findings

Findings

Overview

Recent stand-replacing wildfires in late-successional and old-growth (LSOG) forests have increased land manager interest in fire refugia, which could provide vital habitat for threatened and endangered species during a time of rapid change. The overall goal of this project was to model, map, and share information essential for the conservation of LSOG forest ecosystems in the U.S. Pacific Northwest, within a diverse co-production team of state and federal land managers. We developed statistical models of contemporary (2002-2017) fire refugia, non-stand-replacing fire (NSR), and high-severity fire based on topography, fuels, fire weather, fire behavior and climate. Independent models were built for two ecoregions (Figure 1), one encompassing the Douglas-fir/western hemlock forests of the northwestern portion of our study area and the other encompassing dry-mixed conifer forests of the eastern Cascades and Klamath-Siskiyou region. We used these models to produce probability surface maps for fire refugia, NSR, and high-severity fire under low, moderate, and extreme fire weather and fire growth scenarios. These maps and associated products provide timely information about the likely persistence, change, and loss of LSOG forests under current and future climate conditions.

CASC study area CASC probability maps

Figure 1: Map of the Fire Refugia study area showing the non-fire-prone (yellow) and fire-prone (red) ecoregions (left panel) and probability maps of fire refugia (Refugia), non stand-replacing severity (NSR), and high-severity (High) for the entire region under 10th, 50th and 90th percentile fire weather conditions.

Below are some key findings from our research.

Key lessons
  1. Contrasting ecoregional refugia dynamics: Forests in both ecoregions exhibited broad similarities in the final set of predictors and relative shape of the response function curves. For example, refugia were more likely in both ecoregions for older forests with intermediate to high biomass in lower relative slope positions that burned under milder fire weather conditions. However, our models reveal striking ecoregional differences in the patterns of fire refugia and severity probability that emerge from the unique biogeographic expressions of underlying predictors and higher dimensional variable interactions between them. Our models predicted high refugial probability for the non-fire-prone ecoregion under a range of weather conditions. This is consistent with observational evidence from fires in recent decades that fire refugia comprise an important component (almost 40%) of total burn area (Meigs and Krawchuk 2018). Together, these results suggest that there is broad potential to maintain fire refugia in moist forests of the PNW, depending on the fire weather and growth conditions under which fires occur.

    In these moist forests, refugia are maintained by topographic and vegetative conditions that promote patchy burning, safe-sites, or more fire-resistant microclimates, except under extreme fire weather conditions. In contrast, our models predicted limited, patchy areas of fire refugia for forests of the fire-prone ecoregion. Instead, NSR severity was a prominent feature of our model outputs across most fire weather conditions. Lower refugial probability in forests of the fire-prone ecoregion, where fire-resistant species are more abundant, may seem, at first, counterintuitive. However, this can be explained by different fire severity-mediated pathways to old forest development between the two ecoregions. In fire-prone forests, few areas avoid fire consistently and fire refugia are limited under all fire weather scenarios. However, NSR severity is prominent under most fire weather conditions, resulting in widespread surviving residual tree structures. These disturbance-mediated old forest pathways captured in our statistical models are consistent with old growth forest dynamics theory in each region (Franklin et al. 2002, Spies et al. 2006, 2018, Tepley et al. 2013).
    Read more
  2. Old forest and large tree habitat is an important source of refugia: Live tree biomass was one of the most consistently selected top predictor variables for our models. Fire refugia probability increased in both ecoregions for high biomass old growth forest stands. An empirical analysis of our fire refugia maps with trends in northern spotted owl nesting and roosting habitat over the period 1993-2019 show a net increase of areas mapped as high fire refugia probability and only a marginal decrease in areas predicted as moderate probability, compared with a large decrease in low refugia probability areas (R. Davis, personal communication). Thus, both modeled and empirical results suggest that old forest is an important positive feedback on fire refugia.
  3. Non-stand-replacing fire severity is an important source of large tree habitat: Our models reveal that NSR severity fire was the dominant process generating residual, post-fire structure in fire-prone ecoregions. Although our models demonstrate an important role for fire refugia in the non-fire-prone ecoregion, they also predict high NSR severity probability across both ecoregions under all fire weather and growth scenarios, including the most extreme conditions. These results suggest that NSR severity fire is a critical component of legacy live tree structure in burned forests across the PNW. Future research on the structural characteristics, habitat quality and use of forests affected by low- to moderate-severity fire is an important future research direction.
  4. Multi-decadal depressions in fire refugia probability, and increases in high-severity fire, resulting from past timber harvest: Our models showed a clear and lasting imprint of past timber harvest on fire severity probability. Particularly in the non-fire-prone ecoregion, previously harvested areas showed notable decreases in fire refugia probability, and increases in high severity probability, for several decades after harvest. This finding is consistent with other studies of high-severity risk in managed forests of the region (Zald and Dunn 2018, Evers et al. 2021), but adds an important new perspective through the joint evaluation of refugia and high severity fire. This is a critical land use legacy impact that provides context for current fire severity dynamics and can inform future fire refugia and forest management strategies.
  5. Fire management strategies can promote (or diminish) fire refugia: Our models provide clear evidence that fire refugia outcomes are strongly contingent on fire weather and burning conditions. Importantly, they identify broad windows of fire season weather, based on data from 1986-2018, where mild to moderate conditions promote high refugia and NSR probability. This highlights the risks assumed when aggressive suppression strategies are used to constrain much of the annual area burned to the most extreme weather conditions, when direct fire control fails. Instead, it suggests that more proactive use of prescribed fire or adaptive management of natural ignitions could be an important part of promoting the persistence of fire refugia in PNW forests. Intentional pre-fire planning and fire management strategies that take advantage of these opportune fire weather windows may be especially important as climate change causes increasing warm, long, and dry fire seasons.
  6. Extreme fire growth trumps extreme fire weather: Extreme surface fire weather conditions (e.g. hot, dry, and windy) and extreme fire growth events, such as major blowups documented in recent years (Lareau et al. 2018, Abatzoglou et al. 2021), are often conflated. While these phenomena are often related, their causes are complex and driven by nonlinear, cross-scale dynamics that are not yet well understood and are difficult to predict (Peterson et al. 2017, Coen 2018). Our fire weather and growth models were designed to provide alternative views of the temporal drivers of fire severity, where the latter account for eruptive fire behaviors that are difficult to capture mechanistically.

