How climate affects extreme events and hence ecological population models

Extreme events significantly impact ecosystems and are predicted to increase in frequency and/or magnitude with climate change. Generalized extreme value (GEV) distributions describe most ecologically relevant extreme events, including hurricanes, wildfires, and disease spread. In climate science, the GEV is widely used as an accurate and flexible tool over large spatial scales (>105 km2) to study how changes in climate shift extreme events. However, ecologists rarely use the GEV to study how climate change affects populations. Here we show how to estimate a GEV for hurricanes at an ecologically relevant (<103 km2) spatial scale, and use the results in a stochastic, empirically based, matrix population model. As a case study, we use an understory shrub in southeast Florida, USA with hurricane‐driven dynamics and measure the effects of change using the stochastic population growth rate. We use sensitivities to analyze how population growth rate is affected by changes in hurricane frequency and intensity, canopy damage levels, and canopy recovery rates. Our results emphasize the importance of accurately estimating location‐specific storm frequency. In a rapidly changing world, our methods show how to combine realistic extreme event and population models to assess ecological impacts and to prioritize conservation actions for at‐risk populations. Extreme events are rare large deviations from average conditions (Coles 2001) that often have significant impacts. Ecological examples include hurricanes (Jagger and Elsner 2006), wildfires (Moritz 1997), and disease spread (Thomas et al. 2016). Many times in ecology, extreme events are modeled as random “catastrophes” (e.g., Lande 1993); these models only include event frequency. When intensity is also incorporated, studies often use distributions that misrepresent behavior in the tails, where the extremes occur (e.g., the lognormal in [Solow 2017]). However, other fields use a generalized extreme value (GEV) distribution to accurately model the probability distribution of extreme events. The GEV provides a natural, quantitative description of most extreme events, applied to disciplines including finance (Coles 2001), civil engineering (Gumbel 1958), hydrology (Martins and Stedinger 2000), and climatology (Gumbel 1958, Coles 2001). Although extreme events are important drivers in ecology (Parmesan et al. 2000, Gutschick and BassiriRad 2010), GEV models have been proposed or used, to our knowledge, only in three contexts: sedimentation rate in paleohydrology (Katz et al. 2005), maximum and minimum sea surface