Climate change, woodpeckers, and forests: Current trends and future modeling needs

Abstract The structure and composition of forest ecosystems are expected to shift with climate‐induced changes in precipitation, temperature, fire, carbon mitigation strategies, and biological disturbance. These factors are likely to have biodiversity implications. However, climate‐driven forest ecosystem models used to predict changes to forest structure and composition are not coupled to models used to predict changes to biodiversity. We proposed integrating woodpecker response (biodiversity indicator) with forest ecosystem models. Woodpeckers are a good indicator species of forest ecosystem dynamics, because they are ecologically constrained by landscape‐scale forest components, such as composition, structure, disturbance regimes, and management activities. In addition, they are correlated with forest avifauna community diversity. In this study, we explore integrating woodpecker and forest ecosystem climate models. We review climate–woodpecker models and compare the predicted responses to observed climate‐induced changes. We identify inconsistencies between observed and predicted responses, explore the modeling causes, and identify the models pertinent to integration that address the inconsistencies. We found that predictions in the short term are not in agreement with observed trends for 7 of 15 evaluated species. Because niche constraints associated with woodpeckers are a result of complex interactions between climate, vegetation, and disturbance, we hypothesize that the lack of adequate representation of these processes in the current broad‐scale climate–woodpecker models results in model–data mismatch. As a first step toward improvement, we suggest a conceptual model of climate–woodpecker–forest modeling for integration. The integration model provides climate‐driven forest ecosystem modeling with a measure of biodiversity while retaining the feedback between climate and vegetation in woodpecker climate change modeling.


| INTRODUC TI ON
As global atmospheric CO 2 has increased, the United States has warmed 0.7°C-1.1°C, with most of the warming occurring since 1970 (Walsh et al., 2014) impacting forest ecosystems (Anderson-Teixeira et al., 2013). Globally, forests provide many ecosystem services, including sequestration of ~30% of global annual anthropogenic CO 2 emissions (Pan et al., 2011) and habitat for 77% of the global avifauna (BirdLife International, 2017). Climate warming and changing precipitation regimes have impacted forest ecosystem structure and function (Anderson-Teixeira et al., 2013), including North American avifauna populations (Prince & Zuckerberg, 2015;Tingley, Koo, Moritz, Rush, & Beissinger, 2012). Moreover, predictions indicate that more than half of the forested land cover of North America will experience future climates that differ from historical growing conditions (Charney et al., 2016) with obvious implications for preservation of wildlife biodiversity (Langdon & Lawler, 2015), since forest composition and structure are integral to biodiversity (McElhinny, Gibbons, Brack, & Bauhus, 2005).
Specifically, climate change is expected to cause declines in tree species occurrence (Coops & Waring, 2011a), shifts in carbon stocks (Lenihan et al., 2008), increases in forest mortality events (Allen et al., 2010;McDowell & Allen, 2015), and increases in burned area (Rogers et al., 2011J). These changes will affect avifauna habitat. For example, moderate-to high-severity fires can create open forests, adequate snag density, and minimal midstory vegetation necessary for some woodpecker habitat (Hoyt & Hannon, 2002;Vierling, Lentile, & Nielsen-Pincus, 2008;Zhu, Srivastava, Smith, & Martin, 2012). But even with increases in area burned or fire intensity, models also predict tree species composition shifts that pose adaptation constraints on woodpeckers (Fogg, Roberts, & Burnett, 2014) and potentially reducing habitat and biodiversity.
We propose the woodpecker guild as an ensemble of wildlife species to function as indicators of forest resiliency and biodiversity in a coupled modeled response of vegetation and wildlife to climate change. Woodpeckers are ideally suited as indicator species of forest ecosystem dynamics (Koch, Drever, & Martin, 2011;Segura, Castaño-Santamaría, Laiolo, & Obeso, 2014), because they are ecologically constrained by landscape-scale forest components, such as composition, structure, disturbance regimes, and management activities, in addition to being correlated with forest avifauna community diversity (Archaux & Bakkaus, 2007;Diaz, Armesto, Reid, Sieving, & Willson, 2005;Drever, Aitken, Norris, & Martin, 2008;Patton, 1992). Woodpeckers are also strongly associated with old-growth/structurally complex forests (Drever et al., 2008;Hannon & Drapeau, 2005;Segura et al., 2014), which sustain greater biodiversity (Mazziotta et al., 2016) and are key habitat characteristics that modulate woodpecker population responses.
