Cross-ecosystem subsidies have been a major focus of food web ecology for decades (Marczak et al., 2007; Allen and Wesner 2016) and much of this work has been linked to the meta-ecosystem concept (Loreau et al., 2003b). Terrestrial-aquatic connections provide resource subsidies that support higher growth rates (Sabo and Power 2002), abundances (Sanpera-Calbet et al., 2009; Allen and Wesner 2016), and niche diversity of consumers (Darimont et al., 2009) in adjacent ecosystems. Subsidies of terrestrial invertebrates to streams provide up to half of the annual energy budget to fishes such as salmonids, and emergence of adult aquatic insects comprises between 25 and 100% of the energy budget to nearby terrestrial bird, bat, lizard, and spider populations (Baxter et al., 2004; Baxter et al., 2005; Gratton and Vander Zanden 2009). Although aquatic ecosystems generally receive relatively large quantities of low-quality organic matter subsidies from terrestrial ecosystems (e.g., leaves), and terrestrial ecosystems receive relatively small quantities of high-quality subsidies from aquatic ecosystems (e.g., guano and insects), consumers rely on these subsidies to a comparable extent in both realms (up to 40% of the diet in each) (Soininen et al., 2015), but see Hagar et al. (2012).

Terrestrial-aquatic subsidy impacts can have large spatial extents. Signatures of subsidies to terrestrial environments have been shown to extend up to 5,300 m laterally from stream banks and nearly 1,000 m from lakeshores (Muehlbauer et al., 2014). Bats, for example, consume large quantities of emergent aquatic insects and deposit guano several kilometers away, increasing nutrient concentrations near roosts (Power et al., 2004). Subsidies from one ecosystem may promote biodiversity in an adjacent ecosystem both directly by increasing the spatial and temporal availability of prey to consumers (Schindler et al., 2013), and indirectly by alleviating predation pressure on prey in adjacent ecosystems (Sabo and Power 2002).

Positive relationships have been shown in studies that have quantified cross-realm BEF relationships by linking terrestrial primary production and inland aquatic biodiversity. Controlled mesocosm experiments have revealed complex relationships between terrestrial leaf and insect subsidies and the biodiversity of aquatic systems (Klemmer et al., 2020). At landscape and regional scales, richness of stream insects (Vinson and Hawkins 2003) and plankton diversity (Soininen and Luoto 2012) was positively correlated with watershed Normalized Difference Vegetation Index (NDVI), a measure of terrestrial greenness and productivity. Catchment terrestrial primary production explained significant variation in geographic distributions of anadromous arctic char (Salvelinus alpinus), suggesting that aquatic animals track the spatial availability of terrestrial primary production (Finstad and Hein 2012). Such studies have prompted some aquatic ecologists to advocate using measures of terrestrial primary production to predict broad-scale patterns of aquatic biodiversity (Soininen et al., 2015).

Despite the importance of cross-realm subsidies, studies of subsidy-biodiversity relationships tend to be unidirectional. Little research has documented the reciprocal subsidies (Figures 2I,J) that reflect the true web structure of these relationships (but see Klemmer et al., 2020). BEF research predicts that strong richness-productivity relationships in one system should strengthen the association between richness and productivity in adjacent systems by reducing spatial and temporal variance around the richness-productivity correlation (Loreau et al., 2003a). For example, increased mussel species richness in streams reduces the temporal variance in insect emergence (Allen et al., 2012), and increased soil microbial biodiversity may stabilize hydrological pathways of material transfer from terrestrial to aquatic systems (Bardgett et al., 2001). Although these studies show that terrestrial or aquatic BEF may spill over to adjacent systems, we are not aware of a single study that demonstrates a direct mechanistic connection between terrestrial BEF and aquatic BEF. Therefore, the current understanding of cross-ecosystem subsidies is likely underestimating the strength and importance of cross-realm BEF relationships, which could impact environmental conservation prioritization decisions within and across realms.

