Observed June temperature predicted Alaska natural-fire area in 2006-2025 better than a fixed historical mean

We tested whether the June-temperature response in Duffy et al. (2005) held in later Alaska fire years. We fitted simple log-area models in 1950-2003, fixed their coefficients, and tested 2006-2025. Current BLM fire points gave annual natural-fire area for fires greater than 50 hectares. NOAA statewide June temperature was a proxy for the original station mean. June temperature reduced squared log error by 68% against the old mean (95% three-year block interval 48% to 87%). Slopes were 0.638 and 0.609 log hectares per degree F in the two periods. Skill was 55% against a trailing ten-year mean, but its interval included zero. An interior-temperature check gave 61% skill against the fixed mean. Winter January-February PDO alone gave -7% skill (-24% to +11%); adding it to June gave 63%. An added decomposition showed 48% skill even against the later mean known after the test, so the gain was not only a mean shift. These are tests conditional on observed June weather, not early-season operational forecasts. Limits include 20 test years, current rather than historical data versions, different spatial inputs, natural ignition as a lightning proxy, uncertain older fire records and no test of the full published equation. Code, source hashes, aggregate results and an original figure are provided.