Because these forest components will be impacted by climate change (Allen et al., 2010;Anderson-Teixeira et al., 2013;Parks et al., 2016;Rocca et al., 2014;Weed et al., 2013), the change will have cascading effects on woodpecker responses, rendering them viable indicators in modeling future changes to biodiversity.
We reviewed the current and predicted trends associated with climate change impacts on woodpecker responses to identify ways to integrate woodpecker and forest ecosystem models. In addition, our intent is to provide a collective baseline of woodpecker responses to current and future climate change for integrated modeling efforts to be evaluated against. To identify ways to integrate woodpecker models, we identify inconsistencies between current (observed) and predicted responses, explore the modeling causes, and identify the models pertinent to integration that will address inconsistencies. We acknowledge there are vast syntheses possible when studying the response of woodpeckers to climate change. However, the focus of this review is to seek the information to facilitate identification of the model attributes that can best serve an integrated framework of climate-woodpecker-forest modeling. Having this framework will facilitate including other biodiversity measures (e.g., other species) in future climate modeling efforts.

| ME THODS AND RE VIE WED LITER ATURE
We conducted a systematic literature review of the observed and predicted responses to climate change of 22 North American woodpecker species. We refer to woodpecker response models as any of the following: species distribution, occupancy, abundance, and demographic models. Search terms using Google Scholar and Web of Science included "avian cavity nesters climate change," "woodpeckers climate change," "birds climate change," and "birds breeding climate change." The search spanned all literature through June 2018.
We included all papers that modeled the effects of climate change on woodpecker responses. Models that based predictions on alternative analyses to evaluated datasets (Distler, Schuetz, Velásquez-Tibatá, & Langham, 2015;Rodenhouse et al., 2008;Schuetz et al., 2015) or reported woodpecker responses aggregated at the community level (Stralberg et al., 2009)  These were mostly bioclimatic niche models ( and territory type; Supporting information Table S4).

| PRED IC TED WOODPECK ER RE S P ON S E S TO CLIMATE CHANG E
Generally, geographic forecasts indicate a north-northeast shift of eastern U.S. avifauna species by 2100 ) and a concurrent change in community composition (Langham, Schuetz, Distler, Soykan, & Wilsey, 2015;Stralberg et al., 2009 breeding bird assemblages of northern Canada and Alaska may gain as many as 80 species, while the greatest species loss is predicted along the Canadian-U.S. border and through the Rocky Mountains . Model results show that the resulting dissimilarity to contemporary species composition will be greatest throughout Canada and the Rockies. These trends will downscale to regional extents; for example, upwards of 57% of California may have novel breeding bird species assemblages by 2070 with no current analogs (Stralberg et al., 2009). In addition, central and southern California are areas of peak losses of species in the nonbreeding season .
Among the models reviewed, the model of Langham et al. (2015) is evaluated are predicted to be climate endangered or threatened due to loss of breeding and/or wintering range by the end of the century (Supporting information Table S1). Some of the range losses will be mitigated by climatically suitable range expansions. This results in an overall 53% and 23% of the woodpecker species breeding and nonbreeding ranges to exhibit net contractions by 2080, respectively ( Figures 1 and 2). Overall, all woodpecker species will lose climatically suitable habitat by the end of the century, and even with net gains, a majority are labeled as climate threatened or endangered based on climatic range changes (Supporting information Table S1).
In comparison, a trait-based assessment of climate change vulnerability via assessment of sensitivity, exposure, and adaptability found a mixed response among woodpeckers to those metrics. Most North American woodpecker species are sensitive to climate change. However, all are ranked as low vulnerability because of exposure ("the extent of the species' environment that will change") and/or high adaptive capacity ("the species' ability to avoid the negative impacts of climate change through dispersal and/or micro-evolutionary change"; Supporting information Table   S1; Foden et al., 2013). This discrepancy between the bioclimatic niche predictions  and climate vulnerability of climate vulnerability (i.e., sensitivity and adaptability). Hence, a species may be exposed to shifts in climatically suitable habitat but may have adaptability potential via phenotypic plasticity or not be sensitive to the degree of climate change represented in the bioclimatic niche model.
