Rocky Mountain subalpine forests now burning more than any time in recent millennia

Area burned across all ecosystems in the central Rocky Mountains increased significantly since 1984 (ρ = 0.40, P = 0.015), a trend strongly correlated with average May to September vapor pressure deficit (VPD; ρ = 0.75, P < 0.001; Fig. 1B). VPD reflects atmospheric demand for water and is well-linked to increased fire activity because of its influence on fuel aridity (1, 2, 6). The vast extent (95%) of burning since 1984 occurred in the 21st century, with 2020 alone accounting for 44% of area burned over this period (Fig. 1B and Table 1). Within the subalpine forests of our focal study area, defined by the dense network of fire history records (inset box, Fig. 1A), the 2020 wildfires were even more pronounced, accounting for 72% of the total area burned since 1984 (Table 1 and SI Appendix, Fig. S1).

While the majority of area burned in subalpine forests typically occurs in years with extreme seasonal climate conditions (20, 21)—such as the 1988 Yellowstone Fires and the 2020 fires in our study area—such conditions are occurring more frequently in the 21st century (Fig. 1B and SI Appendix, Fig. S1). In our focal study area, just 5 y account for 99% of the total area burned since 1984, with shortening gaps between extreme years: 2002, 2012, 2016, 2018, and 2020 (SI Appendix, Fig. S1). This trend foreshadows continuing increases in wildfire activity and extreme fire seasons with higher aridity in coming decades (22), as projected across Rocky Mountain forests (18, 19).

The extensive fire activity during the 21st century is unprecedented in the past two millennia in our focal study area (Fig. 2). To directly compare recent burning to paleofire records, we summarized contemporary fire activity using fire rotation periods (FRPs), defined as the time required to burn an area equal in size to the area of interest, which in this case is the total area of subalpine forests in our focal study area. By sampling across this large area, we substitute space for time to characterize contemporary FRPs. During the 21st century (2000 to 2020), the FRP in our focal study area was 117 y. The FRP from 1984 to 2020, incorporating the last 16 y of the 20th century with little fire activity, was 204 y (Table 1), within the historical range of variability defined by tree-ring and lake sediment records (Fig. 2C). Integrating published tree-ring–reconstructed FRPs from eight subregions in our focal study area (11, 15, 17, 23), the median FRP (95% CI) was 261 y (183 to 398) from c. 1630 to 1860 CE (Common Era) and 315 y (295 to 1,520) from c. 1861 to the mid-20th century CE (Fig. 2C and SI Appendix, Fig. S2); the FRPs from these periods are not significantly different (Wilcoxon rank sum statistic = 55, P = 0.195). Similarly, in the subalpine forest watersheds where lake sediments record fire history over the past two millennia, 44% experienced a fire event within any given century on average, analogous to an FRP of 230 y (Fig. 2 A and C). Therefore, the 21st century FRP of 117 y, largely because of the 2020 fire season, represents nearly a doubling of the average rate of burning over the past 2,000 y.

The 21st century FRP also exceeds the maximum rate of burning reconstructed over the past 2,000 y. Modest warming during the early Medieval Climate Anomaly (MCA, 770 to 870 CE) coincided with the maximum of 67% of sites recording fire events within a century, corresponding to an FRP of 150 y (Fig. 2C). The 21st century FRP of 117 y represents 22% more burning per century than the early MCA maximum (Fig. 2C). Fire activity over the past two millennia also broadly tracked paleotemperatures across North America (24) and the Northern Hemisphere (25) (Fig. 2 B and C), consistent with links between climate and fire seen in the contemporary record (Fig. 1B and SI Appendix, Fig. S1). Although we cannot directly compare area-burned statistics from the paleofire records to the size of contemporary wildfires, burning during the early MCA was concentrated within a subset of sites (sites 2 to 13 in Fig. 2A) spanning an area similar in size to the extent of the major 2020 wildfires (Fig. 1A and SI Appendix, Fig. S3).

