The future of suitable habitats of an endangered Neotropical grassland bird: A path to extinction?

Abstract Global changes increasingly worry researchers and policymakers and may have irreversible impacts on Earth's biodiversity. Similar to other phytophysiognomies, natural grasslands suffer from the effects of land use changes and rising temperatures, threatening animal and plant communities. Birds, being very sensitive to these changes, are widely studied and fundamental to understand the dynamics of ecosystems in relation to climate and land use changes. The Campo Miner Geositta poeciloptera is a grassland bird endemic to the Brazilian Cerrado and threatened with extinction that has been widely studied in recent years. We analyze the decrease in its extent of occurrence (EOO) and the effects of climate and land use change to understand the environmental suitability of the species in current and future scenarios. We used 5 common algorithms to produce ecological niche models. For future predictions, we use two general circulation models for two different greenhouse gas emission scenarios with different climate policies, an optimistic (ssp245) and a pessimistic (ssp585), plus two land use models focusing on increasing farmlands and reducing native grasslands. The current EOO represents ~45% of that presented by the IUCN EOO. The models generated for the present were satisfactory (TSS = 0.77 and ROC = 0.90) and showed high environmental suitability in areas where the species is currently found and low suitability where it is already extinct. All future scenarios have reduced suitable areas for the species, and the models of a greater increase in temperature and increase in farmlands and a greater decrease in grasslands were the worse. Our results reinforce the need to care about biome awareness disparity and the importance of actively preserving grassy‐shrub areas. Apparently, the state of Minas Gerais will be the only stronghold of the species in the coming years; however, the lack of protected areas that guarantee its survival needs attention.


| INTRODUC TI ON
Global changes have been a recurring concern among researchers and policy-makers in recent years. According to the latest report of the Intergovernmental Panel on Climate Change, the world may experience an alarming increase in temperature, reaching or exceeding 1.5°C in the next two decades, leading to more dangerous extreme weather events and irreversible damage to the earth (IPCC, 2022).
Effects such as ocean acidification, rising seas, rain, and drought extremes, among others, will intensify, causing, in addition to impacts on human survival, a great loss of biodiversity (IPCC, 2022) that, together with land use change, will be the main causes of species extinction by 2100 in virtually every terrestrial ecosystem of the planet (Sala et al., 2000).
For grassland ecosystems, climate projections show a substantial increase in temperature and, added to land exploitation and habitat loss, may further compromise the composition of native species (Gibson & Newman, 2019;Sala et al., 2000). For the Brazilian Cerrado, a biogeographic province with different types of grassland, the projections are not encouraging either. Hofmann et al. (2021) show that the increase in temperature will reduce the relative humidity by ~15%, making these grasslands increasingly drier and hotter, directly affecting local biodiversity. This finding becomes even more worrying since open areas are still poorly understood and with many erroneous restoration policies (Silveira et al., 2021). Additionally, the Cerrado is one of the biogeographic provinces with great richness of animal and plant species in the world, being also a biodiversity hotspot (Myers et al., 2000).
As a group that is very sensitive to global changes, birds are widely studied and are fundamental to understand the dynamics of ecosystems in relation to climate and land use changes (Borges & Loyola, 2020;Borges et al., 2019;Marini et al., 2009a;Moraes et al., 2021). Among the Cerrado grassland birds, the Campo Miner Geositta poeciloptera is a threatened species that since 2012 has been subject to a long-term study that focused on its basic natural history (Lopes & Peixoto, 2018;Machado et al., 2017), reproductive ecology (Meireles et al., 2018, environmental endocrinology , and movement ecology (Lopes et al., 2023). Endemic to the Cerrado and with its range almost restricted to Brazil (Lopes et al., 2023), the species inhabits open landscapes with sparse grass cover and exposed soil (Lopes & Peixoto, 2018;Machado et al., 2017) and is suffering a marked decrease in its area of occurrence over the years, largely due to land use changes (BirdLife International, 2022;Lopes et al., 2009;Silveira, 2009). Marini et al. (2009a) in their study with several endemic birds of the Brazilian Cerrado, including the Campo Miner, pointed out a drastic decrease in their area of the occurrence until 2100 due to climate changes. However, the study considered only few information about the natural history of the species that were available at that time, not identifying in detail the priority areas for its conservation.
In this study, we used occurrence records obtained from different sources of data, in addition to records collected in the field, to understand the past and current distribution of Campo Miner and to discuss the possible causes of the disappearance of the species in certain regions. Furthermore, using ecological niche modeling, we investigate the impacts of climate and land use changes on suitable habitats for the species in the present and future. With this, we want to understand why its range is being reduced and where the most favorable areas for its occurrence and conservation are located. For this, we evaluated possible changes in the size and location of suitable areas in different scenarios of greenhouse gas emission policies and changes in land use. Due to the lack of prospects for a return to preindustrial emission levels (IPCC, 2022) and effective enforcement policies regarding land use, we expect to find a decrease in suitable areas in all scenarios. However, we expect that the scenario with the larger increase in emissions (and consequently a greater increase in terrestrial temperature) and greater loss of natural grasslands will result in a greater loss of suitable areas.
With this paper, together with the knowledge acquired during the last 10 years of studies with Campo Miners, we expect to better understand the possible causes of its local extinction and draw up plans for its conservation, as well as understand the ecology of grassland ecosystems in the face of climate change and its importance for biodiversity.

