CTFS-ForestGEO: a worldwide network monitoring forests in an era of global change

Plants

Supplementary measurements on plants include liana abundance and diversity, functional traits, high-precision diameter growth, flower and seed production, seedling performance, and DNA barcoding (Table 3). Liana censuses help to characterize the important role of lianas in forest dynamics. Measurements of functional traits – well-defined, measurable properties of organisms that are strongly associated with ecological performance – provide information on key attributes and ecological roles of the species included in the census. High-precision growth measurements provide fine-scale understanding of temporal and spatial variation in tree growth and forest productivity. Flower, seed and seedling censuses enable study of complete tree life cycles, which are critically important for forest regeneration and long-term species persistence. DNA barcoding provides a powerful means of species identification that allows elucidation of phylogenetic relationships and ecological roles (Dick & Kress, 2009; Kress et al., 2009, 2010).

Animals-

Arthropod and vertebrate initiatives (Table 3) yield understanding of the roles of these taxa in forest dynamics through their roles as herbivores, pollinators, seed dispersers, predators, ecosystem engineers, and vectors of microbial diversity. In a unique effort to monitor multitaxon assemblages in tropical rainforests (Basset et al., 2013; but see Leidner et al., 2010 for long-term monitoring of a single taxon), key arthropod groups are being monitored to better understand how interactions between arthropods and plants affect forest dynamics and to evaluate the impact of global change on the full range of forest trophic levels. Vertebrate monitoring is helping to elucidate how mammals differentially affect tree species and how modification of the fauna may impact the future forest (e.g., Wright et al., 2007; Harrison et al., 2013; see below).

Ecosystem and environmental

Supplementary measurements of ecosystem and environmental variables include major aboveground C stocks and fluxes (aboveground biomass, standing dead wood and coarse woody debris, ANPP_wood, litterfall, net ecosystem exchange); soil nutrients, C, and fine root biomass; bio-micrometeorology, and meteorology (Table 3). These measurements provide a basis for understanding environmental and biotic controls on C stocks and fluxes within forest ecosystems and how these may respond to global change. Soils measurements provide a basis for understanding the critical role of soils in determining species composition, forest structure, and primary productivity, as well as their globally significant role as an important C reservoir. Bio-micrometeorological measurements further elucidate the important role of forests in climate regulation through ongoing exchange of CO₂, H₂O, and energy between the ecosystem and the atmosphere. Meteorological data are critical for understanding how the biotic community and whole ecosystem processes respond to climate variables over half-hourly to multiannual time scales.

Combined applications

In combination, the core tree census and supplementary measurements enable unique analyses of the many interacting components of forest ecosystems, yielding a holistic picture of forest dynamics. For instance, core census data have been combined with data on lianas, vertebrates, seeds, seedlings, and reproductive functional traits to link decreasing populations of seed dispersers to changing patterns of plant reproduction, liana abundance, and tree growth and survival (Wright & Calderon, 2006; Wright et al., 2007; Ingwell et al., 2010; Harrison et al., 2013). Core census, functional trait, and DNA barcoding data have been combined to understand the roles of phylogeny and functional traits in shaping habitat associations and diversity in space and time (Pei et al., 2011; Swenson et al., 2011). The combination of core census data, plant functional traits, ecosystem measurements, soils data, and weather data lend themselves to parameterizing and evaluating ecosystem and earth system models. Thus, the broad suite of standardized measurements at CTFS-ForestGEO sites (Tables 1 and 3) provides opportunities to address a multitude of questions on forest dynamics and their responses to global change pressures.

Are forests changing?

Change is the natural condition of forests (e.g., Baker et al., 2005; Laurance et al., 2009), which makes it challenging to detect and attribute directional responses to global change pressures. A key finding from the network is that forests generally, and in particular tropical forests, are highly dynamic; for instance, in the first 18 years of monitoring at BCI, >40% of trees ≥1 cm DBH (or 34% ≥10 cm DBH) turned over, and 75% of all species changed in abundance by >10% (Leigh et al., 2004). Superimposed upon this dynamism, forests are responding to global change pressures. Data from the network reveal some generalities and long-term trends of change in forests worldwide.

Forest composition in terms of species and functional groups has changed at multiple sites across the network, in different directions at different sites (Condit et al., 1996; Chave et al., 2008; Feeley et al., 2011; Makana et al., 2011; Swenson et al., 2011). An analysis of data from twelve CTFS-ForestGEO sites reveals that environmental variability – as opposed to demographic stochasticity – is the most important factor driving tree population dynamics on decadal time scales (Chisholm et al., 2014). Across relatively undisturbed tropical forests, the dominance of slow-growing species increased at nine of ten sites analyzed (significantly so at five sites), indicating that these forests may be recovering from past disturbances, even as they are impacted by a variety of global change pressures (Chave et al., 2008). In addition, at six tropical sites monitored over more than 10 years, there have been long term increases in the proportions of flowers and seeds produced by lianas (Fig. 6; Wright & Calderon, 2006; Wright, unpublished analysis) – a trend that corresponds with long term increases in the abundance of lianas observed on BCI (Panama) and elsewhere in the tropics (Ingwell et al., 2010; Schnitzer & Bongers, 2011). While community change appears to be the rule rather than the exception across the network, and while there have been some instances of rapid change in forest composition (e.g., Condit et al., 1995; Chave et al., 2008), there have not been any hugely dramatic changes such as a forest die-off affecting the majority of large trees at the network sites.