Subalpine forest fire regimes are unlikely to return to the late-Holocene range of variability in this century (Fig. 2C and SI Appendix, Fig. S2C). It would take two decades with no additional burning following 2020 to return the 21st century FRP (i.e., 2000 to 2040) in our focal study area to the late-Holocene average of 230 y. Even returning to the 1984 to 2020 rate of burning for the next three decades would keep the 21st century FRP near the late-Holocene limit of 150 y (Fig. 2C and SI Appendix, Fig. S2C). The 2020 fire season thus marks the emergence of 21st century fire regimes with distinctly higher rates of burning not just from the late 20th century but relative to the past two millennia.

The primary importance of climate in enabling widespread burning, in the past and present, does not mean that nonclimatic factors are unimportant for fire activity at smaller scales. Extreme winds over hourly and daily time scales drove extraordinary growth of individual fires in 2020, with Colorado’s East Troublesome Fire crossing the fuel-barren Continental Divide. The 2020 fires also burned forests with extensive insect-caused tree morality, and while this likely altered stand-level fire behavior (26), it does not explain the West-wide pattern of increased burning in recent decades (27). Furthermore, while fire suppression and prior land uses have altered fire regimes in lower-elevation forests, long fire-free intervals (Fig. 2A) and less intensive forest management minimize these impacts in subalpine forests. Instead, the increasing magnitude and frequency of extreme moisture deficits in the 21st century (Fig. 1B), rather than increased fuel abundance, lacks precedent in recent millennia and has driven the 21st century shift in fire activity.

The combination of increased burning (18, 19) and more stressful postfire climate conditions for tree regeneration (28) in upcoming decades foreshadows the potential for widespread loss of subalpine forest resilience to wildfire (29). The extensive burning during the early MCA, for example, transformed some closed-canopy forests near the tree line into lower-density ribbon forests, a structure that persists today (30). Additionally, extensive burning during the MCA reduced the landscape connectivity of late-successional forests, likely explaining why elevated burning was not sustained even as warming continued for several centuries (Fig. 2C) (12). Such decreased forest density or connectivity could eventually create a negative feedback with fire activity (18, 29), but this is unlikely in upcoming decades. Even at the 2010 to 2020 rate of burning, it would take six decades to burn an area equal to all subalpine forests in the focal study area. Area burned will likely continue to increase before fuel limitations, other ecosystem changes, or long gaps between extreme fire years reduce subalpine forest burning to late-Holocene levels.

Rates of burning now nearly twice the late-Holocene average and exceeding the maxima of the MCA underscore the high sensitivity of regional fire regimes to climate change. From the paleoecological perspective, unprecedented burning in the 21st century is not surprising, given that Northern Hemisphere temperatures have now risen ∼0.5 °C above the MCA maximum (Fig. 2B). Our findings are also consistent with the predicted development of novel fire regimes across the Rocky Mountains by the early to mid-21st century (18, 19). The emergence of unprecedented burning by 2020 suggests that the central Rocky Mountains are following a trajectory consistent with the warmer and drier climate scenarios used in projections of future fire activity. Only by returning to the 1984 to 2020 average rate of burning would 21st century fire regimes in our focal study area realign with the highest rates of burning over the past 2,000 y, an unlikely scenario under even the most modest climate change projections (18, 19).

As the 21st century rate of burning moves beyond the range of late-Holocene variability, planning across all scales—from individuals to utility companies, municipalities to the federal government—can no longer be reasonably based on expectations from the past. The 2020 fire season serves as an example of how increasingly fire-conducive climate conditions have made Rocky Mountain subalpine forests more flammable now than at any point in recent decades or millennia, a trend expected to continue with climate warming.

We characterized contemporary fire activity and climate in the central Rocky Mountains within Bailey’s M331H and M331I ecosections (Fig. 1A). Fire perimeters were obtained from the Monitoring Trends in Burn Severity program (MTBS, https://www.mtbs.gov/) for wildfires from 1984 to 2018 (as 2019 fires were not yet available) and from the National Interagency Fire Center (NIFC, https://data-nifc.opendata.arcgis.com/) for wildfires from 2019 and 2020. To be consistent with the fire-size cut off in the MTBS dataset, we only used fires >405 ha (1,000 acres) from 2019 and 2020. Annual average May to September VPD for the study region was obtained from gridMET (31) (http://www.climatologylab.org/gridmet.html). We used Spearman rank correlation to assess trends in area burned and VPD and to compare these two time series.

