A fragile future for pink birds: habitat suitability models predict a high impact of climate change on the future distribution of flamingos

ABSTRACT Climate change is one of the most impactful global phenomena, affecting multiple ecosystems, particularly wetlands and water bodies, as well as important species that depend on these areas. Flamingos are unique and distinctive species that live exclusively in these environments and are highly impacted by any changes in their breeding or non-breeding wetlands. To address and measure the direct impact that future climatic changes could have on the distribution of the six extant species of flamingos, I used citizen science data and climatic variables to construct habitat suitability models. These models were used to predict the future gain or loss of climatic suitability areas in the short, medium, and long term, under four different Shared Socioeconomic Pathways and eight Global Circulation Models. The results predicted that five out of the six species of flamingos will experience continued loss of habitat over the next few decades in all scenarios. Dramatic changes in distribution are expected for all species. The data also indicate a higher impact of climate change on more habitat restrictive species and on wetlands along the borders of their distributions. Finally, the research highlights the importance of combined efforts from public communities, scientists, and policymakers to create mitigation and conservation plans that could avoid the intensification of climate change effects on wetlands and prevent the future reduction of flamingo populations.


Introduction
Climate change is one of the most pressing issues facing the planet and is having significant impacts on all ecosystems and species (Martin and Maron 2012;Thornton et al. 2014).Wetlands are among those ecosystems that are expected to be the most impacted by climate change, with projections that nearly 70% of these ecosystems could disappear by 2100 under current scenarios (Kingsford et al. 2016;Salimi et al. 2021).Wetlands are important habitats for many species of animals, including fish, birds, and mammals, which are often already threatened by human activities (Greb et al. 2006;Junk et al. 2006;Janse et al. 2019).Wetlands also provide a variety of important ecological services to humans, such as acting as natural filters, removing pollutants and sediment from water, regulating water flow, and preventing flooding (Neri-Flores et al. 2019;Bhowmik 2022).Additionally, these ecosystems often provide economic sustainability to local communities through activities such as extraction, fishing, and the use of water for domestic or agricultural purposes (Adger and Luttrell 2000;Schuyt 2005).However, the changes in rainfall patterns and temperature caused by climate change in recent decades have put all of these benefits at risk, potentially leading to drastic hydrologic alterations and, in some cases, complete drying of waterbodies around the world (Salimi et al. 2021).
Climate change is also having a significant impact on bird populations that depend on wetlands (Žalakevičius and Švažas 2005;Galewski and Devictor 2016).Drastic temperature changes and water quality alterations promoted by climate change can also reduce food resources or even lead to eutrophication of waterbodies, forcing animals to leave or threatening their survival.High mortality of wetland birds has already been observed in Little Auks (Alle alle), and Tree Swallows (Tachycineta bicolor) (Shipley et al. 2020), possibly as a consequence of changes in resources availability due to climate change (Bowler et al. 2019;Riddell et al. 2019).Furthermore, rising sea levels and increased frequency of flooding/drought events are making it difficult for some birds to find suitable breeding and nesting sites, directly affecting their ability to remain viable in the future, threatening their populations with interbreeding and the risk of extinction, as has already been reported for Greater White-fronted Goose (Anser albifrons) and Mallard (Anas platyrhynchos) (Von Holle et al. 2019;Teng et al. 2021;Kahara et al. 2022).
Flamingos are one of the world's most recognisable bird species, found exclusively in wetlands and shallow lakes in Africa, South America, southern Europe, western Asia and the Caribbean (Bildstein et al. 1993, Baldassarre andArengo 2000;Caziani et al. 2001).They are known for their distinctive pink to reddish plumage and unique behaviours, related to their gregarious and highly social lifestyle (Delfino and Carlos 2021).Of the six species of flamingos, currently only two of them are considered safe from the threat of extinction, the Greater Flamingo (Phoenicopterus roseus) and Caribbean Flamingo (Phoenicopterus ruber) (IUCN 2023).The other species are considered either near threatened, such as the Chilean Flamingo (Phoenicopterus chilensis), the Puna Flamingo (Phoenicoparrus jamesi), and the Lesser Flamingo (Phoeniconaias minor) or Vulnerable (Andean Flamingo Phoenicoparrus andinus) (IUCN 2023).By depending on wetlands, flamingo populations are currently exposed to threats such as pollution, excessive economic exploitation of waterbodies, human disturbance, among others, as well as the impact of climate change (Nasirwa 2000;Valqui et al. 2000;Kingsford et al. 2016;Kumar and Rana 2021).
