Seasonal variation in climate and landcover niches of a migratory bird

Background: Resource utilization strategies of avian migrants are a major concern for conservation and management. Consequently, the ability to predict potential shifts in migratory species distribution and migratory behaviour is a pressing challenge for predictive ecology under global changes. There are two main strategies of resource use adopted by migratory birds: niche tracking for similar environmental conditions and niche switching between different environmental conditions between seasons. Our objective was to examine whether the “niche tracker” or “niche switcher” hypothesis would better explain seasonal variations in the ecological niche breadth and overlap of the American White Pelicans (AWPE, Pelecanus erythrorhynchos). We also tested whether recent changes in the non-breeding ground land-use have altered the land-cover niche breadth of AWPE. Methods: We built Maximum Entropy Models (MAXENT) to predict the AWPE breeding and non-breeding ranges using GPS locations, climate variables, and land-cover variables. We then compared the estimated climatic and land-cover niche breadth and the overlap between the breeding and non-breeding grounds. Results: Our climate, land-cover, and combined species distribution models had a good to excellent predictive performance. Our ndings supported the hypothesis that AWPE would be climatic niche switchers. American white pelicans showed little climatic niche overlap between nesting and wintering seasons. Migrants on the breeding grounds showed broader climatic niche than both residents and migrants on the non-breeding grounds. Finally, declines in availability of food resources provided by commercial aquaculture on the non-breeding grounds appeared to increase land-cover niche breadth. Conclusions: Climatic niche switching suggests that AWPE may adapt to future climate changes with unexpected spatial distributions under global change. Future declines in wetlands and food resources may restrict AWPE spatial distributions. Future studies need to investigate AWPE demographic consequences of climate and land-use changes.


Background
Understanding habitat and resource selections by animals represents a key topic in ecology (Boyce and McDonald 1999, Manly et al. 2002, McLoughlin et al. 2010, Bell 2012). If resources are used by an organism more frequently than expected from availability, they are considered selected by the organism (Aebischer et al. 1993). Organisms often exhibit seasonal and long-term shifts in habitat and resource selection in adaptation to uctuating living environments, particularly under global changes and land-use changes (Martin 2001, Araujo et al. 2004, Tingley et al. 2009, Hoffmann and Sgro 2011, Gomez et al. 2016). In niche theory, the total living environmental conditions in which populations of organisms survive and grow de ne the fundamental ecological niche (Hutchinson 1957). The occupied or selected portion of the fundamental niche is the realized niche (Hutchinson 1957). Furthermore, ecological niche has multi-dimensions, with a subset or axis (e.g., climatic conditions and land covers) of the total living environmental conditions de ning a dimension of ecological niche (e.g., climatic niche and land-cover niche; Hutchinson 1957, Sexton et al. 2009). Empirical studies of variations in the seasonal breadth and overlap of climatic and land-cover niches may provide insight into ecological mechanisms, which enable migratory species to shift their geographic ranges under anthropogenic disturbances (Tingley et al. 2009, Gomez et al. 2016).
The long-distance latitudinal migration between the breeding and non-breeding grounds represents a model system to study seasonal variations in the ecological niche of migratory species (Cox 1985, Newton 2008. Migration not only enables migrants to take advantage of resources (e.g., food and breeding habitat) available in different seasons, but also imposes physiological challenges (e.g., inclement climate and energetic costs of migration) to migrants (Marini et al. 2013, Gomez et al. 2016, Winger et al. 2019. Therefore, seasonal variation in climate and food availability between the breeding and non-breeding grounds may be two selective forces of the evolution of migration and annual resource utilization strategies (Newton 2008). The strategies that migratory birds use to cope with physiological challenges may vary between basic extremes: niche tracking and niche switching. Migrants either track a set of conditions year-round ("niche tracker") or adapt to the new environment when moving from one ecological regime to another ("niche switcher") (Nakazawa et al. 2004, Laube et al. 2015. Niche switchers, therefore, may have a broad ecological niche with minimal overlap between the breeding and non-breeding grounds (Wiens et al. 2010). Niche trackers have more constant, narrow niches throughout the year (hence, greater niche overlap between seasons) than niche switchers. However, few studies have determined the tracking or switching strategies for both climatic and land-cover niches of migratory birds ).
The American white pelican (Pelecanus erythrorhynchos, hereafter AWPE) is a short-to middle-distant migrant and the largest ying bird of North America . American white pelicans have been of conservation and restoration interests since the 1970s, and have become a substantial economic issue more recently due to their impacts on channel cat sh (Ictalurus punctatus) aquaculture (King 2005). Since the 1990's, AWPEs have been known to use commercial aquaculture ponds consistently for feeding primarily on cat sh in the Northern Gulf of Mexico (NGOM; King 2005 (Walther et al. 2002, IPCC 2014. The decline may be exacerbated in the southern part of North America (Seager et al. 2007). The changes in precipitation regimes, in conjunction with the land-use changes (i.e., aquaculture rise and fall), are likely to affect the habitat selection and land-cover niche of waterbirds such as AWPE. For these reasons, we investigated the potential effects of reduced aquaculture activities in the nonbreeding range on AWPE habitat selection patterns. We expected that AWPE may select alternative food and shelter resources due to the decline in aquaculture water acreage in the NGOM.
Our objective was to address the following questions in this study: (i) Is the migratory behaviour of the AWPE populations better explained as niche trackers or niche switchers? (ii) Do AWPE migratory populations on the breeding grounds show wider climate niche but narrower land-cover niche (temporally more stationary) compared to AWPE on the non-breeding grounds? (iii) Is there a greater land-cover niche overlap than climatic niche overlap between AWPE on the breeding and non-breeding grounds because of tracking similar foraging habitats in wetlands? and (iv) Has the land-cover niche of AWPE on the nonbreeding grounds expanded after the decline of aquaculture activities?