Trends in various components of aboveground net primary productivity (ANPP) have also been monitored at some sites. Across the network, the woody component of NPP (ANPP_wood) has increased or decreased, as a function of both climate change and succession. Forests globally are mixed in terms of their productivity trends (Laurance et al., 2004; Clark et al., 2010; Gedalof & Berg, 2010; Wright, 2010). For instance, decreases in ANPP_wood were observed in tropical forests in Panama (BCI) during 1981–2005 and Malaysia (Pasoh) during 1990–2000 (Feeley et al., 2007b) and increases in ANPP_wood were observed in secondary forests in Maryland, USA (SERC; McMahon et al., 2010). Notably lacking is evidence of consistent increases in ANPP, as might be expected based solely on increasing atmospheric CO₂ concentration (e.g., Norby et al., 2005). In the tropics, allocation of NPP to reproduction appears to have shifted; at five of six tropical sites where flower and seed production has been monitored for more than 10 years, there has been a long-term increase in flower production but not seed production (Wright & Calderon, 2006; Wright, unpublished analysis). Ongoing monitoring of NPP and flower and seed production will be vital to characterizing trends in productivity and C allocation.

Finally, changes in standing biomass over time have been detected. Across ten relatively undisturbed tropical forests, highly resolved estimates of net biomass change show that aboveground biomass increased on average 0.24 ± 0.16 Mg C ha⁻¹ yr⁻¹ (Fig. 7; Chave et al., 2008). This value is comparable to (though slightly lower than) values recorded for networks of small forest plots in Amazonia (0.62 ± 0.23 Mg C ha⁻¹ yr⁻¹; Baker et al., 2004), and Africa (0.63 ± 0.36 Mg C ha⁻¹ yr⁻¹; Lewis et al., 2009a). Combining published data for the CTFS-ForestGEO, RAINFOR, and AfriTRON tropical forest sites leads to an overall average of 0.34 ± 0.11 Mg C ha⁻¹ yr⁻¹ based on a total of 8243 ha-years of monitoring (Muller-Landau et al., 2014). Ongoing monitoring will be important for quantifying trends in biomass in the global forests represented by CTFS-ForestGEO.

What are the mechanisms by which global change impacts forests?

While data from the CTFS-ForestGEO network add to abundant evidence that forests globally are changing (e.g., Soja et al., 2007; Lewis et al., 2009b; Allen et al., 2010; Wright, 2010), it is difficult to identify the mechanisms underlying such changes given ubiquitous and simultaneous changes in multiple global change drivers (Fig. 3). The information-rich nature of CTFS-ForestGEO sites has yielded insights into the mechanisms of response to global change pressures.

Warming is expected to alter forest dynamics, but predicting effects at the ecosystem scale remains a major scientific challenge (e.g., U.S. DOE, 2012). Monitoring, physiological measurements, and nearby warming experiments combine to yield insights into how warming may impact forest dynamics. The effects of warming are perhaps most dramatic at Scotty Creek, Canada – the highest latitude site, which is experiencing rapid warming (Figs 2 and 3; Tables S3 and S4) – where accelerating permafrost thaw is resulting in tree functional decline near forested plateau edges (i.e., reduced sap flow, radial growth, and leaf area) and driving loss of forest area at a rate of 0.5% yr⁻¹ (Baltzer et al., 2014). At another Canadian site (Haliburton Forest), a heat wave event during spring leaf-out in 2010 resulted in a >50% decline in leaf area of the dominant tree species (Filewod & Thomas, 2014), and a large net ecosystem carbon loss in the same year (Geddes et al., 2014). However, at most temperate and tropical sites, the impacts of warming are less obvious and tend to be confounded by other aspects of global change (Fig. 3). Data from four tropical forest sites (BCI, Huai Kha Khaeng, Lambir, and Pasoh) indicate that tree growth rate correlates negatively with nighttime temperature, as expected from increased respiration rates causing a reduced carbon balance (Feeley et al., 2007a; Dong et al., 2012) – a trend that has also been observed at an external site in Costa Rica (Clark et al., 2010). In contrast, warming experiments associated with two of the sites reveal that warming may also directly or indirectly increase woody productivity; specifically, soil warming at Harvard Forest has increased tree growth through increased N mineralization (Melillo et al., 2011), and chamber warming experiments in Panama revealed that increased nighttime temperatures increased seedling growth rates (Cheesman & Winter, 2013). Ongoing monitoring, experimentation, and modeling will be necessary to disentangle the diverse productivity responses of forests to warming. Warming may also shift C allocation to reproduction; flower production at BCI, Panama has increased with increasing temperature (Pau et al., 2013). Future warming (Fig. 2) will inevitably impact forests, and ongoing monitoring at CTFS-ForestGEO sites should help to document and explain these changes.