To predict and measure the potential future impacts of climate change on various ecosystems and environments, scientists and climatologists have developed different climate change scenarios, taking into account varying levels of carbon emissions and patterns of pollutant circulation on the planet, from optimistic to pessimistic projections (Wiens et al. 2009;Giling et al. 2019).In this research, my aim was to use these projections to predict the changes in climatically suitable habitat for the six species of flamingos in the future and how their distribution will be affected by climate change in the short, medium, and long term.Additionally, I aim to examine the ways in which climate change is impacting these wetlands and flamingo populations, and to provide insight into the steps that need to be taken to protect these iconic birds and their habitats.

Occurrence data collection and filtering
The occurrence data for the six species of flamingos from 2002 to 2022 was collected from the Global Biodiversity Information Facility (GBIF) (https://www.gbif.org/)database.I chose this website because they provide a large amount of open-access data, mostly contributed by citizens, birdwatchers, institutions, and zoological collections, and which include observations from iNaturalist (https://www.inaturalist.org/)and eBird (https://ebird.org/home)(Kosmala et al. 2016;Ding et al. 2022).Using citizen science data comes with certain limitations that need to be considered.Firstly, the data collected by citizen scientists may be biased and inconsistent in terms of sampling effort and coverage across regions (Tulloch et al. 2013;Jäckel et al. 2021).Additionally, citizen science data may be subject to spatial and temporal biases, with certain areas or time periods being over-or underrepresented in the observations, which can hinder the assessment of trends (Tulloch et al. 2013;Loss et al. 2015).Despite these limitations, citizen science data can still provide valuable insights into bird distribution and demography and can be a valuable tool in conservation policy and management decisions (Greenwood 2007;Dickinson et al. 2012).
The data on flamingos in GIBF were filtered to eliminate inaccuracies and decrease the potential bias in the model.The selected time frame aimed to represent the most recent distribution of the species and avoid outdated, inaccurate, or vague reports, as GPS and coordinates were not always available or reliable in the past (Johnston et al. 2021).Only data points with detailed coordinates from human or direct observation of the species were used, and data points with excessively rounded or invalid coordinates, or those outside of their normal distribution were excluded from the analysis (Johnston et al. 2021).I used the current distribution areas for each species to include only points within these areas, removing many records of vagrants, occasional sightings, or misidentifications outside the species' normal range (Leitão and Santos 2019).To make the analysis feasible, I selected only points within a 10 km buffer of waterbodies or wetlands, as these are the main habitats for flamingos; occurrences outside of this limit are likely to be errors or rounded coordinates (Morganti et al. 2019;Ferrarini et al. 2023).To avoid spatial autocorrelation in the data, only one occurrence point per pixel was used in the maps, avoiding overrepresentation of wetlands with many records and underrepresentation of wetlands with fewer records (Hui et al. 2006;Harisena et al. 2021).
The spatial resolution of maps was of 2.5', the same as the resolution of bioclimatic variables (see below).The original distribution area shapefiles for the species were obtained from IUCN (2023), and the shapefiles for water bodies and wetlands of the world were obtained from the Global Lakes and Wetlands Databases (Lehner and Döll 2004).Geospatial point filtering was performed using QGIS (QGIS Development Team 2022), and the correction of spatial autocorrelation was performed using the 'spapstat' package in program R (Baddeley and Turner 2005).