| Study species and area
The Campo Miner is a threatened grassland terrestrial passerine (Machado et al., 2017;Ridgely & Tudor, 2009) classified as Vulnerable in Brazil (MMA, 2022) and globally (BirdLife International, 2022). In the state of São Paulo, the species is considered regionally extinct (Alesp, 2018). For these reasons, and also for the low protection of the species in reserves (Marini et al., 2009b), it was included in the National Action Plan (PAN) for the Conservation of Cerrado Birds (ICMBio, 2021a).
The Campo Miner inhabits the more open grasslands of the Cerrado savannas (Lopes & Peixoto, 2018;Machado et al., 2017;Ridgely & Tudor, 2009), a Brazilian biogeographic province that suffers from anthropogenic impacts (ICMBio, 2021b) and climate and land use changes, with severe impacts upon the species' conservation (Hofmann et al., 2021;Marini et al., 2009a). There are also scarce records of the species for the Cerrados of Bolivia and Paraguay, from where it is known from historical specimens (del Castillo et al., 2005;Herzog et al., 2016). The Campo Miner is a habitat specialist, living in very open grasslands growing on shallow soils, which show patches of exposed soil and that suffer a high incidence of erosion processes, which expose the soil banks where the species excavate the burrows where it nest (Lopes & Peixoto, 2018;Meireles et al., 2018).
Due to its distribution, we used the entire boundary of the

| Species record data
To understand the distribution of the Campo Miner, we used the occurrence records previously collected by Lopes et al. (2023). These data are from a literature review conducted in two search engines (https://schol ar.google.com and www.biodi versi tylib rary.org) using the following keywords: Campo Miner, Geositta poeciloptera, and Geobates poecilopterus (a previous name of the species). This dataset also includes data from online sound archives (http://macau layli brary.org, www2.ib.unica mp.br/fnjv), citizen science platforms (www.xeno-canto.org, www.wikia ves.com.br, and http://ebird.org), museum databases (www.vertn et.org, http://splink.cria.org.br) and general biodiversity databases (www.gbif.org). A total of 15 Brazilian and overseas ornithological collections were also personally visited (see Lopes et al., 2023).
Due to the authors' previous field experience and knowledge about the species' natural history, habitat preference, and strict nesting requirements (Machado et al., 2017;Meireles et al., 2018Meireles et al., , 2021, we excluded from the analysis records obtained in sites that do not harbor suitable habitats for it.
We also obtained occurrences records from other ornithologists coming into direct contact with them, in addition to records made in the field by the authors themselves (see Lopes et al., 2021Lopes et al., , 2023Lopes & Peixoto, 2018;Machado et al., 2017;Meireles et al., 2018Meireles et al., , 2021. For all records found, we organized the database by location, date (<1950, 1950-1990, and >1990), and geographic coordinates. The choice of cut-off dates was based on the study by Lopes et al. (2023).
For those records to which it was not possible to accurately identify their exact geographic location, we adopted the municipal seat as a reference. After constructing our database of occurrence data, we removed all duplicate points.