Spatially, there is an emergent pattern of predictions among woodpeckers relative to their contemporary distributions. The climatically suitable ranges of species with contemporary northern or western distribution centroids (i.e., those associated with conifer/ boreal forests) are projected to contract .
This is in concordance with other model results of climate-induced declines in avifauna abundance and species richness in conifer/boreal habitats of North America (Stralberg et al., 2015) and Europe (Virkkala, Heikkinen, Leikola, & Luoto, 2008 However, species at the southern edge of their range within this region (e.g., American Three-toed Woodpecker (Picoides dorsalis) and Black-backed Woodpecker (Picoides arcticus)) may diminish because of the encroachment of hardwoods from lower elevations into their primary habitat (spruce-fir; Rodenhouse et al., 2008). Nevertheless, coastal and southern regions of the United States are predicted to provide climates amenable to many wintering species .
Laying date advancement and increase in reproductive productivity of Northern Flickers (Colaptes auratus) were observed along the U.S. Pacific coast (Wiebe & Gerstmar, 2010). The authors showed that the response is spatially explicit; it correlates with increases in local ambient temperatures instead of broad regional climate indices or range-wide temperature gradients. Moreover, differing climatic conditions is producing similar phenology responses within the same species. Red-cockaded Woodpeckers (Leuconotopicus borealis) are laying earlier, and those that do are more productive (Schiegg, Pasinelli, Walters, & Daniels, 2002). The climate factors that correlate to these responses differ between populations; one population is responding to increases in temperature and the other increases in precipitation (Schiegg et al., 2002). Mechanistically, this may be occurring via genetic diversity and age-based experience, which increases plasticity (Schiegg et al., 2002 The integrated framework of climate-woodpeckerforest modeling (d) resulting from the linking of separate model types (a-c). (a) Climate-forest prediction models include a spectrum of model types: dynamic global vegetation models (DGVMs) to GAP models to dynamic community process-based forest landscape models (i.e., dynamic communities, spatial interactions, and ecosystem processes); (b) Climate-woodpecker prediction models include bioclimatic envelope models; (c) Woodpecker-forest models include realized niche models (e.g., occupancy), potential niche models (e.g., habitat suitability), and demographic models et al., 2008, 2013), their preferred nesting tree (Martin, 2015). This is rendering some species more vulnerable because of sensitivity to changes in nesting tree availability and lack of observed adaptability.
Martin (2015)  TA B L E 3 The predicted 2020 breeding range size relative to the 2000 range  and observed contemporary breeding range changes (Bateman et al., 2016) that is, the probability of occurrence was positively associated with climatic trends and was independent of abundance trends (Supporting information   Table S3; Stephens et al., 2016).
In addition, the northward winter range shift is occurring without a concurrent population abundance change (Supporting information Figure S1; Soykan et al., 2016 and Yellow-bellied Sapsucker is strongly contributing to these winter community composition changes (Prince & Zuckerberg, 2015).
However, only the Pileated and Red-bellied Woodpecker populations, both resident migrants, exhibited a concurrent increase in abundance during the winter season (Supporting information Figure   S1; Soykan et al., 2016 (Tingley et al., 2012(Tingley et al., , 2009 shifting species upslope and downslope, respectively (Tingley et al., 2012). Comparing 1911-1929, Tingley and Beissinger (2013 found avian populations decreased across all elevational gradients, species richness was lower, and compositions changed. However, woodpecker responses differed slightly from the community response with more than half not declining. The adaptive capacity of these woodpeckers is considered high (Supporting information Table S1; Foden et al., 2013), so climate change alone may not drive responses and community dynamics may not scale to the species level. Thus, accounting for two-dimensional climate space interactions (Tingley et al., 2012) and subsequent niche constraints in models is important for montane populations.  (Bled, Sauer, Pardieck, Doherty, & Royle, 2013) was attributed to maturing forest, backyard bird feeders, (Jackson and Davis Jr 1998;Meade, 1988), and planted trees in the Great Plains (Shackelford, Brown, & Conner, 2000). Although climate is likely influencing these broad-scale range changes and expansions, it is difficult to ascribe change to climate, if it can be explained by other spatially explicit variables, for example, habitat patterns (Bled et al., 2013). Currie and Venne (2017) found that among some passerines, their realized niche temperatures have changed in the last three decades and that represents changes in ambient temperature and not necessarily species movements. That is, species did not maintain more constant thermal niches through time or exhibit strong poleward shifts especially at the higher latitudes; therefore, climate change, more specifically temperature, is not always the major driver of continental species' range shifts (Currie & Venne, 2017). Moreover, observed lag responses to contemporary climate change are likely to occur in the future resulting in miss-estimations of range change based on climate condition-only models (Hovick et al., 2016;La Sorte & Jetz, 2012;La Sorte et al., 2009). Factors other than broad-scale climate are confounding distribution and habitat use responses. The mechanisms underlying observed shifts are numerous (Currie & Venne, 2017;Hitch & Leberg, 2007;Hovick et al., 2016;La Sorte & Thompson III, 2007;Tingley et al., 2009) and require further consideration, especially within modeling frameworks, if climate-induced distribution changes are to be accurately predicted.