Methods
Study areas and GPS tracking American white pelicans Processing and ltering of GPS data for the breeding and non-breeding grounds We ltered and thinned the presence locations to control for spatial bias and unbalanced numbers of GPS locations among individual , Williams et al. 2017). First, we identi ed the GPStracked individual with the smallest number of GPS locations among the 36 AWPEs. We then reduced the number of GPS locations of the other individuals to the smallest number by randomly sampling from all available locations of an individual. Second, we resampled each of the 36 subsets of presence locations so that each location was separated from the nearest one by a minimum distance of 5 km. In a preliminary analysis, we found that average hourly movement distance of AWPE during the active hours (0900-1700 hr) on both the breeding and non-breeding grounds was about 5 km. Last, to ensure that we excluded those locations where AWPE were ying at higher altitude above ground, we also excluded from analyses those locations in the upper decile of ying speed (> 4.45 m/s) (Illan et al. 2017).
In a preliminary analysis, we built SDMs using data on the GPS locations from migrants and year-round residents on the non-breeding grounds during winter. The results of SDM tting and niche analyses did not differ in either direction or signi cance between the migrants and residents. Thus, in further analyses, we combined the winter GPS locations of the migrants and the residents at the non-breeding grounds.
To test for the effects of the aquaculture decline in AWPE habitat use and ecological niches, we divided our AWPE GPS location data into two subsets: before and after the 2005 peak of aquaculture activities on the non-breeding grounds. All comparative analyses were based on SDMs developed with these two sets of data (hereafter, pre-and post-aquaculture). We expected that SDMs would show different land-cover niche breadth before and after the peak of aquaculture.