Bioclimatic data collection
Current bioclimatic spatial data were obtained from WorldClim v2.1 (https://www.worldclim.org/data) and used to adjust the habitat suitability models (Fick and Hijmans 2017).Nineteen different bioclimatic variables at a 2.5' spatial resolution were initially used in model construction, including mainly temperature and precipitation attributes, such as annual average temperature, daily temperature variation mean, isothermality, temperature seasonality, maximum temperature in the hottest month, minimum temperature in the coldest month, annual temperature range, mean temperature of the wettest quarter, mean temperature of the driest quarter, mean temperature of the hottest quarter, mean temperature of the coldest quarter, annual precipitation, precipitation in the wettest month, precipitation in the driest month, precipitation seasonality, precipitation of the wettest quarter, precipitation of the driest quarter, precipitation in the hottest quarter, and precipitation in the coldest quarter (Waltari et al. 2014;Fick and Hijmans 2017).All bioclimatic variables were cropped to the extent of the distribution of each flamingo species using QGIS (QGIS Development Team 2022).

Model construction
I calculated a Variance Inflation Factor (VIF) for the bioclimatic variables, verifying the presence of multicollinearity among the response variables utilised in each model (Viana et al. 2022).Only variables with VIFs lower or equal to five were used, while the others were removed (Zuur et al. 2010).Using the selected variables, habitat suitability models were constructed for each species using the MAXENT algorithm, a method that shows high performance in adjusting habitat suitability models using animal occurrence data when compared with other similar algorithms such as Random Forest or Generalized Additive Models (GAM) (Elith et al. 2006(Elith et al. , 2011;;Phillips et al. 2006).Each model was run 100 times with 10 replicates of pseudo-absence records with the same amount of occurrence data.The pseudoabsences were weighted equally to presence records, since unbalance records can reduce the accuracy of model selection and prediction (Elith et al. 2006).Since all species had higher than 100 occurrence points, I used 70% of randomly selected records to train the model; meanwhile, the remaining 30% were used to evaluate model validity (Elith et al. 2006;Wenger and Olden 2012).Based on the tests and model adjustments, the model with the lowest Akaike Information Criterion (AIC) and those with Area Under the ROC Curve (AUC) values higher than 0.75 were selected for each species of flamingo (Elith et al. 2006).
To predict current and future ecological distribution and habitat suitability for flamingo species, the same bioclimatic variables were collected from WorldClim v2.1, considering eight different scenarios of Global Circulation Models (GCM) (BCC-CSM2-MR; CanESM5; CNRM-CM6-1; CNRM-ESM2-1; IPSL-CM6A-LR; MIROC6; MIROC-ES2L and MRI-ESM2 -0) and four different Shared Socioeconomic Pathways (SSPs), from the most optimistic, SSP1-2.6, to the most pessimist, SSP5-8.5, but also including the intermediate scenarios: SSP2-4.5 and SSP3-7.0(Fick and Hijmans 2017).The predictions included three different time periods: short term (2021-2040), medium term (2041 -2060), and long term (2061 -2080).For each SSP and time, I constructed a consensus model of all GCM, using the arithmetic average of each model.Lastly, for each species, 13 different models were constructed (one for the current distribution and 4 for each future period considered).The models were cropped to the current IUCN distribution areas for each species and binarized using the 10 th percentile training presence threshold (Elith et al. 2006).
The VIF calculations were performed using the package 'usdm' and models were constructed and trained using the package 'sdm', both in R-studio software (Naimi and Araújo 2016).

Measuring distribution changes
I examined the potential changes in the climatic suitability area for a certain species over time by comparing the current and future climatic distributions in the binarized projection generated above (Collins et al. 2012).I calculated the areas of the current (Ac) and future (Af) distributions for each period and each SSP projected, as well as the area of overlap between them (Acf).With these measures, I calculated the relative area gained Equation (1), the relative area lost Equation (2), and the relative balance between gains and losses for each species in each distribution projected, adapted from Pinto et al. (2023).Negative values of B indicated loss of distribution area in the future, whereas positive B values indicated that the species will increase its distribution area in the projected scenarios (Pinto et al. 2023).