Changes in aridity and drought severity have the potential to impact forests worldwide, including those in wet climates (Allen et al., 2010; Choat et al., 2012). Across the tropics, increases in aridity or the occurrence of severe droughts have led to forest “browning”, mortality episodes, or fires (Van Nieuwstadt & Sheil, 2005; Lewis et al., 2011; Zhou et al., 2014), and there is concern that potential future increases in aridity in some parts of the tropics could result in severe tropical forest dieback (e.g., U.S. DOE, 2012). Research across the CTFS-ForestGEO network has yielded insights into the role of aridity in shaping tropical forest dynamics. Droughts in Panama (BCI, San Lorenzo, and Cocoli) and Malaysia (Lambir) have revealed differential drought sensitivity by size class, microhabitat association, and functional type (Condit et al., 1995, 2004; Potts, 2003). In Panama, mild or even fairly strong drought increased both woody productivity and production of flowers and seeds – presumably because of increased solar radiation (Condit et al., 2004; Wright & Calderon, 2006). At a tropical dry forest in India (Mudumalai), drought increased mortality rate, but with a 2–3 year lag for larger trees (Suresh et al., 2010). These findings yield insight into how moist tropical forests may respond to predicted changes in aridity (Fig. 2; Table S4; IPCC, 2013).

Beyond climate, impacts of other global change drivers have been observed across the CTFS-ForestGEO network. Nitrogen deposition (Fig. 1; Table S5) has altered forest biogeochemistry across the globe. Temperate forests are typically N limited; however, high N deposition at Haliburton Forest, Canada, has caused a shift from N to P limitation (Gradowski & Thomas, 2006, 2008), providing evidence of constraints on increases in temperate forest productivity driven by elevated CO₂ and/or nitrogen deposition. Because tropical forests are typically limited by elements other than N, N deposition is not expected to increase the productivity of these forests (Matson et al., 1999). At the two tropical CTFS-ForestGEO sites where relevant measurements have been made, increased ¹⁵N concentrations in plant tissues suggests substantial N deposition and altered N cycles (Hietz et al., 2011). Specifically, on BCI, leaf N and δ¹⁵N in recent (2007) samples were elevated relative to herbarium samples (~1968) (Hietz et al., 2011). These changes have been mechanistically linked to increased N availability through a nearby fertilization experiment, which increased foliar N concentrations and δ¹⁵N by similar amounts but did not affect productivity (Wright et al., 2011; Mayor et al., 2014a,b). A similar increase in δ¹⁵N was observed in wood from Huai Kha Khaeng, Thailand (Hietz et al., 2011). These results imply that, in tropical forests, N deposition is accelerating N cycling without increasing productivity, and reduced cation availability resulting from N deposition may be one potential explanation for observed declines in tree growth rates at some tropical sites (see above; Matson et al., 1999).

Habitat fragmentation (Fig. 5) and faunal degradation have also been linked to altered dynamics at CTFS-ForestGEO sites. The CTFS-ForestGEO site near Manaus, Brazil, is part of the Biological Dynamics of Forest Fragments Project (BDFFP), which has revealed that forest fragmentation rapidly and profoundly alters tree, arthropod, bird, and primate communities, reducing species diversity and shifting composition toward dominance of more disturbance-adapted species (Laurance et al., 2006). Across the network, more highly fragmented sites (e.g., Witham Woods, UK; Bukit Timah, Singapore; Lambir, Malaysia; Heishiding, China; Fig. 3; Table S5) tend to have degraded faunas, as indicated by the absence of apex predators and larger vertebrates that were present historically, whereas faunal communities tend to remain more intact in unfragmented forests such as Yasuni (Ecuador), Rabi (Gabon), and Scotty Creek (Canada) (Turner & T Corlett, 1996; LaFrankie et al., 2005; Laurance et al., 2012; Harrison et al., 2013;W.F. Laurance, personal communication). As detailed below, faunal degradation – whether caused by habitat fragmentation, hunting, or other pressures – has strong impacts on forest structure and dynamics.