| Extent of occurrence (EOO)
We used the occurrence records obtained above from 1990 to 2021 to estimate the current Extent of Occurrence (EOO) of Campo Miners. According to the IUCN (2012) definition, the EOO is "the area contained within the shortest continuous imaginary boundary that can be drawn to encompass all the known, inferred or projected sites of present occurrence of a taxon, excluding cases of vagrancy." For this, we build a convex hull polygon in the ConR package (Dauby, 2020). We also performed a comparison between the

| Variables
To build the variables database, we obtained bioclimatic data from South America from the WorldClim v2.1 database (Fick & Hijmans, 2017). We performed a test of Variance Inflation Factor (VIF) analysis using the usdm package (Naimi et al., 2014) to minimize multicollinearity in the data (the default limit of 10 was used).
We also obtained the land use models from data provided by Li et al. (2017). All data were in 2.5-min resolution, Datum WSG84, and were cropped according to the study area ( Figure 1). To minimize spatial autocorrelation and clustering, we create a 5 km buffer around each presence record using the spThin package (Aiello-Lammens et al., 2015).

| Present prediction
To predict the current environmental areas suitable for Campo Miner, we only used occurrence records with exact coordinates and after 1990 (n = 47) since the records prior to this date are located in anthropized areas or even where the species is considered extinct (Bressan et al., 2009;Lopes et al., 2009), as well as the historical records (<1950) that were also not considered because the location of the record may not correspond to the current environmental reality of the species. We built all models using the biomod2 package (Thuiller et al., 2021) and five algorithms of ecological niche modeling (ENM; Table 1) with 10-fold cross-validation. To train and test the models, 75% and 25% of the data were kept during each run, respectively. Model performance was determined by a threshold >0.7 and assessed based on the average True Skill Statistic-TSS- (Allouche et al., 2006) and receiver-operating characteristic-ROC-curve (Fielding & Bell, 1997).
Given that the engines of selected algorithms are different, we followed Barbet-Massin et al. (2012) recommendations to generate background and pseudo-absence data. For that, we divided the algorithms into two groups according to the type of "absence" data they require, as shown in Table 1, being 100 pseudo-absence points for BRT, RF, and GLM algorithms (weighting the "absences" based on the number of presences) and 10,000 background points for Maxent and SRE algorithms, in accordance to Phillips et al. (2006).
We selected the best models for ensembling based on the TSS value, maximizing sensitivity and specificity, generating five ensembles, one per algorithm (Liu et al., 2005). Afterwards, we used the weighted average of the binarized projections of the determined TSS limit, resulting in model agreement for each group, following the method used by Köhler et al. (2020). From the two ensemble model groups generated, we superimposed the maps to compute the average of the overlapping pixel values to generate a single map of the present prediction.

| Future predictions
For future predictions, we used two General Circulation Models   (2002) (CMIP6) by Eyring et al. (2016) and the data were also obtained from the WorldClim v2 database (Fick & Hijmans, 2017).
Land use models for the future were obtained from data provided by Li et al. (2017). We resorted to two different scenarios, one with a moderate increase in farmlands and a moderate decrease in grassland areas (A1B) and the other with a greater increase in farmlands and a greater decrease in grassland areas (A2). We then related each model to the respective climate scenario: (1) A1B to ssp245; (2) A2 to ssp585. Due to the lack of a period comprising the ana-

| RE SULTS
A total of 1110 occurrence records were found for the species in 67 municipalities in 6 Brazilian states and Bolivia and Paraguay. After filtering to remove duplicate points, we obtained about 264 records.
The first records of the species were obtained in the state of São Paulo in 1819 in Ipanema (today Ipanema National Forest, municipality of Iperó) and along the road to the municipality of Sorocaba.
Almost all records for the state of São Paulo were obtained before 1950, with only one record obtained between 1950 and 1990.
Records for the northwest part of Mato Grosso, southwest and eastern part of Mato Grosso do Sul were also obtained before 1950.
Most of the records, irrespective of the period considered, are from the Brazilian states of Goiás, Minas Gerais, and Distrito Federal.
Records for Bolivia and Paraguay were obtained in the years 1989 and 1938, respectively (Figure 2).