| COMPARING CLIMATE-INDUCED OBS ERVED AND PRED I C TED TRENDS
We found that 7 of 15 species short-term breeding geographic range predictions under one or both emissions scenarios are not in agreement with observed trends ( Woodpecker climatically suitable range is predicted to contract substantially in the short term (Table 3); however, observed trends from 2005 to 2015 indicate an increasing population .
The disagreements between short-term predictions and observed trends highlight the potential incongruencies between future potential climatic niches and realized niches based on climate-woodpecker bioclimatic niche models.
We hypothesize that woodpecker responses derived from climate-woodpecker models are likely not in agreement with observed trends because additional niche characteristics (e.g., forest composition) are responding differently to climate change, and these changes are not represented in the models being used. Therefore, mismatches in observed and future trajectories will continue to arise as actual vegetation cover (i.e., habitat) differs from theoretical because of climate conditions interacting with landscape-scale processes (e.g., fire, seed dispersal; Hampe & Jump, 2011). A comparison between climate-woodpecker model projections and habitat responses of such species in climate-forest models emphasizes the potential for such inconsistencies.
For example, western montane and boreal woodpecker species such as the American Three-toed Woodpecker, Red-naped Sapsucker, Williamson's Sapsucker, and White-headed Woodpecker are predicted to lose climatically suitable habitat based on the bioclimatic niche models (Figures 1 and 2; Supporting information Table S1). Climate-forest models associated with these woodpeckers' habitats project shifts in species distribution and composition (McKenney, Pedlar, Lawrance, Campbell, & Hutchinson, 2007). In other words, climate-woodpecker models indicate a range loss due to climate change, but climate-forest models report a mixed response of the underlying habitat. Assuming tree species of this region (associated with woodpeckers' suitable habitat) track their climate niches (i.e., the climatically suitable range of woodpeckers is more closely associated with a congruent shift in vegetation), forest composition change projections are mixed leading to the potential for habitat persistence. Lodgepole pine (Pinus contorta), black spruce (Picea mariana), and aspen geographic ranges will likely decline (Coops & Waring, 2011a, 2011bMcKenney et al., 2007;Rehfeldt, Ferguson, & Crookston, 2009), ponderosa pine (Pinus ponderosa) range projections show mixed results (Coops & Waring, 2011b;McKenney et al., 2007), and Douglas fir (Pseudotsuga menziesii) range is predicted to increase (Coops & Waring, 2011b;McKenney et al., 2007). However, tree species will exhibit some level of delayed climate niche tracking (McKenney et al., 2007) because tree species migration will likely not keep pace with projected climate change (L. R. Iverson, Schwartz, & Prasad, 2004). This will result in a lag effect between changing climatically suitable geographic range and subsequent woodpecker species colonization because contemporary vegetation patterns will not perfectly track climatic shifts. This will increase the likelihood of the persistence of suitable habitat or refugia (Beever et al., 2016) through the 21st century, which are undetectable with bioclimatic niche models (Wiens & Bachelet, 2010).
Using climatic conditions associated with contemporary distributions can under-predict the areas that are suitably post-climatic change (Early & Sax, 2014) because landscape-scale processes can cause a lag in vegetation (Wu et al., 2015) or animal (Menéndez et al., 2006) responses. Processes that create a mismatch between expected and actual vegetation could result in the persistence of suitable habitat patches that mitigate short-term climate change pressures on some populations (Kellermann & van Riper, 2015).
