Climate change is often reduced to two core phenomena: rising temperatures and sea levels. However, as we've seen in the past few years, climate change is far-reaching in how it induces more extreme weather: more frequent heat and cold waves, atmospheric rivers, and hurricanes. Each one of these events affects a variety of stakeholders ranging across agriculture, city planning, energy grids, supply chain management, and more.
In this article, we take a closer look at heat waves, which may occur due to a variety of reasons. Heat waves can wreak havoc on infrastructure and pose severe health risks, especially for the young and elderly. Accurately forecasting heat waves, therefore, is crucial to ensuring that cities and governments can take necessary precautions to ensure resilience against the damages brought on by extreme weather.
2022 Western U.S. Heat Wave
In the late summer of last year, a long-lasting heat dome settled around the western U.S. The heat dome prevented moist air from the Pacific from cooling the land down, resulting in historically high temperatures, including a global temperature high at Death Valley (127°F).
Existing physics-based forecasts, such as CFSv2, are generally pretty good at forecasting the signs of extreme weather: they can tell you whether to expect abnormally warm or cold weather. However, they're not as great at predicting the magnitude of extreme weather, which is important for events such as the 2022 heat wave.
At Vāyuh.ai, we've used the best of both AI and physics-based models to improve forecasting. This approach pays dividends in forecasting both the direction and magnitude of extreme events.
Table 1: Temperature predictions and observations (columns) across several locations (rows) for a 28-day forecast. Columns denote the Vāyuh.ai predicted anomalies (left), CFSv2 predicted anomalies (middle), and ground truth anomalies (right).
For example, on July 26, 2022, cities across the West and mid-west experienced positive temperature anomalies, where the daily temperature exceeded that of the historical averages. The Pacific Northwest experienced temperatures more than 10 degrees above historical averages, while Texas saw temperature 5-6 degree anomalies (Table 1: right column). A month in advance, CFSv2 failed to pick up on the severity of the heat wave (Table 1: middle column). Vāyuh.ai models, however, were already picking up on the fact that a heat wave might occur (Table 1: left column).
Table 2: Temperature predictions and observations (columns) across several locations (rows) for a 28-day forecast. Columns denote the Vāyuh.ai predicted anomalies (left), CFSv2 predicted anomalies (middle), and ground truth anomalies (right).
Vāyuh.ai's ability to capture the heat wave improved closer to the forecast date. At 14 days in advance, both CFSv2 and Vāyuh.ai picked up on the severity of the heat wave in the Pacific Northwest. However, Vāyuh.ai's accuracy included Texas, which CFSv2 failed to capture (Table 2). Leveraging both physics and AI in a predictive model improves forecasting of extreme anomalies.
As we move towards summer, more warming anomalies are expected. How will you be impacted by them? The forecast above was put in actual service by a global top 5 bank to help them with their understanding of the grids. What would you do with the world's best medium-term weather forecast?
If you are a potential customer, feel free to reach out to us at email@example.com.