| Extent of occurrence
The current extent of occurrence with records over 1990 covers an area of 889,509 km 2 , ~39% smaller than that estimated by the IUCN (1,457,020 km 2 ). Both estimates cover the states of Minas  (Figure 2).

| Predictions
After verifying the multicollinearity of the 19 bioclimatic variables, only 7 were selected (Table 2). Annual mean temperature (BIO1) proved to be the main variable representing 64.5% of contribution to the occurrence of Campo Miners, followed by precipitation of the wettest quarter (BIO16) with 32.8% of contribution (Table 2).
All 50 models generated from the 5 chosen algorithms (10 runs for each one) had good performances, presenting an average of the values of TSS = 0.77 and ROC = 0.90, higher than the 0.7 threshold previously determined.
The estimate of areas above 50% suitable for the occurrence of Campo Miners was ~348,293 km 2 and covered, as expected, mainly the Cerrado (Figure 3) and all regions where the species was currently recorded (see Appendix S1: Figure 1), which shows a good performance of the models generated for the present.
As with the current prediction, the maps generated by the two groups of algorithms were superimposed, generating a total of four projections. They were separated by period ( Figure 2 would need to be excluded from the current EOO of Campo Miners. Therefore, the current EOO will be ~659,981.3 km 2 , about 55% smaller than the EOO proposed by the IUCN.

| Present prediction
The fact that the variables Annual Mean Temperature (BIO1) and (3) Emas National Park, Goiás; and (4) Serra da Canastra National Park, Minas Gerais. All of them harbor vast expanses of grasslands and can be of paramount importance for the conservation of the species (ICMBio, 2021b). However, in Brasília National Park the species seems already to be currently very rare (Braz, 2008, WikiAves data).
The areas with the lowest environmental suitability (<50%) are mostly in regions where the species has already been declared extinct, as follows: 1. São Paulo state: Although researchers still indicate the possibility of the species occurring in the state (Bressan et al., 2009) surveys did not find the species, and only century-old museum records are known from there (Dornas, 2020;Lopes et al., 2009). In brief, all records available for these regions occurred before 1950.
3. Bahia state (Chapada Diamantina): As mentioned earlier, only one record of the species was made there in the 1990s (Parrini et al., 1999). As it is a highly touristic area and with a high flow of people (including birdwatchers), the lack of records in the last 30 years (see www.wikia ves.com.br) seems to confirm the hypothesis of extinction in the region. The possible cause of the extinction of the species in these regions is attributed to the wiping of natural grasslands, replaced by exotic pastures and crops (Lopes et al., 2009;Silveira, 2009). The lack of proper fire management needs also to be considered, as the species is fire-dependent, occupying recently burned areas (Lopes et al., 2023;Machado et al., 2017;Willis, 2004). None of the studies point to climate change as a possible cause of the extinction of the Campo Miner.