To explicitly compare contemporary burning to fire activity reconstructed over the past 2,000 y, we further defined a focal study area, reflecting subalpine forests represented by the network of paleofire records. The focal study area was defined as the area of subalpine vegetation, within the ecosections noted above, within the area from 39.75 to 41.70° N latitude and 105.0 to 107.8° W longitude (Fig. 1A). Subalpine vegetation classes were derived from LANDFIRE’s Environmental Site Potential product (https://landfire.gov/; SI Appendix, Fig. S3). In defining the spatial extent of our focal study area, we balanced the need to capture enough area to characterize fire regimes over recent decades while still reflecting the forest types represented by the network of paleofire records. We used the FRP to characterize the rate of contemporary subalpine forest burning, defined as the amount of time it takes to burn an area equal in size to our focal study area: time period considered/(total area burned/size of study area).

To characterize fire history over the past several centuries in subalpine forests from within the focal study area, we utilized four published tree-ring reconstructions of past fire extent. Each study provides FRP estimates for varying study areas, eight in total: six in Rocky Mountain National Park (11, 17), one in the Park Range in northern Colorado (23), and one in the Medicine Bow Mountains in southern Wyoming (15) (Fig. 1A and SI Appendix, Fig. S3). Fire extent was reconstructed using dendrochronology to date forest age classes and identify precise fire years with fire scars. Three of the four studies (11, 15, 23) estimated FRPs for two distinct time periods, generally c. 1600s through the mid-1800s and from the mid-1800s through the early to mid-1900s. We calculated FRP statistics for the fourth study to approximate these time periods based on data presented in SI Appendix, Table S4 in Sibold et al. (21) (i.e., for 1654 to 1863 and 1864 to 2000). The precise cutoff dates among studies varied based on the oldest fires confidently reconstructed and the timing of fires in the 1800s. To generate a pooled estimate of the FRP representing the focal study area, we calculated the median FRP from among the eight FRP estimates within two time periods: c. 1650 to 1870 and 1870 to 2000. We estimated the 95% CIs around the median FRP from 1,000 bootstrapped samples.

We characterized fire history over the past 2,000 y using a network of 20 published fire reconstructions based on distinct peaks in macroscopic charcoal in high-resolution lake sediment records (12–14, 16, 32, 33). All records come from small (<10 ha) lakes surrounded by subalpine forests dominated by lodgepole pine (Pinus contorta var. latifolia) at lower elevations and Engelmann spruce (Picea engelmannii) and subalpine fir (Abies lasiocarpa) at higher elevations. Chronologies were based on ²¹⁰Pb and ¹⁴C dates, and macroscopic charcoal (>150 μm) was sampled contiguously, yielding an average c. 5 to 20 y/sample in each record. Distinct charcoal peaks were identified using the CharAnalysis program and interpreted as fire events, representing one or more fires within ∼1 to 3 km of the lake within the sample interval. We used the fire history reconstructions presented in the original publications, 19 of which were developed by the authors of the current study.

We summarized regional paleofire activity from the lake sediment records by calculating the percent of sites with fire events within overlapping 100-y periods and smoothed this time series using locally weighted regression with a 10-y window. To compare fire activity across our paleofire network to FRP statistics from tree-ring and contemporary records, we calculated a paleo-FRP following Calder et al. (12) defined as the amount of time it takes to record a total number of fire events equal to the number of sites recording. At the 100-y time intervals used to calculate the percent of sites burned, the paleo FRP is defined as the following: 100 y/% sites burned.

We thank T. Hudiburg, K. McLauchlan, and S. Dobrowski for comments. This work was funded by NSF Grants DEB-1655121 and OISE-0966472 to P.E.H. and DEB-1655189 and BCS-0845129 to B.N.S. We thank the numerous people involved in field and laboratory work associated with the original publications utilized here and the federal agencies granting research permits to perform this work.