Fire Detection as a Proxy for Combat: The Economist

Key Takeaways

• Satellite fire data quantifies combat intensity across Ukraine
• Machine‑learning filter isolates war‑related fires with 95% consensus
• Dataset reveals artillery, drone, missile strikes beyond front lines
• Russian casualties estimated hundreds of thousands to over one million

Summary

The Economist used NASA’s FIRMS satellite fire detection system combined with a 100‑model machine‑learning filter to identify war‑related heat signatures in Ukraine. By requiring 95 of 100 models to flag an anomaly, the approach isolates artillery, drone and missile strikes with high confidence. The resulting dataset records 3,368 combat‑related fires in the past month and estimates that Russia controls roughly 20% of Ukrainian territory, having seized an additional 12.8% since the invasion. Analysts also link fire intensity trends to Russian casualty estimates ranging from several hundred thousand to over one million.

Pulse Analysis

The Economist’s new combat proxy leverages NASA’s FIRMS satellite system, which sweeps Ukraine twice daily, capturing thermal anomalies that signal explosions or artillery impacts. By training a hundred machine‑learning models on pre‑war fire patterns and demanding a 95‑model consensus before labeling a fire as war‑related, the methodology filters out civilian or agricultural heat sources. This rigorous statistical filter creates a standardized, continuously updated dataset that can be compared across regions and timeframes, offering analysts a reliable alternative to contested on‑the‑ground reports.

The dataset’s early findings paint a nuanced picture of the conflict’s geography. In the last thirty days, 3,368 war‑related fires were logged, spreading across both front‑line zones and rear areas, suggesting that strikes are no longer confined to traditional battle lines. The analysis estimates Russian forces now control about 20% of Ukrainian territory, having added roughly 12.8% since the invasion began, while Ukraine reclaimed 93 square kilometres in the past month. By correlating fire spikes with casualty models, analysts infer Russian losses in the hundreds of thousands to over one million, providing a rare quantitative glimpse into the human cost of the war.

Despite its strengths, the fire‑based approach has limitations. Cloud cover and the twice‑daily satellite pass can miss smaller or obscured events, and not every combat action generates a detectable heat signature. Nonetheless, the open‑source code and data on GitHub invite further refinement and integration with other intelligence streams. As satellite monitoring becomes more affordable and algorithms improve, fire detection could become a cornerstone of transparent, data‑driven conflict analysis, aiding governments, NGOs, and investors in assessing risk and humanitarian needs.