For example, fire potential and frequency are predicted to increase across most of the United States and more specifically the Rocky Mountains (Liu, Goodrick, & Stanturf, 2013;Rocca et al., 2014). This is proposed to fundamentally change the western U.S. fire regime to dynamics not observed in the historical and paleoecological record, that is, a novel fire-climate-vegetation relationship is predicted (Westerling, Turner, Smithwick, Romme, & Ryan, 2011). Bioclimatic range projections can track climate change assuming processes occurring under current climatic conditions persist. However, bioclimatic niche models do not fully capture the shifting woodpecker niche constraints resulting from novel climate-vegetation-disturbance interactions. It is possible that increases in fire severity and or frequency may be beneficial to some woodpecker species in the western United States (Hutto & Patterson, 2016) and that climatic changes that do not pose direct physiological constraints on woodpeckers may result in suitable habitat via forest composition and structure changes. Therefore, accounting for vegetation and the ecosystem processes underlying vegetation dynamics is important in the climate-woodpecker-forest integration framework.
There are instances where climate-woodpecker models agree with observed trends, and future predictions are supported by climate-forest projections of the underlying habitat vegetation composition. However, the mechanisms underlying these observed and predicted trends are nuanced and identifying them will improve model integration. For example, the Yellow-bellied Sapsucker has short-term predictions that are in agreement with observed trends (Table 3) and long-term predictions indicate range contractions Matthews et al., 2011). The Yellow-bellied Sapsucker has been increasing in abundance at its southern range extent since 1966 , shifting south, expanding east, and increasing in geographic range (Bateman et al., 2016;Hitch & Leberg, 2007;Zuckerberg et al., 2009), though this is despite climatic factors (Supporting information Table S3; Stephens et al., 2016).
They favor early-successional forests and are currently increasing because of the reversion of post-European settlement agricultural land use to forests (Walters, Miller, & Lowther, 2002). The contemporary geographic breeding range is projected to decrease by 2080 and shift north under the highest emissions scenario (A2 model; Figure 1); this will result in an overall geographic range reduction of 31%  and a breeding range almost entirely in Canada (National Audubon Society, 2017). Further, the predicted decline (Supporting information Table S1) is in agreement with results from a climate-woodpecker-forest model for the eastern and northeastern regions of the United States Rodenhouse et al., 2008), which represents the southern portion of the breeding range.
Although these climate-forest bioclimatic niche tree models may suffer from under-prediction errors (Early & Sax, 2014), a process-based model of these forest ecosystems indicates a seral stage shift (Thompson, Foster, Scheller, & Kittredge, 2011), which will affect Yellow-bellied Sapsucker habitat suitability. The contemporary early-successional forests of the northeast United States will change by midcentury; at the southern edge of the Sapsucker's breeding range, a shift toward late-successional species is expected and possibly accelerated as climate change has a net positive impact on growth (Thompson et al., 2011). In addition, the contemporary Sapsucker population is likely above historical size because of the large-scale changes in land use post-European colonization . It is likely the current population size and range extents are not sustainable because of antecedent land use change and forest succession; however, climate change will synergistically interact with successional trajectories.
The predicted declines of climatically suitable range of the As the niche constraints (e.g., forest composition, structure) associated with woodpeckers respond to climate change (Ganey & Vojta, 2012;Westerling, Hidalgo, Cayan, & Swetnam, 2006), climate variables may poorly approximate woodpecker species responses compared to measures of ecosystem dynamics, for example, forest net primary productivity (Tingley et al., 2009) or forest composition. Therefore, ecosystems predicted to be climatically unsuitable (per bioclimatic niche models) but predicted to maintain or increase key habitat species or functions (per process-based climate-forest models) may still be suitable habitat for woodpeckers because of resource persistence. Accounting for associated niche constraints in a climate-woodpecker-forest modeling framework will produce more informative responses.

| FR AME WORK INTEG R ATI ON
Development of forest management strategies aimed at increasing or preserving wildlife species in a changing climate requires modeling efforts that include the coupled response of vegetation and wildlife to climate change. We suggested woodpeckers as indicator species of forest resiliency and biodiversity in an integrated forestwildlife modeling framework, because they are ecologically constrained by forest structure, composition, and processes, which also affect a diversity of other organisms. Based on our comparison of predicted and observed woodpecker responses to climate change, we propose a framework for integration of climate, woodpecker, and forest modeling (Figure 3).