    Predictions from the fire weather and growth models were broadly congruent, especially at low to moderate extremity scenarios. However, fine-scale differences in the mapped predictions were apparent even in these scenarios and large differences existed for the extreme scenarios, with fire growth models predicting much higher severity (and lower refugia and NSR probability). The conceptual model of nested fire growth-severity relationships we developed (Fig. 10) reflects that fire severity-mediated outcomes are more sensitive to, and directly influenced by, fire growth than to fire weather. As a result of this, our fire weather models may, in fact, underestimate fire severity under blowup conditions.

    A notable example that demonstrates the utility of the fire growth models is the 2020 Labor Day fires. Fire weather conditions during the 2020 fires were substantially more extreme than anything in our model training dataset (Abatzoglou et al. 2021, Higuera and Abatzoglou 2021), which included data from 2002-2017. As a result, extreme, rare weather conditions like that which drove the 2020 fires may not be reflected in our model predictions. This is evident in the prevalence of intermediate probabilities of high-severity fire predicted by the fire weather models, even under the most extreme scenario. In contrast, the fire growth model predicted widespread high-severity fire in the non-fire-prone ecoregion under the extreme scenario that is more consistent with the observed severity patterns resulting from the 2020 fires. This is a striking result. It suggests either that similar fire behavior was captured in our training dataset, despite the lack of similarly extreme fire weather conditions, or that the fire growth model more effectively captured a threshold response in the growth-severity relationship that was useful in predicting outcomes for an unprecedented event. This decoupling of fire weather and behavior is an important operational and conceptual tool that should be informative in real-time fire management decisions, post-fire ecological assessments, and fire severity modeling.
    Read more
  7. Opportunities for fire refugia and old forest structure under extreme burning conditions: Although fire refugia extent in the non-fire-prone ecoregion was greatly reduced under extreme fire conditions, our models identify some consistent areas of refugial persistence. Under extreme fire growth, refugia are strongly constrained to valley bottoms and areas of cold air-pooling, especially in the non-fire-prone ecoregion, with areas of intermediate refugia probability extending further upslope in the fire-prone ecoregion. Biogeographic areas of moderate to high refugia probability existed in portions of the Coast Range, Olympic Peninsula, northwestern Cascades, and portions of the southeastern Cascades. Companion work to this study (Downing et al. 2021) that evaluated the drivers and biogeography of refugia in repeatedly burned areas found that a similar set of topographic factors were associated with persistent refugia. Knowledge of these fire refugia areas that may persist under more extreme burning conditions or repeated burns are a critical anchor to consider in future LSOG reserve design and management planning.
Final report

Naficy, C. E., G. W. Meigs, M. J. Gregory, R. Davis, D. M. Bell, K. Dugger, J. D. Wiens, M. A. Krawchuk. 2021. Fire refugia in old-growth forests—Final report to the USGS Northwest Climate Adaptation Center. Oregon State University, Corvallis, OR. 39 p. (pdf)

Publications
  • Downing, W. M., G. W. Meigs, M. J. Gregory, and M. A. Krawchuk. 2021. Where and why do conifer forests persist in refugia through multiple fire events? Global Change Biology 27:3642–3656. https://doi.org/10.1111/gcb.15655.
  • Meigs, G. W., and M. A. Krawchuk. 2018. Composition and Structure of Forest Fire Refugia: What Are the Ecosystem Legacies across Burned Landscapes? Forests 9:243. https://doi.org/10.3390/f9050243.
  • Meigs, G.W., Dunn, C.J., Parks, S.A., and Krawchuk, M.A. 2020. Influence of topography and fuels on fire refugia probability under varying fire weather in forests of the US Pacific Northwest. Canadian Journal of Forest Research. https://doi.org/10.1139/cjfr-2019-0406.
  • Krawchuk, M.A., Meigs, G.W., Cartwright, J., Coop, J.D., Davis, R., Holz, A., Kolden, C., Meddens, A.J.H. 2020.Disturbance refugia within mosaics of forest fire, drought, and insect outbreaks. Frontiers in Ecology and the Environment 18:235-244 Special Issue on Climate Change Refugia. https://doi.org/10.1002/fee.2190.
Resources
  • Eco-Vis: A web-based map portal for data download, exploration, and visualization.
  • Click here to view our project page on USGS ScienceBase.
  • See Outreach Materials for links to all datasets and a Manager's brief summarizing the "so what?" for integrating these data into project decisions.