Introduction:
Considerable recent research has investigated how species’ ecological niches evolve over time. Most of these studies have examined whether and at what speed niches evolve in speciating lineages (e.g., Peterson et al. 1999; Graham et al. 2004; Knouft et al. 2006; Losos 2008; Pearman et al. 2008; Evans et al. 2009; Vieites et al. 2009; Nyári and Reddy 2013; Owens et al. 2017; García-Navas and Rodríguez-Rey 2018). Methods for estimating fundamental ecological niches (Peterson et al. 2011; Hijmans and Elith 2015) and inferring macroevolutionary patterns from phylogenies (Swofford and Maddison 1987; Lanyon 1993; Freckleton et al. 2002; Pagel et al. 2004; O’Meara 2012; Revell 2012) have improved greatly in recent decades. These developments have facilitated a paradigm shift toward investigating biogeographic history in the context of reconstructed ancestral ecological niche characteristics (e.g., Rice et al. 2003; Graham et al. 2004; Knouft et al. 2006; Pearman et al. 2008; Anciães and Peterson 2009; Evans et al. 2009; Vieites et al. 2009; Smith and Donoghue 2010; Nyári and Reddy 2013; Ribeiro et al. 2016; Owens et al. 2017).
Researchers have used several approaches to characterize ecological niches when attempting to reconstruct the evolutionary history of species’ niches. Some of the earliest studies used means and standard errors of suitable abiotic niche characteristics in their reconstructions (Rice et al. 2003; Anciães and Peterson 2009). Soon, however, researchers began characterizing niches using minimum and maximum suitable abiotic niche values (Graham et al. 2004; Yesson and Culham 2006), central tendencies of suitable niche values (i.e. mean or median; Ackerly et al. 2006; Kozak and Wiens 2010; Cooper et al. 2011), or distributions of suitable niche values (Evans et al. 2009; Smith and Donoghue 2010). These data were derived either directly from the occurrence data (e.g. Ackerly et al. 2006; Kozak and Wiens 2010; Cooper et al. 2011) or from ecological niche model outputs (e.g. Rice et al. 2003; Smith and Donoghue 2010; Nyári and Reddy 2013).
Fundamental ecological niches, however, are rarely characterized completely and unambiguously, owing to pervasive partial representation of fundamental ecological niches when assessed and characterized over real-world landscapes (Fig. 1; Saupe et al. 2012; Veloz et al. 2012; Owens et al. 2013; Guisan et al. 2014; Warren et al. 2014; Saupe et al. 2017). The fundamental ecological niche is defined as the set of conditions under which the species is able to maintain populations without immigrational input (Soberón 2007). The ability of a species to occupy a particular fundamental ecological niche is the result of phenotypic traits subject to natural selection (Peterson et al. 2011). However, the full suite of environmental conditions in a species’ fundamental niche is not necessarily represented across Earth at any given time. Furthermore, existing abiotic conditions are not necessarily accessible to a species or coincident with suitable biotic conditions (Barve et al. 2011). A species’ realized niche (i.e. where the species is found) is determined by available resources, biotic factors such as competition, availability of suitable environments, and/or dispersal barriers and dispersal capabilities (Soberón 2007). As such, any characterization of fundamental ecological niches that relies on inference from species’ geographic distributions (i.e. realized niche) will either be incomplete or will have to be inferred via extrapolation (Saupe et al. 2012; Owens et al. 2013). Hence, while the estimated niche of a lineage through time may show variation in response to inherited adaptations that alters the lineage’s fundamental niche, that variation may also derive from changes in the set of environments accessible to that lineage, which do not represent a genetically-inherited set of adaptations (Araújo et al. 2013).
Methodologies that use estimates based on species’ realized niches to characterize ecological niches in phylogenetic analyses are known to overestimate true amounts of niche change (Saupe et al. 2017). Here, we present a new framework to characterize species’ fundamental niches, which further emphasizes and illustrates the importance of considering gaps in knowledge as part of coding species’ fundamental niches as characters for comparative phylogenetic analyses. The framework relies on consideration of areas sampled by researchers and accessible to the species over relevant time periods (M ; Phillips et al. 2009; Barve et al. 2011; VanDerWal et al. 2009). Estimating and accounting for this region has already been recognized as important in model excersises that use background or pseudo-absence data for calibration of suitable habitat (Phillips et al. 2009; Barve et al. 2011). If sampling effort and regions accessible to a species are ignored when selecting the geographic extent for model calibration, fitted models may be closer to models of survey effort and/or erroneously omit suitable niche conditions. However, even niche estimates derived from presence-only data (i.e., without a modelling component) should consider M , as doing so provides one of the only ways of assessing whether estimates are likely to be truncated. Specifically, unknown tolerance limits may be highlighted when environments across M do not encompass conditions in excess of those where the species in question is observed. In these cases, it may be prudent to indicate potential suitability of environments outside of the environmental bounds of M , whilst noting uncertainty in these estimates. Thus, a broader suite of environmental conditions must be considered, often derived from the region occupied by a focal clade rather than the M of a single species. The conceptual advance for our new approach is in assessment of niche estimates in the context of available environments withinM , considering shared environmental space across all focal species to account for potential cases of niche truncation.
Specifically, our new character coding method decomposes species’ niches into discrete bins across the broader environmental background occupied by a clade, and scores each bin as within, outside, or uncertain for each species’ fundamental niche (Fig. 1). We illustrate the utility of summarizing species’ niches this way by performing a simulation that compares evolutionary rates estimated from characters coded using our method to those estimated in a more traditional analysis. We detail how to use our new coding scheme to infer ancestral ecological niches and demonstrate the utility of our approach with an empirical example—inferring patterns of ecological niche evolution in New World orioles (Icterus spp.).