Models used to project future abundances and distributions
of North American woodpecker species have largely been developed independently of process-based models of forest vegetation responses to climate change (Table 1; Figure 3). The available bioclimatic niche models that predominate the predictions about woodpeckers (Figure 3b) provide potential broad-scale range distribution trends (Pearson & Dawson, 2003) and ground cover composition (James, Hess, Kicklighter, & Thum, 2011), which are indistinguishable at the scale of plant functional groups. Therefore, even with the persistence of the needle-leaved evergreen biome or long-leaf pine successional stages within this region (Costanza, Terando, McKerrow, & Collazo, 2015), finer scale niche attributes are important (Schiegg et al., 2002) and should be included in model integration.
Dynamic community process-based forest landscape models (Scheller & Mladenoff, 2007) such as the LANDIS models (LANDIS-II and LANDIS PRO; Figure 3a) that incorporate finer scale climate-vegetation-disturbance interactions compared to bioclimatic DGVMs are ideally suited for this integration (Di Febbraro et al., 2015;LeBrun et al., 2016;Tremblay, Boulanger, Cyr, Taylor, & Price, 2018). These models could improve woodpecker distribution modeling, especially within the context of multi-objective management scenarios (Martin, Hurteau, Hungate, Koch, & North, 2014). Many of the key habitat characteristics and processes (e.g., forest composition and structure, disturbance type, intensity, and temporal trends) that modulate woodpecker population responses are already output variables of forest landscape models, allowing for points of integration between the two modeling disciplines (Figure 3a,c). In addition, these models can be modulated by climate data, which is the crucial integration element in the climate-woodpecker-forest framework (Figure 3d).
Integration examples support this proposed framework. LANDIS-II model projections by Martin et al. (2014) found that managing longleaf pine habitat for carbon storage decreases biodiversity and Redcockaded Woodpecker habitat at the expense of increased carbon sequestration. Similarly, the Black-backed Woodpecker in boreal forest of Canada are predicted to decline under climate change or business as usual harvest practices (Tremblay et al., 2018). The LANDIS models ( Figure 3a) allow for climate data integration, simulate ecosystem processes that produce emergent vegetation dynamics that constrain woodpecker distributions, and output variables that can inform woodpecker-forest models (Figure 3c).
In summary, after evaluating the predicted and observed woodpecker trends associated with climate change, we found there are inconsistencies between climate-woodpecker predictions and observed woodpecker responses, highlighting the uncertainty of future woodpecker distribution and population predicted responses.
We conclude that implementation of climate smart management strategies aimed at increasing or preserving wildlife species will require modeling efforts to include the coupled response of climatewildlife-forest (Figure 3). The use of an indicator species of climate effects on forest biodiversity and resiliency is an improvement to ecosystem modeling. The general principle of coupled modeling frameworks is not a new proposal with regard to climate change (Root & Schneider, 1993). However, to date, we are aware of no model (Figure 3d) that has managed to fully combine wildlife niche modeling into a climate-forest model; meaning modeling activities have utilized multiple models in tandem with data handoffs rather than have the models interact with feedbacks to processes. Our review suggests that fully integrating climate-woodpecker-forest models will address the limitations of climate-woodpecker models, while providing a biodiversity measure for climate-forest modeling efforts. Selection of the proper models within the framework will improve the resolution of fine-scale woodpecker population responses to climate change and support multi-objective management through integration of a habitat evaluation metric.

ACK N OWLED G M ENTS
We thank Jeffrey Stenzel, Betty Kreakie, and two anonymous reviewers for valuable input and comments on this manuscript. This work was supported by the NSF Idaho EPSCoR Program and by the National Science Foundation under award number IIA-1301792 and USDA NIFA McIntire-Stennis project 1004594.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
EW performed all data acquisition and synthesis. EW and TH wrote the manuscript. All authors contributed to editing and revising the manuscript. All authors read and approved the final manuscript.

DATA ACCE SS I B I LIT Y S TATE M E NT
All summarized data in this study are available in Supplemental   Tables S1-S4.