Two separate 2025 studies are doing the load-bearing work here, and they count different things. One attributes 41,380 excess US deaths a year to wildfire smoke across 2011-2020 [4]. The other estimates that the record 2023 Canadian season may have shortened the lives of 87,000 people worldwide who sat under its plumes [6]. Lives shortened by an unstated amount is not the same quantity as deaths per year, and one figure covers a single season globally while the other is a decadal average for one country. Add them and you get a number that describes nothing.
Take the projection on its own terms instead. Moving from 41,380 a year to 71,420 by 2050 is an increase of 30,040 deaths a year, about 73 percent [1][2]. The baseline is the part that gets less attention: hold the 2011-2020 rate flat across its own decade and it implies roughly 413,800 excess deaths already banked [3].
The 2050 figure is conditioned on warming of 2.8 to 4.6 degrees Celsius by 2100 [5], a band wide enough that the output is a family of outcomes rather than a single one. The thing it does not tell you is how the increase splits between more smoke and more people standing in it. A larger, older US population in 2050 would raise the count with no change in the air at all. Adaptation is the other unpriced term, since filtration and indoor rescheduling both sit between an index reading and a lung.
The mechanism is less contested than the counting. Smoke exposure can trigger heart attacks, asthma attacks and premature births, and it worsens existing heart disease and lung conditions including COPD [8]. The number of large wildfires in the US West doubled between 1984 and 2015, according to NOAA [11], and Harvard's Mary M. Johnson says nobody foresees a near-term decrease: size, intensity and frequency are all expected to rise [10]. Marshall Burke of Stanford, senior author on the excess-death projection, calls smoke the fastest-growing environmental threat to health in the US [7]. Even holding warming to 1.5 C leaves fires outside tropical latitudes more frequent [18].
On timing, the February 2026 Geophysical Research Letters paper is a cleaner read than the death counts because it models one variable. Senior author Hongliang Zhang of Fudan University puts it as dryness [14], and the counterintuitive part is that a warmer atmosphere holds more water, so a place can get more total precipitation and more drought at the same time as rainfall concentrates into heavier bouts [19]. Under about 4.4 C by 2100, boreal Alaska and Canada see fire season arrive roughly a week earlier, with some far northern areas shifting by up to 23 days [15]. The Southwest goes the other way, starting a week later and ending more than a month later, which nets out to a season at least three weeks longer [17][4]. California's peak burning slides later [16].
Which brings the argument back to the index. Grand Rapids logged 616 on July 16, about 1.8 times its own previous record of 350 [2][5]. Both readings land in the same advisory bin, because anything above 301 counts as hazardous [3]. One top category is workable for a rare event. If the projection is even roughly right, the range of severity that matters for who stays indoors, and for how long, now falls entirely inside a bin that does not distinguish it.
