城市和研究人员正在使用人工智能来绘制和预测极端高温风险，但新研究表明，它被更多地用于识别人们处于危险中的位置，而非帮助官员作出应对。
Cities and researchers are using to map and predict extreme heat risk, but new research suggests it's used far more to identify where people are in danger than to help officials respond. Why it matters:
为何重要：气候变化加剧了极端高温，除非政府能够将预测转化为行动，否则更好的预测无法保护脆弱社区。
climate change intensifies extreme heat, better predictions won't protect vulnerable communities unless governments can turn them into action.
- 根据耶鲁大学2025年的一项研究，美国与高温相关的死亡人数从2000–2009年每年平均2670人上升了53%，达到2010–2020年的4000人以上。
• U.S. deaths associated with high temperatures rose 53% from an annual average of 2,670 in 2000–2009 to more than 4,000 in 2010–2020, according to a 2025 Yale study.
推动新闻：发表于《自然·气候变化》的一项研究审视了现有研究，探讨人工智能和机器学习如何被用于管理极端高温。
Driving the news: study published in examined existing research on how AI and machine learning are being used to manage extreme heat.
- 在被审查的研究论文中，98.4%聚焦于预测，59.2%聚焦于预防。相比之下，仅有10.4%聚焦于应对，3.2%聚焦于恢复，1.6%聚焦于准备——作者称这种差距为“研究与实施之间的鸿沟”。
• Of the research papers reviewed, 98.4% focused on prediction and 59.2% on prevention. That's compared with just 10.4% on response, 3.2% on recovery and 1.6% on preparedness — a disparity the authors call a "research-to-implementation gap."
- 研究主要集中在利用人工智能预测危险高温和识别处于风险中的人群，而较少关注它如何帮助官员决定下一步行动。
• The research is heavily focused on using AI to forecast dangerous heat and identify who's at risk, with far less attention on how it can help officials decide what to do next.
- “人工智能实际上能发挥价值的地方，是预测成为决策核心的时候，无论是规模还是提供更细粒度的信息，”兰德政策研究员、该研究合著者Karishma V. Patel表示。
• "Where AI can actually add value is where prediction is central to the decision-making, and that's in scale or providing more granularity," said Karishma V. Patel, a RAND policy researcher who co-authored the study. State of play:
现状：将复杂的高温预测转化为政府行动不仅需要先进的模型。
Turning sophisticated heat predictions into government action requires more than advanced models.
- 康奈尔大学研究员Jennah Gosciak告诉Axios，拥有“能够摄入多种来源数据并使其对从业者可用的稳健数据系统”才是这里的关键所在。
• Cornell researcher Jennah Gosciak told Axios that having "robust data systems" that can ingest data from a variety of sources and then make it usable to practitioners "is most of the battle here."
举例说明：谷歌6月宣布，其人工智能驱动的热弹性数据已从最初的14个城市试点扩展到全球超过50个城市，为城市规划者提供建筑级别的数据，以帮助精准开展热害干预措施。
Case in point: Google in June announced it had expanded its AI-powered Heat Resilience data from an initial 14-city pilot to more than 50 cities worldwide, giving urban planners building-level data to help target heat interventions.
- 该数据利用人工智能和高分辨率影像，绘制出单栋建筑的屋顶反射率。
• The data uses AI and high-resolution imagery to map roof reflectivity down to individual buildings.
- 谷歌表示，试点项目的数据帮助多个城市做出决策，包括制定冷屋顶条例和适应计划。
• Google said data from the pilots helped inform decisions in several cities, including cool-roof ordinances and adaptation plans.
现实检验：更好的信息并不一定意味着更好的结果。
Reality check: Better information doesn't necessarily equal better outcomes.
- 戈西亚克曾在纽约市应急管理部门担任公共利益技术研究员，她最近将更简单的热脆弱性指数与更复杂的预测算法进行了比较，用于政府的热害决策。
• Gosciak, who spent time as a Public Interest Technology Fellow with New York City Emergency Management, recently compared simpler heat-vulnerability indexes with more complex predictive algorithms for government heat decisions.
- 她指出一个重要区别在于“慢数据与快数据”。预测算法可能对与特定触发事件相关的时效性决策更有用，而更简单的指数则可能更适合长期规划。
• She described one important distinction as "slow versus fast data." Predictive algorithms may be more useful for time-sensitive decisions "tied to a specific trigger event," while simpler indexes can be better suited to longer-term planning.
- 戈西亚克补充说，更简单指数的一个优势在于，官员可以明确选择优先保护哪些脆弱人群。她目前是康奈尔大学信息科学系的博士生。
• One advantage of simpler indexes is that officials can explicitly choose which vulnerable populations to prioritize, added Gosciak, a PhD student in Cornell's Information Science Department. What's next:
接下来：下一个前沿可能不在于让热浪预测逐步变得更好，而在于将预测与干预措施联系起来。
The next frontier may be less about making heat forecasts incrementally better and more about connecting predictions to interventions.
- 帕特尔表示，人工智能的价值可能在于通过整合变化中的信号并提供“细粒度的超局部预测”，帮助官员进行“近乎实时的决策”。
• AI's value could lie in helping officials make "near real-time decisions" by integrating changing signals and providing "hyperlocal forecasting at granular levels," Patel said.
- 这可能有助于官员实时应对热害风险，并更好地理解涉及高温、空气污染、野火烟雾和干旱的复合威胁。
• That could help officials respond to heat risks in real time and better understand compound threats involving heat, air pollution, wildfire smoke and drought, she said.
结论：“重要的是，我们不仅能够构建具有人工智能能力的技术，而且该技术实际上能够解决问题或消除瓶颈，”帕特尔说。
The bottom line: "What is important is not just that we can build AI-enabled technology, but it actually solves a problem or removes a bottleneck," Patel said.