I used chi-square tests to verify if different future scenarios were significantly different from each other in each species, impacting the B values differently.I also used chi-square tests to evaluate whether the B values were significantly different from zero and each other among the different time periods considered in the analysis, by species.Finally, to verify if the total change and balance were linked to the size of IUCN distribution areas, I performed a Spearman correlation test between the log of geographical extent and the relative total change in area.Geospatial measures were performed using QGIS (QGIS Development Team 2022), while the statistical analysis was performed using the base statistical package of program R (R Core Team 2023).For all tests of significance, an alpha of 0.05 was considered (Zuur et al. 2010).

Results
I collected a total of 413,958 observation records for the six species of flamingo, with the species with most records being Greater Flamingo (n = 255,216) and Caribbean Flamingo (n = 72,372), with the least recorded species being the Andean Flamingo (n = 4,016) and Puna Flamingo (n = 3,450), both from South America.After the data filtration process, the points selected for the model construction for each species were 7,684 for the Greater Flamingo, 1,405 for the Caribbean Flamingo, 4,604 for the Chilean Flamingo, 554 for the Andean Flamingo, 505 for Puna Flamingo, and 2,490 for Lesser Flamingo.With the selected points, initial tests were conducted to verify the VIF of the bioclimatic variables for each species, selecting only the variables with a VIF less than 5.For all species, the variables Mean Diurnal Range (BIO 2), Mean Temperature of Driest Quarter (BIO 9), Precipitation of Driest Month (BIO 14), and Precipitation Seasonality (BIO 15) were selected, with the other selected variables appearing independently for only some species (Table 1).After the selection of occurrence points and the elimination of multicollinear variables, the models for each species were constructed and tested (Table 1).The models selected for each species included the ones with lower AICc and AUC higher than 0.75.In general, variables such as Mean Diurnal Range (BIO 2) and Mean Temperature of Driest Quarter (BIO 9) were the most relevant for the habitat suitability models and were the ones with more importance in the selected models, despite a slight difference between the species (Table 1).For Puna and Andean Flamingo, the Precipitation Seasonality also appear to be important in the models.
After the model selection and prediction construction, I obtained 13 binarized distributions for each species of flamingo studied, including one current and four future predictions for each time interval studied (Figures 1 and 2).The current distribution area of flamingos varied from 14,607,733 Km 2 in the case of Lesser Flamingo to 524,873.8Km 2 in the case of Puna Flamingo (Table 2).In all scenarios and time intervals studied, five of the six species of flamingos are predicted to lose area, suggesting that climate change will lead to a decrease in the distribution of these species (Table 2).In the short term, the species that will lose more area are the Puna Flamingo (� x= −12.27%) and the Lesser Flamingo (� x = −12.62%),even in the more optimistic scenarios, while in other cases the losses will be less drastic or even positive in the near future, such as in for Chilean Flamingo (� x = −5.05%),Caribbean Flamingo (� x = −6.94%),and Greater Flamingo (� x = 5.37%) (Figures 1 and 2).
For the medium and long term, the effect of climate change on the distribution of flamingos will be accentuated.In the medium term, species such as the Andean, Lesser, and Puna Flamingo are expected to experience a change in suitable area of −10.32%, −18.74%, and −21.38%, respectively, which will be accentuated in more intermediate or pessimistic CO2 emission scenarios (Table 2) (Supplementary Material 1).By 2080, species such as Lesser, Caribbean, and Puna Flamingo will lose more than a quarter of their original areas in the more optimistic scenarios, and up to almost half of their current distribution in more pessimist scenarios (Table 2).In all scenarios, the only species that is predicted to gain area is the Greater Flamingo, gaining a mean of 4.01% by 2080, despite experiencing drastic changes in suitable area, mainly in Central Africa (Figure 2).Despite experiencing overall gains, total gains experienced by the Greater Flamingo will decrease in the more pessimist scenarios.
Gains and losses predicted for each SSP were significantly different among them during the three different periods studied, indicating that the magnitude of impacts is related to the severity of the climate change (Figures 1 and 2).There were significant differences between almost all the scenarios tested in suitable area gains and losses (Supplementary Material 2).In relation to the correlation between the balance predicted by the models for each species and their original distribution area, a negative tendency was found between area and the potential distribution loss in the future in at least 2 time periods of each scenario, indicating that species with less current distribution were the ones that lost more in future scenarios, thus, the ones in more danger, despite no significant results were obtained due the low number of samples tested for each period (Figure 3) (Supplementary Material 3).