The strong influence of fauna on forest composition and dynamics (e.g., Wright, 2010; Estes et al., 2011; Schmitz et al., 2013) has been documented at several CTFS-ForestGEO sites. At Mpala, Kenya, an experiment excluding herbivores of different sizes and replicated across a rainfall gradient revealed that herbivores of different sizes influence the biomass and growth rates of trees and understory plants, plant community composition, and small mammal communities (Goheen et al., 2013). At Mudumalai, elephants (Elephas maximus) cause high mortality among the small- to medium-sized stems, particularly in a few favored forage species (Sukumar et al., 2005). At SCBI (Virginia, USA), where white-tailed deer (Odocoileus virginianus) populations greatly exceed their historical levels, 20 years of deer exclusion from a 4-ha subsection of the CTFS-ForestGEO plot has resulted in a >4-fold increase in sapling abundance relative to heavily browsed forest outside the exclosure (McGarvey et al., 2013). Large impacts of mammalian herbivores have also been found in an exclosure study adjacent to the Pasoh plot site in Malaysia (Ickes et al., 2001), where native pigs (Sus scrofa) have a dramatic effect on tree recruitment. In Panama, comparison of forest plots protected from bushmeat hunting with those exposed to poachers revealed that by reducing the abundance of frugivores and seed dispersers, hunting decreases the abundance of plant species with seeds dispersed by these animals while increasing the abundance of species with seeds dispersed by bats, small birds, or mechanical means (Wright et al., 2007). The latter includes lianas whose seeds are much more likely to be dispersed by wind (60% of liana species vs. 25% of canopy trees and <10% of midstory and understory trees and shrubs). Lianas have thus increased disproportionately in abundance where hunters remove the frugivores that disperse the seeds of most tree species, hence hunting may have unforeseen consequences for carbon sequestration (Jansen et al., 2010). Directional change in tree communities driven by faunal degradation has also been demonstrated. At Lambir, where populations of large mammals and birds have been severely impacted by hunting, tree community dynamics changed profoundly from 1992 to 2008 (Harrison et al., 2013). Specifically, sapling densities increased and regeneration of tree species with animal-dispersed seeds decreased and became more spatially clustered (Harrison et al., 2013). Thus, ongoing faunal degradation due to hunting and habitat fragmentation in many forests globally is expected to alter forest community composition, tree dispersal and regeneration, species diversity, forest structure, and carbon cycling.

CTFS-ForestGEO research has also shed light on community interactions that will act to either magnify or buffer forest responses to global change. Species are linked to one another through complex webs of interaction. For example, mapping of quantitative trophic foodwebs at Wanang (Papua New Guinea) and current efforts to document tritrophic foodwebs of seeds, seed predators and parasitoids at this same location, at Khao Chong (Thailand) and Barro Colorado Island (Panama) demonstrates the complexity of ecological interactions in forest ecosystems (Novotny et al., 2010). Studies of seed dispersal and seedling recruitment demonstrate the critical role of vertebrates and insects in tree reproduction and the composition of future forests (e.g., Wright et al., 2007; Harrison et al., 2013). It is therefore unsurprising that global change impacts on one group cascade through the ecosystem. For example, as described above, dramatic reduction in large mammal and bird populations at Lambir, Malaysia has altered the dynamics of tree dispersal and regeneration (Harrison et al., 2013). Similarly, in the light-limited moist tropical forests of Panama, El Niño events bring relatively cloud free, sunny conditions that enhance fruit production while subsequent La Niña events bring rainy, cloudy conditions, and lower levels of fruit production that can lead to famines, particularly among terrestrial frugivores and granivores (Wright et al., 1999; Wright & Calderon, 2006). Climate change is bringing changes in cloud cover and atmospheric transmissivity to PAR (Table S3) with cascading effects on frugivores, granivores, and the plant species with which they interact.

At the same time, the diversity and complexity of forest communities may serve to provide some resilience to global change. A diversity of tree physiological strategies implies a wide range of responses to global change that helps to provide ecosystem resilience (e.g., Isbell et al., 2011; Mori et al., 2013). For example, Panamanian tree species have displayed a wide range of physiological responses to temperature variation (Cheesman & Winter, 2013; Slot et al., 2014), and trees of different species have generally responded differently to experimental manipulation of CO₂, temperature, or precipitation globally (Anderson-Teixeira et al., 2013). The resilience enabled by species diversity may be exemplified by the stability of biomass, size structure, and functional composition of the BCI forest (Chave et al., 2008; Swenson et al., 2012) despite severe droughts that impacted drought-sensitive species (Condit et al., 1995, 1996). In addition, in the tropics, pervasive negative density dependence – i.e., elevated mortality of a plant species in areas where it is abundant – may buffer change because as a species becomes rare, it will suffer less from negative density dependence (Comita et al., 2010). Thus, accounting for biodiversity in ecosystem models will be important for predicting forest responses to climate change. While such complexity makes it challenging to predict forest responses to global change, it may serve to partially buffer forest response to global change, which might otherwise be more dramatic.