Physical alterations of landscapes by organisms can significantly influence the connections between terrestrial and inland aquatic systems. The effects of these “ecosystem engineers” (Jones et al., 1994) often span ecological realms (Hastings et al., 2007). Beavers (Castor spp.), for example, influence the riparian density, biomass, and species composition (Johnston and Naiman 1990; Wright et al., 2002) as well as the hydrologic profile of freshwater systems. Beaver dams, in turn, increase the diversity of aquatic invertebrates important for terrestrial consumers (McCaffery and Eby 2016) and increase the supply of terrestrially derived organic matter in streams (Anderson and Rosemond 2010). Beaver dams elevate soil moisture and nitrogen levels (Naiman et al., 1994) that, along with beaver herbivory, affect forest succession and increase terrestrial and wetland plant species diversity. Hippopotami (Hippopotamus amphibius) create new river channels which redistribute nutrients, enhance productivity, and facilitate the movement of fish populations (Mosepele et al., 2009); they also transport terrestrial carbon and nutrients to aquatic systems (Subalusky et al., 2015).

Physical modification by sediment-binding plants also influences both aquatic and terrestrial systems. The abiotic environment influences dune grass growth, which in turn affects dune shapes (Zarnetske et al., 2012; Emery and Rudgers 2014); the resulting dune geomorphology and dune plant community influence the dune hydrological regime including ponds and wetlands in dune slack areas (Grootjans et al., 1998). A wide range of ecosystem engineers affect both terrestrial and inland aquatic realms, illustrating the potential for strong BEF connections through direct physical alteration of the landscape.

Strong hydrologic connections among terrestrial and inland aquatic realms exist in both river floodplains and dryland ecosystems. In the evolution of river floodplains terrestrial vegetation diversifies inland aquatic habitats, providing a range of colonization options for different organisms (Ward et al., 2002). This example suggests that terrestrial primary production and plant species diversity may promote inland aquatic biodiversity by increasing habitat heterogeneity, improving forage quality and variety, and supporting more diverse assemblages of organisms. Research in dryland ecosystems shows that low-productivity systems with high levels of salinity in both water and soil have negative effects on biodiversity. Increased levels of salinity in eastern Australia, for example, caused riparian trees to die, which caused aquatic salinity levels to rise, in return causing further tree death (Briggs and Taws 2003). Increases in salinity can also decrease the diversity of aquatic invertebrates, riparian and aquatic vegetation, microalgae, and fish (Hart et al., 2003). As aquatic invertebrates and fish are important predators of mosquitoes, a reduction in predation along with more mosquito breeding habitat from salinity-induced rising water tables can increase mosquito-borne virus transmission to humans (Jardine et al., 2007). Given the dominant role that hydrology plays in both terrestrial and aquatic realms, there are likely many other examples of hydrological connections influencing BEF relationships, though studies do not often use the BEF framework (e.g., Kneitel and Lessin 2010; Vander Zanden and Gratton 2011; Stahlschmidt et al., 2012; Schriever et al., 2014).

The relative importance of processes that impact BEF relationships across realms is likely to differ across scales, as has been demonstrated in single-realm studies (e.g., Thompson et al., 2018). We present hypotheses about how terrestrial and inland aquatic linkages may change across spatial scales of analysis (Figure 3). Similar hypotheses could be developed to consider connections across temporal scales; however, temporal scaling is beyond the scope of this paper. At fine grains and local extents, we expect strong connections between terrestrial and inland aquatic systems—the condition of an individual lake matters to the watershed that drains into it, and a watershed has direct impacts on the water bodies it feeds (Jackrel and Wootton 2014) (Figure 3A). Gratton and Vander Zanden (2009), for example, estimated that fluxes of aquatic insects to the surrounding terrestrial ecosystems would move a maximum of 300 m inland from a lake shore. At intermediate grains and regional extents greater variability in environmental conditions should result in weaker or noisier terrestrial-aquatic connections. For example, Sabo and Hagen (2012) found that within river networks, the strength of connection between a river and its surrounding watershed was highly dependent on the morphometry of the river network. Some watersheds may be highly connected, whereas others are less connected, or connections are strong only at small extents (Sabo and Hagen 2012; McCullough et al., 2019) (Figure 3B).