Table 1.Climatically suitability models were performed using citizen science data for each species of flamingo to further predict the distribution area changes in the future.The model selected for each species used two main criteria: the lowest akaike Information criterion (AIC) and an area under the ROC curve (AUC) value higher than 0.75, indicating a better fit for the presence data used in the model construction (N).For each model, a series of bioclimatic variables with a variance inflation factor lower than 5 were used, with different importance for each variable.

Discussion
The results showed that five of the six flamingo species, which are directly dependent on wetlands for breeding and survival, are expected to lose suitable areas by 2080, and all of them will experience significant changes in their distribution areas, potentially affecting the survival of many of these species' populations.For the six species of flamingo, three of them raised most concern due the decreasing populational trends in the last decades (IUCN 2023), specially the three South American species: Andean Flamingo (estimated population: 77,949 animals), Puna Flamingo (estimated population: 154,001 animals), and Chilean Flamingo (estimated population: 515,530 animals) (Marconi et al. 2021).From the mid-80s to the mid-90s, population of Andean Flamingo declined from near 100,000 to only 34,000, and the populations of Chilean Flamingo and Puna Flamingos faced a drastic population decline in the end of the 20th century, mainly due the quick urbanisation of wetlands surrounding areas and the changes in the environmental conditions for successful breeding (del Hoyo et al. 1992;Rocha 1997;Valqui et al. 2000;Marconi et al. 2011).In more recent years, the effort for wetland conservation and the increasing research in flamingo ecology and management lead to the recovery and stabilisation of some populations, despite some breeding and wintering colonies still present decline trends in current days (Marconi et al. 2021;Delfino et al. 2023).
Nevertheless, the true populational assessment for flamingos is still hard due the large distribution of the species, the vagrant behaviour of animals and the inaccessibility of many wetlands, preventing scientists to better measure the true impact of climate change on these species (Rocha 1997;Valqui et al. 2000).
Reports of massive droughts are already emerging from some flamingo-inhabited lakes in South America, such as Lagoa do Peixe in southern Brazil, Poopó lake and Colorada lagoon in Bolivia, and lake Pozuelos and Llancanelo lagoon in Argentina, where the loss of water has already had a negative impact on flamingo presence (Mascitti 2001;Rocha et al. 2009;Carbone et al. 2015;Satgé et al. 2017;Alvarez Guerrero et al. 2018;Delfino and Aldana-Ardila 2020;Zubieta et al. 2021).Nevertheless, it is predicted that these phenomena will continue in the short, medium, and long term, reducing the suitability of many of these wetlands for Andean, Puna, and Chilean Flamingo populations (Junk 2013;Salimi et al. 2021).Furthermore, the drought in feeding and foraging areas is also expected to affect important sites for flamingos in Central Africa, Northern Europe, and India, directly impacting some populations of Lesser and Greater Flamingos, despite their currently large populations numbers (respectively 2,730,000 animals and 938,000 animals) (Schuyt 2005;Čížková et al. 2013;Belle et al. 2018;CSR7 2018).
