Regional differences in rapid evolution during severe drought

Abstract

Climate change is increasing drought intensity, threatening biodiversity. Rapid evolution of drought adaptations might be required for population persistence, particularly in rear-edge populations that may already be closer to physiological limits. Resurrection studies are a useful tool to assess adaptation to climate change, yet these studies rarely encompass the geographic range of a species. Here, we sampled 11 populations of scarlet monkeyflower (Mimulus cardinalis), collecting seeds across the plants' northern, central, and southern range to track trait evolution from the lowest to the greatest moisture anomaly over a 7-year period. We grew families generated from these populations across well-watered and terminal drought treatments in a greenhouse and quantified five traits associated with dehydration escape and avoidance. When considering pre-drought to peak-drought phenotypes, we find that later date of flowering evolved across the range of M. cardinalis, suggesting a shift away from dehydration escape. Instead, traits consistent with dehydration avoidance evolved, with smaller and/or thicker leaves evolving in central and southern regions. The southern region also saw a loss of plasticity in these leaf traits by the peak of the drought, whereas flowering time remained plastic across all regions. This observed shift in traits from escape to avoidance occurred only in certain regions, revealing the importance of geographic context when examining adaptations to climate change.

Keywords: Adaptation; Erythranthe cardinalis; climate change; dehydration avoidance; dehydration escape; flowering time; plasticity; resurrection study; specific leaf area.

Figure 1

Climatic moisture deficit (CMD) across the range of M. cardinalis during historical conditions and extreme drought. Sites are arranged by latitude and color coded by region (blue = North, orange = Center, red = South). (A) 11 populations of M. cardinalis included in this resurrection study spanning California and Southern Oregon. (B) The median and distribution of yearly CMD for each year between 1979 and 2009 for each site. Boxes represent the interquartile range, while the black line is the median. Dots are years further than 1.5 times the interquartile range. (C) Coefficient of variation for CMD experienced between 1979 and 2009. (D) CMD anomaly (CMDA) during the studied drought cycle. The single‐digit numbers represent the last digit of the year (e.g., 0 = 2010, 4 = 2014). The black line delineates the historical average (1979‐2009) for each site. Higher CMD implies a drier site, while positive CMDA implies the site is drier than the historical average.

Figure 2

Evolution of specific leaf area (SLA) and date of flowering across the range of M. cardinalis from the least to most drought‐impacted year. SLA shows (A) slight increases in the North, (B) decreases in the Centre, and (C) loss of plasticity in the South. Date of flowering shows evolution of later flowering time across the (D) North, (E) Center, and (F) South. Each point represents residuals from a mixed model including Region*Year*Treatment model and Site, Family, and Block as random effects. The lines are linear models run on the residuals with 95% confidence intervals given for both wet and dry treatments.

Figure 3

Slopes capturing the change of drought‐associated traits over time from the least to most drought‐impacted year. (A) SLA slopes vary considerably across regions and treatments and show a significant three‐way interaction (P = 0.003). Positive slopes represent the evolution of dehydration escape, while negative slopes are consistent with evolution toward dehydration avoidance. (B) Flowering time slopes are positive for most regions and water treatments, although slopes are not significantly different since the favored model is Region*Treatment + Year (P = 0.014). Positive slopes represent evolution of dehydration avoidance, while negative slopes show evolution of dehydration escape. Slopes vary much less for (C) water content (Treatment‐only model; P = 0.001), (D) carbon assimilation (Region*Treatment + Year; P = 0.06), and (E) stomatal conductance (Treatment‐only model; P = 0.055). Error bars show standard errors.