At very coarse grains and broad extents, we expect to find the weakest relationships between terrestrial and inland aquatic realms (Figure 3C). Broad-scale heterogeneity in climate, landform, and biogeography, along with regional-scale processes such as land use/cover and glaciation, should add noise to or diminish the connections between aquatic and terrestrial biodiversity and productivity (Holland et al., 2011; Stachelek et al., 2020). Terrestrial and inland aquatic systems are likely to respond to these macroscale phenomena in different ways and at different rates.

BEF studies often focus on a single trophic level, often primary producers. Since as little as 10% of the energy produced at a given trophic level is transferred to the next higher trophic level (Elser et al., 2000), we might expect BEF relationships across realms to weaken when higher trophic levels are considered. However, the question of whether food web controls are bottom-up (resource controlled) or top-down (consumption controlled) continues to be fundamental in ecological research across realms (Lynam et al., 2017; Vidal and Murphy 2018).

Taking a green-world, bottom up perspective (Polis 1999), we expect strong relationships between inland aquatic and terrestrial productivity. Because all productivity is generally controlled by the same suite of processes, regardless of system (Grimm et al., 2003), productivity between realms should be related (Figure 4A). Terrestrial productivity should also have a strong correlation with aquatic producer (autotroph) diversity (Figure 4D), and vice versa (Figure 4B)—more productive landscapes should have higher autotroph biodiversity regardless of realm. Similarly, aquatic producer diversity and terrestrial producer diversity (phytoplankton and plants) should follow similar biogeographic patterns and therefore be correlated (Edvardsen and Økland 2006; Suurkuukka et al., 2014; Stoler and Relyea 2016) (Figure 4E). We expect aquatic consumer (heterotroph) biodiversity to have a weaker but still positive correlation with terrestrial productivity (Dolný et al., 2013; Schriever et al., 2014; Kautza and Sullivan 2015) (Figure 4G), and aquatic productivity should be weakly correlated with terrestrial consumer biodiversity (Stahlschmidt et al., 2012; Sarneel et al., 2014) (Figure 4C).

Moving from lower to higher trophic levels, from a bottom up perspective we expect organisms to be able to access resources from more food web compartments (e.g., Marczak et al., 2007) and so their biodiversity should be less closely tied to any individual source of productivity (Figures 4F,H). Finally, we expect connections between aquatic and terrestrial consumer biodiversity to exist (Purdy et al., 2012); however, due to food web complexities, from a bottom-up perspective we would expect these relationships to be weak (e.g., Corti and Datry 2016) or undetectable in some systems (Hagar et al., 2012; Tonkin et al., 2016) (Figure 4I). In the case of streams and rivers, these connections may further be complicated by the fact that water and nutrients are moving rapidly through the ecosystem; this could lead to spatial disconnects between terrestrial and aquatic productivity (Vannote et al., 1980).

While bottom up control of consumer diversity is important, it is also likely that top down control, the brown-world perspective (Bond 2005), is a major influence in at least some contexts and at some spatial scales (Borer et al., 2005). Terrestrial predators (high trophic level consumers) may regulate the diversity and abundance of aquatic prey, which in turn affects the diversity and abundance of aquatic primary producers (Nakano et al., 1999). In systems where top-down forces are important relative to bottom-up forces, an alternative set of hypotheses may be more realistic (Figure 4, right panel). From a top-down perspective, we would expect a stronger positive relationship between aquatic and terrestrial diversity at higher trophic levels, and there are examples of trophic cascades within higher trophic levels connecting terrestrial and aquatic consumers (e.g., Wesner 2012) (Figure 4R). Cross-realm trophic cascades may then impact relationships between consumer and producer diversity (Figures 4O,Q). Depending on the strength of these higher trophic level connections, this situation could lead to a decoupling of terrestrial and inland aquatic productivity (Gergs et al., 2014) (Figure 4J). In many systems, both top-down and bottom-up processes may operate simultaneously, and experimental approaches may be required to disentangle these processes. These examples demonstrate the importance of carefully considering spatial scales, complex life history strategies, and trophic levels when studying BEF connections across realms.