Delayed effects of climate on vital rates lead to demographic divergence in Amazonian forest fragments

Abstract

Deforestation often results in landscapes where remaining forest habitat is highly fragmented, with remnants of different sizes embedded in an often highly contrasting matrix. Local extinction of species from individual fragments is common, but the demographic mechanisms underlying these extinctions are poorly understood. It is often hypothesized that altered environmental conditions in fragments drive declines in reproduction, recruitment, or survivorship. The Amazon basin, in addition to experiencing continuing fragmentation, is experiencing climate change‐related increases in the frequency and intensity of droughts and unusually wet periods. Whether plant populations in tropical forest fragments are particularly susceptible to extremes in precipitation remains unclear. Most studies of plants in fragments are relatively short (1–6 years), focus on a single life‐history stage, and often do not compare to populations in continuous forest. Even fewer studies consider delayed effects of climate on demographic vital rates despite the importance of delayed effects in studies that consider them. Using a decade of demographic and climate data from an experimentally fragmented landscape in the Central Amazon, we assess the effects of climate on populations of an understory herb (Heliconia acuminata, Heliconiaceae). We used distributed lag nonlinear models to understand the delayed effects of climate (measured as standardized precipitation evapotranspiration index, SPEI) on survival, growth, and flowering. We detected delayed effects of climate up to 36 months. Extremes in SPEI in the previous year reduced survival, drought in the wet season 8–11 months prior to the February census increased growth, and drought two dry seasons prior increased flowering probability. Effects of extremes in precipitation on survival and growth were more pronounced in forest fragments compared to continuous forest. The complex delayed effects of climate and habitat fragmentation in our study point to the importance of long‐term demography experiments in understanding the effects of anthropogenic change on plant populations.

The inclusion of immediate and lagged climate responses amplifies the effect of climate autocorrelation on long‐term growth rate of populations

“…Currently, few empirical studies have tested for the presence of TVR. However, previous studies searching for TVR have found evidence for them (Evers et al, 2021;Scott et al, 2021;Tenhumberg et al, 2018),…”

“…Currently, few empirical studies have tested for the presence of TVR. However, previous studies searching for TVR have found evidence for them (Evers et al, 2021; Scott et al, 2021; Tenhumberg et al, 2018), suggesting that TVR might be common. Our recent study (Evers et al, 2021) conducted a review of literature published between 1997 and 2017 and found that most demographic studies consider only a single climate timeframe: typically, the first 12 months prior to vital rate responses.…”

“…Iler et al, 2019), alpine species could be a prime example of species with TVR. The presence of below‐ground rhizomes is another physiological characteristic that has been linked to lagged climate drivers in Heliconia acuminata (Scott et al, 2021). For this species, more immediate climate responses have been observed as well (Westerband & Horvitz, 2017).…”

“…This research agenda can advance via both empirical investigations and population modelling studies. Empirically, there are still too few studies that test for the existence of TVR, perhaps because such studies require long‐term data (Evers et al, 2021; Scott et al, 2021; Tenhumberg et al, 2018; van de Pol et al, 2016). Thus, we encourage researchers with long‐term demographic data to explicitly test for TVR.…”

Antecedent effect models as an exploratory tool to link climate drivers to herbaceous perennial population dynamics data

“…However, single studies report successful results with smaller datasets. For example, a study using functional smoothing methods on a 10-year long dataset (Scott et al, 2022), and a study using a sliding-window approach (van de Pol et al, 2016) on two 15-years long individual datasets . While these results are encouraging, we did not find a similar predictive ability in our 12 datasets which were longer than 20 years.…”