Table 2. Percent change in area of occurrence for each species of flamingo, based on prediction maps that were binarized using the 10 th percentile training presence threshold; 13 different models were constructed taking into consideration four different Shared Socioeconomic Pathways (SSPs), three for each time period studied and eight different global circulation models (GCM), combined; SSP1 is considered the most optimistic scenario, while SSP5 is the most pessimistic.Although non-breeding areas are key to the future of flamingo populations, it is the disappearance of breeding areas and breeding conditions that most directly impacts population stability and maintenance of all flamingo species (Kitaysky and Golubova 2000;Carey 2009).Flamingos are highly specialised breeders and depend on a specific set of conditions for courtship, breeding, and nest-building, mainly related to water conditions, water levels, and food availability, as already reported for species such as Chilean, Caribbean, Lesser, and Greater Flamingos (Brown 1971;Pickering et al. 1992;Cezilly et al. 1995;Studer-Thiersch 2000;Derlindati et al. 2014).The alteration of these conditions, caused by severe drought or excessive rain, can cause flamingos to skip breeding seasons, sometimes for multiple years, to avoid raising offspring under adverse conditions (Studer-Thiersch 2000;Tavecchia et al. 2001;Pradel et al. 2012).Furthermore, stressful climatic events and environmental changes during the breeding season can also potentially lead to nest abandonment and death of nestlings (Béchet et al. 2009;Scott et al. 2012).For instance, in 2010, from the sixteen reported location for Andean Flamingo nesting, only six presented active breeding colonies and from those, in only one of them chicks fledge the nest, the same patterns observed for Puna Flamingos and Chilean Flamingos, that also all reported nest and eggs abandonment (Lee et al. 2011).In eastern and Central Africa, models predict a change in suitability for important breeding areas of Lesser and Greater Flamingos, such as for the Nakuru, Bogoria, and Elmenteita lakes in Kenya, the Makgadikgadi and Sua Pan wetlands in Botswana, and the Etosha Pan in Namibia (Mitchell 2013).In South America, areas such as Colorada lagoon in Bolivia, Mar Chiquita in northern Argentina, Surire lagoon and Los Flamencos wetlands in Chile are also crucial areas expected to change in the future (Junk 2013;Xi et al. 2021).
Caribbean Flamingos are, among all flamingo species, the one with the most particular case concerning the effects of climate change, according to the multiple scenarios provided by the models.Although they are currently considered a non-threatened species with an increasing population size (population estimated: 307,490 animals) (IUCN 2023), Caribbean Flamingos still experienced a loss in climatically suitable areas, particularly in important breeding areas such as Ría Celestun and Ría Lagartos in the Yucatan Peninsula, Mexico (Torres-Cristiani et al. 2020), which predicts a potential future decline for this species, despite an increase in suitability in some areas on the northern coast of South America and southern USA.The Caribbean Sea and the Gulf Coast, which lie within a central part of this species' distribution, are part of what is known as Hurricane Alley, a region where high sea temperatures and wind flow patterns, particularly during the summer, make it easier for tropical storms and hurricanes to form (Cooper 1992;Lugo 2000).The occurrence of these events in the region makes wetland conditions unstable for Caribbean Flamingos and can be considered one of the main threats to the species (Cambers 2009;Taylor et al. 2012).However, the outlook for the climate change scenarios is for these types of events to increase not only in frequency but also in intensity, further affecting the future populations of flamingos in the area (Lewsey et al. 2004;Day et al. 2008).
The abandonment of nests, the disbandment of colonies, or the skipping of breeding seasons directly impacts the population numbers of flamingos by causing a decrease in birth rates and promoting an increase in mortality (Saether et al. 2004;Jenouvrier et al. 2018).I expect that these effects will have an even greater impact on species such as Andean, Puna, Chilean, and Lesser Flamingos, which are already undergoing population declines and are already considered Vulnerable or Near Threatened (IUCN 2023).I expect that they will also lead to a decline in the populations of Greater and Caribbean Flamingos, which are not currently considered conservation priorities.The data shows that for the most threatened and endangered species, the loss of climatically suitable habitats will be even more drastic, especially in intermediate or pessimistic scenarios, leading to a series of events that could potentially cause the extinction of these species in the medium or long term (Velásquez-Tibatá et al. 2013;de Moraes et al. 2020).Although the results show that the Greater Flamingo is the only species that will gain more suitable habitats than it loses, it's important to note that the analysis was restricted to climatic suitability and did not take into account microclimatic and microhabitat quality, factors that could also potentially negatively impact this and other flamingo species across their distribution (Guo et al. 2017;Jähnig et al. 2020).