Environmental variables
We used two types of environmental predictors in raster format: (i) climate (14 variables) and (ii) land cover ( ve variables).
Climate: Climatic variables were obtained from the WorldClim (version 2) database (Hijmans et al. 2005; http://www.worldclim.org), which provides a variety of monthly climatic data averaged over the years 1970-2000. Three main reasons persuaded us to develop our models using the WorldClim data. First, the WorldClim data include part of the AWPE non-breeding range in the Southern GOM outside the USA, which is not encompassed by an alternative raster climatic data PRISM from Oregon State University (PRISM Climate Group 2012). Second, wind direction and strength are two main drivers of AWPE movements and behaviour (Illan et al. 2017). However, none of the alternative climate data sources included data on wind currents at the desired spatiotemporal scale. Last, we regressed data obtained from two alternative climate sources, Climate Research Unit (CRU) of University of East Anglia (Harris et al. 2014) and PRISM against the same variables from the WorldClim data. Correlations were highly signi cant (R 2 > 0.8; p < 0.001) in all climatic variables considered.
The choice of climate predictors re ected energy and water constraints on the spatial distribution of birds. We selected July for the warmest month on the breeding grounds and January for the coldest month on the non-breeding grounds of AWPE (King et al. 2017). The arrival and departure dates of AWPE seasonal migration vary substantially from year to year. The two months correspond to the time of year when AWPEs are present on the breeding and non-breeding ranges ). Initially selected climatic variables included minimum, maximum and average temperature, total precipitation, solar radiation, water vapour pressure and wind speed.
Land cover: We used North American Land Change Monitoring System (NALCMS, 2005; Homer et al., 2015) (https://landcover.usgs.gov) based on MODIS satellite imagery to obtain land cover data. We chose NALCMS also because the US Land Cover Data Base does not include the AWPE geographic range outside the USA. The classi cation of land cover types was designed with three hierarchical levels using the Food and Agriculture Organization (FAO) Land Classi cation System (Homer et al, 2015). Given the importance of the presence of water for AWPE foraging habitat, we used a layer called "wetlands" (which discriminates between different types of wetlands) from the Global Lakes and Wetlands Level 3 Database (GLWD-3) (Lehner and Döll, 2004). We also calculated the variable "distance to water" as the Euclidean distance from every tracked location to the nearest permanent water body. For each individual AWPE and for each GPS location, we extracted the selected land cover variables using ArcMap spatial analyst (ESRI, 2014).
To eliminate multicollinearity, we used a backwards stepwise process to calculate the variance in ation factor (VIF) starting from the initial pool of 19 predictor variables. We only selected predictors of VIF < 4 (Graham 2003, O´Brien 2007, Kock and Lynn 2012). With the resulting set of variables, we calculated a matrix of pairwise Pearson's correlation r and discarded any remaining predictors of r > 0.7. All land cover raster les were resampled to the same spatial resolution of climatic layers at the 30 arc-second grid cell size (equivalent to 1 km resolution). The 1-km resolution was chosen for a balance between the computational burdens and the accuracy of SDMs, given the spatial extent of North America in this study. Our selected spatial resolution is less than the average hourly movement distance (5 km) of AWPE, su cient for AWPE species distribution modelling, and is consistent with the spatial resolution of the climatic raster les (Table 1).

Results
The number of birds that satis ed the ltering and inclusion criteria for analyses were 19 migrants on the breeding ground (eight individuals with 1,154 GPS locations for the pre-aquaculture SDM; 11 individuals with 2,123 GPS locations for the post-aquaculture SDM) and 30 individuals on the non-breeding grounds (15 individuals with 1,584 GPS locations for the pre-aquaculture; 15 individuals with 1,494 GPS locations for the post-aquaculture). After the multicollinearity removal, the resulting nal set of non-correlated environmental predictors included four climatic and four land cover variables (Table 1).
Model performance evaluation with the 10-fold cross-validation showed that the TSS values ranged from 0.64 to 0.90 for all different SDMs (Table 2), which were considered to be a good to excellent performance. The average AUC values of all different SDMs ranged from 0.85 to 0.95, which were also considered to be a very good to excellent performance. Climate and the combined models performed marginally better than the land-cover models ( Table 2, Fig. 2). Table 2 The performance evaluation of species distribution models for American white pelicans using the True Skill Statistic (TSS). Average values for the 10 replicates and the associated standard deviation are shown for each model.