| Future predictions
Even with differences between models, it is possible to observe a marked reduction of suitable areas for all future scenarios, mainly in the states of Mato Grosso, Mato Grosso do Sul, Goiás, and Distrito Federal. This reduction is more dramatic in the pessimistic scenario, which projects a better suitability for the state of Minas Gerais.
These results agree with other studies that evaluated the influence of climate and land use changes on bird communities in the Cerrado, which showed that the southern region of it will hold the largest refuge area for these species (Borges et al., 2019;Borges & Loyola, 2020;Marini et al., 2009aMarini et al., , 2009b. However, as acknowledged by Borges et al. (2019), the southern region is the most developed in the Cerrado, with much of its vegetation converted to pastures and crops (Sano et al., 2010), in addition to having the smallest conservation units (Françoso et al., 2015;Sano et al., 2019), which can be an impediment to the establishment of new bird populations or even to the conservation of species already existing in the region.
Even if the areas predicted by the model are indeed environmentally suitable for Campo Miners, other factors must also be considered. As a specialist grassland bird that tolerates only little changes in land use (Lopes et al., 2023;Machado et al., 2017;Meireles et al., 2018), the Campo Miner needs vast areas of natural grasslands, however, these areas are less common (Borges et al., 2019) F I G U R E 4 Future prediction of suitable areas for the occurrence of Campo Miners in two different scenarios . See all the scenarios in Appendix S1: Figure 2. and poorly represented in reserves. Protected areas are important because, by suffering less anthropogenic pressures and land use changes, they can, in addition to preserving species and environments, help to mitigate the effects of long-term climate change (Dudley et al., 2009). Although Minas Gerais is the region with the best conditions for Campo Miners in the future, currently only one protected area (Serra da Canastra National Park) can be a stronghold for the preservation of the species in the state.
Some regions identified here proved to be extremely important for the creation of new protected areas, such as the Upper Rio Grande Grasslands (URRGs), which seems to be the most important stronghold for this resident species. However, the URGGS is poorly protected, with only two noteworthy reserves in the region,  (Souza et al., 2018). This region also does not have protected areas with suitable habitats to ensure the survival of the species.
The reserves would also protect a landscape that is naturally vulnerable and highly susceptible to erosive processes that lead to the loss of soil, habitat, and biodiversity (Lima et al., 2011

| CON CLUS ION
Campo Miner is a threatened species that has been suffering a marked reduction in its range in recent years and its survival in the face of global changes is not guaranteed. The effects of local extinctions are not only limited to the impact caused on its conservation, because interspecific interactions can be affected. Campo Miners build their nests within cavities they excavate in steep soil banks that are later used by other animals as shelter or even for nesting , so its disappearance can cause impacts at the community level.
We show in this study a not optimistic future for the species since important areas with ideal environmental suitability will be considerably reduced. Even in an optimistic scenario of reduced greenhouse gas emissions and greater rigidity regarding land use policies, the species will lose a large part of its suitable areas. However, in a scenario of high emissions and high land degradation, the areas lost will be even greater, so an aggressive plan to mitigate CO 2 emissions by limiting warming to 1.5°C by the end of the century as proposed (IPCC, 2022) and greater inspection of land use with effective policies and the creation of protected areas are very important to guarantee the future of the species.
Reducing CO 2 emissions is crucial to ensure the future not only of the Campo Miner but also of several other bird species (Borges et al., 2019;Borges & Loyola, 2020;Marini et al., 2009a). However, as suggested by Marini et al. (2009a), plans are needed that assess not only climate change, but also land use change, as species distribution can be influenced by different factors. In addition, there is no point in areas that show good climatic suitability if they may have already been converted to pasture, agriculture, or even urbanized.
The state of Minas Gerais will shelter the largest area climatically suitable for the species in approximately 50 years. However, the lack of protected areas in the state with adequate habitats may be a hindrance to its preservation. As it is a nonmigratory bird (Lopes et al., 2023), it is likely that the main refuges for the species in the future will be within reserves, so the creation of new protected areas or even as mentioned by Lopes et al. (2023), a mosaic of fully protected and sustainable use reserves that cover their locations is very important to ensure the survival of Campo Miners. This is especially true in the Upper Rio Grande Grasslands, an area that is being silently devastated and that shelters the most important population of the species outside protected areas.

ACK N OWLED G M ENTS
We are grateful to many people who contributed occurrence data to our study. We also thank Ubirajara Oliveira and Argemiro Teixeira for their valuable discussions. This work was supported in

CO N FLI C T O F I NTER E S T S TATEM ENT
No potential conflict of interest was reported by the authors.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used to generate this article are available at doi.org/10.5281/ zenodo.7547448 and in public data platforms stated in the methods.