescales, between two weeks and two months, have traditionally been a blind spot: conventional weather forecast models are not tailored to these scales. However, the adoption of Artificial Intelligence (AI) may hold the key to filling this gap and reliably predicting the occurrence of heat stress episodes weeks in advance. This would bring enormous societal benefits by enabling early warnings and emergency planning.
In this project, we explore an innovative way to generate subseasonal forecasts of droughts and heatwaves, and their resulting human heat stress. A hybrid approach will be adopted, combining physics-based models with AI algorithms. Building on this framework, we will deepen our understanding of the climatic drivers of heat stress and evaluate the potential of land-based adaptation strategies to reduce its impacts. These strategies include afforestation, crop selection, and large-scale irrigation.
Altogether, HEAT will foster our preparedness and resilience to future heat stress episodes by improving their prediction, investigating the mechanisms that trigger them globally, and identifying realistic and effective land adaptation strategies to mitigate them.
During the first phase of the project, HEAT made important scientific and technical progress toward improving the prediction and understanding of hazardous heat and drought events. The work focused on three main areas: developing new forecasting tools, investigating what drives these compound extremes, and preparing climate simulations to test solutions for reducing their impact.
A key achievement has been the development of a new global dataset that tracks how heat and moisture move through the atmosphere during extreme weather. This allows the team to analyse how dry and hot conditions emerge and spread across regions. At the same time, we have reviewed and tested several artificial intelligence (AI) models that may help forecast these events weeks in advance, something current weather forecast systems struggle with.
To better understand the underlying causes of extreme events, we studied how land conditions, such as dry soils or stressed vegetation, interact with the atmosphere. Using global and ecosystem-scale data, we identified regions where these land–climate feedbacks are particularly strong, and began applying new methods to detect causal relationships. This work helps explain why some regions are more vulnerable than others, and which environmental signals could serve as early warnings.
We also made progress on the question of how land use changes — such as planting forests or changing irrigation — could help mitigate heat and drought impacts in the future. To do so, we connected a state-of-the-art global climate model with our tool to trace the flow of heat and moisture in the air. Initial test simulations are already running and will be expanded in the next phase of the project.
Two major scientific papers connected to HEAT were published in Science and Nature. One showed that drylands can accelerate their own expansion by disrupting downwind rainfall, while the other revealed that future dry spells may last longer than previously expected. These results confirm the importance of land processes in shaping climate extremes and highlight the need for better forecasting and adaptation tools, confirming the relevance of the activities planned for the second phase of the project.
A key achievement has been the development of a new global dataset that tracks how heat and moisture move through the atmosphere during extreme weather. This allows the team to analyse how dry and hot conditions emerge and spread across regions. At the same time, we have reviewed and tested several artificial intelligence (AI) models that may help forecast these events weeks in advance, something current weather forecast systems struggle with.
To better understand the underlying causes of extreme events, we studied how land conditions, such as dry soils or stressed vegetation, interact with the atmosphere. Using global and ecosystem-scale data, we identified regions where these land–climate feedbacks are particularly strong, and began applying new methods to detect causal relationships. This work helps explain why some regions are more vulnerable than others, and which environmental signals could serve as early warnings.
We also made progress on the question of how land use changes — such as planting forests or changing irrigation — could help mitigate heat and drought impacts in the future. To do so, we connected a state-of-the-art global climate model with our tool to trace the flow of heat and moisture in the air. Initial test simulations are already running and will be expanded in the next phase of the project.
Two major scientific papers connected to HEAT were published in Science and Nature. One showed that drylands can accelerate their own expansion by disrupting downwind rainfall, while the other revealed that future dry spells may last longer than previously expected. These results confirm the importance of land processes in shaping climate extremes and highlight the need for better forecasting and adaptation tools, confirming the relevance of the activities planned for the second phase of the project.
The HEAT project is advancing the frontier of climate science by targeting one of its most pressing challenges: how to better predict and reduce the impacts of dangerous heat and drought events in a warming world. The project stands out for its integrated approach, combining physical climate models, artificial intelligence, land surface data, and health-relevant indicators to develop new predictive tools and adaptation strategies.
A central innovation is the development of a hybrid forecasting system for compound extremes. Traditional weather models lose accuracy beyond a 10-day forecast horizon. HEAT aims to overcome this limitation by merging physical understanding of the atmosphere with machine learning methods that can extract hidden patterns from observational data. This hybrid system is designed to improve forecasts at so-called subseasonal timescales—between two weeks and two months—where early warnings could save lives and reduce economic losses.
Another novel element is the use of causal inference techniques to disentangle the drivers of hot and dry extremes. While earlier studies have relied on correlations, HEAT uses advanced statistical methods to identify whether land conditions such as soil moisture and vegetation stress actively contribute to the intensification of extreme events. This knowledge will help identify high-risk regions and guide the development of early-warning indicators that are physically meaningful.
HEAT also pioneers the assessment of land-based adaptation strategies to dampen the severity of compound dry–hot events. Using a state-of-the-art Earth system model, we will simulate how interventions like afforestation, irrigation, or crop changes could influence regional climate and reduce the severity of heat and drought events. Importantly, these simulations account not only for local effects but also for downstream impacts, offering a realistic picture of how land management can shape the atmosphere at larger scales.
By the end of the project, HEAT will deliver:
– A prototype forecasting tool for predicting compound heat and drought events weeks in advance;
– New scientific insights into how land–atmosphere interactions affect the frequency, intensity, and spread of extreme events;
– A suite of climate simulations evaluating the effectiveness of nature-based and agricultural adaptation strategies;
– Policy-relevant knowledge to support climate resilience planning in vulnerable regions.
Taken together, these outcomes will help societies move from reactive crisis management toward proactive preparedness and climate adaptation.
A central innovation is the development of a hybrid forecasting system for compound extremes. Traditional weather models lose accuracy beyond a 10-day forecast horizon. HEAT aims to overcome this limitation by merging physical understanding of the atmosphere with machine learning methods that can extract hidden patterns from observational data. This hybrid system is designed to improve forecasts at so-called subseasonal timescales—between two weeks and two months—where early warnings could save lives and reduce economic losses.
Another novel element is the use of causal inference techniques to disentangle the drivers of hot and dry extremes. While earlier studies have relied on correlations, HEAT uses advanced statistical methods to identify whether land conditions such as soil moisture and vegetation stress actively contribute to the intensification of extreme events. This knowledge will help identify high-risk regions and guide the development of early-warning indicators that are physically meaningful.
HEAT also pioneers the assessment of land-based adaptation strategies to dampen the severity of compound dry–hot events. Using a state-of-the-art Earth system model, we will simulate how interventions like afforestation, irrigation, or crop changes could influence regional climate and reduce the severity of heat and drought events. Importantly, these simulations account not only for local effects but also for downstream impacts, offering a realistic picture of how land management can shape the atmosphere at larger scales.
By the end of the project, HEAT will deliver:
– A prototype forecasting tool for predicting compound heat and drought events weeks in advance;
– New scientific insights into how land–atmosphere interactions affect the frequency, intensity, and spread of extreme events;
– A suite of climate simulations evaluating the effectiveness of nature-based and agricultural adaptation strategies;
– Policy-relevant knowledge to support climate resilience planning in vulnerable regions.
Taken together, these outcomes will help societies move from reactive crisis management toward proactive preparedness and climate adaptation.