Model
Pre Ordination based = 0.63) between the breeding and non-breeding grounds (Figs. 3a and 4). Subsequently, we present the niche analysis of the Maxent-based methods. Mean I statistic was 0.24 for climatic niche overlap versus 0.73 for land-cover niche overlap, suggesting that migrating AWPE were likely to switch climatic niche and track land-cover niche. Our results showed the lowest niche overlap for the combined models, which appeared to be more in agreement with the niche switcher hypothesis (Fig. 3a).

Discussion
Ecological niche needs to be considered a dynamic concept and have a time dimension in its quanti cation to better understand the responses of avian migrants to global change and anthropogenic disturbances ( Our results indicated that AWPEs were a climatic niche switcher, but tended to track similar land covers more than climatic conditions between the breeding and non-breeding grounds. American white pelicans may be forced to adapt to substantially different climatic conditions between the breeding and nonbreeding grounds. Likewise, adaptation to more spatially heterogenous climatic conditions may result in broader climatic niche in temperate breeding grounds than in sub-tropical non-breeding grounds. After the aquaculture decline, AWPEs might have diversi ed wetland habitat use to compensate for decreases in food availability, broadening land-cover niches. Therefore, middle-and long-distance migratory species breeding in the northern temperate regions may need plastic climatic niches to adapt to dramatic differences in ecological conditions during the annual migration cycle. Previous studies have found evidence of climatic "niche trackers" following the same climatic Winter climate harshness is a driver of avian migration which shifts the wintering range to low latitudes (Bell 2000, Newton 2008).
Tracking food availability is an important driver of the evolution of avian migration (Newton 2008).
Habitat provide birds with food, shelter or space, and nest sites for reproduction (Fuller 2012). Given little overlap of seasonal climatic niches as shown in this study, it is unlikely that AWPE tracked similar land covers for similar thermal conditions between the breeding and non-breeding grounds. Furthermore, there is no evidence that year-round AWPE residents breed in Alabama, Arkansas, Louisiana, and Mississippi.
American white pelicans feed on shes, salamanders, and craw sh living in the freshwater wetlands (King and Michot 2002). Therefore, selection of similar foraging habitat in both the breeding and nonbreeding grounds may result in appreciable seasonal overlap of land-cover niches. It is plausible to hypothesize that the broader land-cover niches on the non-breeding grounds would be a product of ancestral exploratory movements when AWPE shifted their wintering range to the Gulf of Mexico.
Additionally, nesting activities may also restrict habitat selection and movements on the breeding grounds. Therefore, AWPE migration may be an example of the northern home hypothesis that avian migration originated from the shift of wintering grounds from temperate to sub-tropical regions (Bell 2000).
Our ndings support our expectation that the recent drastic decline in aquaculture activities starting around 2005 (King 2005) had broadened AWPE land-cover niche on both breeding and non-breeding grounds. We found that land-cover niche breadth had expanded considerably on the non-breeding grounds after 2005, suggesting that AWPE were forced to nd additional sources of food and/or shelter due to the disappearance of the aquaculture ponds. However, we also found land-cover niche expansion on the breeding grounds. A plausible explanation is that migrants will broaden land-cover niche during their breeding season as a carryover from the non-breeding grounds. We believe this is one of the most important results of our study, and has considerable economic implications for sh farming in the Southeastern US. Aquaculture, apart from being an economic driver in the region, is possibly acting as a sustainable system to maintain viable long-term AWPE populations. Finding a balance between the economic viability of aquaculture and the maintenance of the near perfect foraging habitats for pelicans seems crucial for the sustainability of the ecosystem (Murphy 2005, King and Anderson 2005).

Conclusions
Our study addressed a major concern in the conservation and future management of avian migrants, including AWPE, regarding their ability to cope with ongoing climate change. Our results suggest that   The predictions of maximum entropy models for the breeding and non-breeding ranges of American white pelicans by (a, b) climate, land covers (c, d), and climate -land covers combined (e, f). The left column (panels a, c, e) is for the breeding range and the right column (b, d, f) for the non-breeding range.
Maps show in warmer colours the higher habitat suitability for each model and each range.