Flamingos also have highly variable movement patterns, being classified as migratory, nomadic, or irruptive, with much variation between years, populations, and species (Delfino and Carlos 2021).Their movements between wetlands are largely driven by food availability and wetland conditions, but they are also affected by intra-population behaviour and colony loyalty (Johnson 1989;Pretorius et al. 2020).A common pattern observed in flamingos is their movement from breeding colonies in the centre of their distribution to more peripheral areas near the edges of their range, while others remain at breeding colonies or in nearby areas (Sanz-Aguilar et al. 2012;Delfino and Carlos 2021).The models indicate that wetlands in these peripheral areas will become quickly unsuitable, highlighting the need to prioritise these areas to maintain the populations.Examples of such areas include Melincué Lagoon in Argentina for Andean and Puna Flamingos, Lagoa do Peixe in Brazil for Chilean Flamingos, Kamfers Dam in South Africa for Greater Flamingo, and Indian wetlands for Greater and Lesser Flamingos, among others.These areas are often used as non-breeding areas, where the birds feed, forage, and rest before the breeding season, and are crucial for the social cohesion of the group and the re-establishment of optimal future flock reproductive conditions (McCulloch et al. 2003;Derlindati et al. 2014;Delfino and Carlos 2022).The disappearance and rapid unsuitability of these areas can have a high impact on these flamingos' populations, as it may also affect their breeding capacity, the health of adult individuals, and the overall fitness of the populations (Holyoak and Heath 2016).
It is imperative that measures and policies are created to protect flamingo wetlands and that mitigate the current and future consequences of climate change in these areas, particularly by identifying priority areas and preventing other activities that could exacerbate these consequences (Kingsford 2011;Kingsford et al. 2016).As a start, it is important to conduct individual impact studies in the most affected wetlands in order to assess the current threats, the actual impact of climate change, and the feasibility of actions to protect these areas used by flamingos (Moomaw et al. 2018).These studies should encompass habitat selection and suitability, water quality, demographic variation over time and space, flamingo ecology and behaviour, the relationship between the animals and nearby communities, and the political sphere (Moomaw et al. 2018;Vélez et al. 2018).Based on these studies, it will be possible to develop management plans with specific goals and objectives, which involve the scientific community, local populations, and government agencies in the conservation of flamingo wetlands worldwide, with adequate financial investment to not only reduce the human impact on these areas but also raise awareness about the importance of wetlands and flamingos (Finlayson et al. 2017).To ensure the successful implementation of these plans, it is important to establish local committees that evaluate the effectiveness of these actions and their positive impacts on both the wetlands (such as improved water quality or decreased effects of drought) and on flamingo populations (such as increased flamingo numbers, breeding success, nest viability, and behavioural diversity) (Finlayson et al. 2017;Kingsford et al. 2016;Vélez et al. 2018).Only by working together, all sectors of society can help avoid a precarious future for one of the planet's most unique birds, the flamingos.

Figure 1 .
Figure 1.Current distribution of Chilean Flamingo (a), Andean Flamingo (b), Puna Flamingo (c), and Caribbean Flamingo (d) in green (I), and the predicted areas gained, in blue, or lost, in red, for the short term (2021-2040) (II), medium term (2041-2060) (III), and long term (2061-2080) (IV), obtained and calculated from the habitat suitability models under the SSP5-8.5 scenario.Global maps obtained from Wikimedia Commons and species figures adapted from Del Hoyo et al. (1992).

Figure 2 .
Figure 2. Current distribution of Greater Flamingo (a) and Lesser Flamingo (b) in green (I) and the predicted areas gained, in blue, or lost, in red, for the short term (2021-2040) (II), medium term (2041-2060) (III) and long term (2061-2080) (IV), obtained and calculated from the habitat suitability models under the SSP5-8.5 scenario.Global maps obtained from Wikimedia Commons and species figures adapted from del Hoyo et al. (1992).

Figure 3 .
Figure 3. Correlation between the total change in area (%) and the current size of the distribution of each of the six species of flamingo (log[km 2 ]) for the short term (2021-2040) in blue, medium term (2041-2060) in orange, and long term (2061-2080) in grey, with their respective linear tendency, for each Shared Socioeconomic Pathways (SSPs).