Evaluation of a potential umbrella species using favourability models: the case of the endangered little bustard (Tetrax tetrax) and steppe birds

Farmland biodiversity is in alarming decline worldwide due to agriculture intensification. In this context, the umbrella species concept may help in better targeting conservation efforts, focusing on species whose requirements may best cover those of other components of biological communities. We test this idea using the little bustard (Tetrax tetrax), a strongly declining steppe bird depending on extensive agricultural landscapes of south-western Europe, to explore the degree to which its habitat requirements can predict those of other sympatric endangered steppe birds. We use little bustard and other nine species distribution data at 10 × 10 km scale in Castilla-La Mancha (the most important region for the little bustard in Spain and the EU) and habitat favourability models to identify variables explaining little bustard favourability that can robustly predict habitat favourability also for the other nine species. Models fitted with variables explaining little bustard favourability and applied on co-occurring species yielded varying performance results. Models support the role of the little bustard as umbrella species only for a part of the steppe bird community, and more precisely, for species linked to cereal and grassland-dominated landscapes, but not for landscape generalist species, distributed over mosaic landscapes including shrublands and woody crops. Results also highlight the importance of favourable extensive cereal steppes for the conservation of strongly endangered species (little and great bustard, Montagu’s harrier, pin-tailed sandgrouse, calandra lark), some of which are largely understudied (pin-tailed sandgrouse and calandra lark), despite their marginal coverage by the Natura 2000 protected area network.


Introduction
The current worldwide loss of biodiversity requires the use of surrogate approaches that help optimise the always-limited resources (funding, time, expertise) allocated to conservation action (Roberge and Angelstam 2004;Caro 2010). In this context, the concept of indicator and umbrella species can be of use to overcome such limitations (Simberloff 1998;Roberge and Agelstam 2004;Caro 2010). Umbrella species are considered a type of indicator species (i.e., species whose ecological characteristics make them useful in assessing the state of other species or the environment when it is difficult to evaluate directly) whose requirements cover those of other organisms in the community (Roberge and Angelstam 2004;Caro 2010).
An important aspect of the umbrella species concept has to do with spatial requirements, which are associated with body size. Thus, species treated as umbrella are frequently large vertebrates because their rather ample home ranges may encompass a large proportion of the space and resources required by most species in the community (Roberge and Angelstam 2004). Another criterion used is rarity (Niemi et al 1997;Rodrígues et al. 1998), so that rare species would be suitable umbrellas for certain habitats and associated communities given their habitat-specificity (although this is not the only cause of species rarity, see Fleishman et al. 2000). However, even if rare species may coexist with a larger number of co-occurring taxa, widespread ones may indeed be more effective umbrellas due to management operational reasons, as shown by Fleishman et al. (2000) with butterflies. Moreover, this study supports the use of some butterfly species as umbrellas in prioritizing habitat remnants for conservation. Nevertheless, the criteria that should be followed to identify umbrella species or groups of species continue to be debated. For example, a metaanalysis carried out by Branton and Richardson (2010) showed that, while the presence of putative umbrella species did favour richness and abundance of co-occurring species, this was not significantly associated to any of the criteria tested (i.e., level of ecological specialisation, trophic level and body size). Therefore, there is a number of species that have proven to exert an umbrella effect on co-occurring taxa, but no clear suit of life-history or other biological characteristics associated to that effect has been consistently identified.
The use of umbrella species with the aim of conserving larger sets of species or ecosystem components can follow two different approaches (Roberge and Angelstam 2004). In the multi-species approach, conservation measures are directed to a number of species whose ecological requirements delimit a range of ecological niche space that comprises the requirements of the community (i.e., they set the minimum ecological conditions required to conserve the largest possible part of the community, Lambeck 2003). This type of approach is recommended by, for example Fleishman et al. (2000) and has been used to, for example, establish the minimum corridor area needed to favour the dispersal of species belonging to different taxa in a community (Breckheimer et al. 2014). In the single-species approach, measures focus on the requirements of one single species whose ecological niche largely overlaps those of a large as possible set of co-occurring species. A good example of the monospecific approach is the conservation of sage grouse (Centrocercus urophasianus) populations and habitats, which has favoured several other components of the North American sage prairie ecosystem, from songbirds to ungulates (Copeland et al. 2014;Donnelly et al. 2017). On the other hand, it is in the context of anthropogenic systems managed for production such as agricultural, grazing or forestry landscapes that umbrella species can be of greater use to evaluate the management measures that can maximize the conservation value of such landscapes (Fleishman et al. 2000;Lambeck 2003;Hawkes et al. 2019).
Umbrella species have been usually considered or evaluated for the protection of coexisting species of the same taxon. For example, Sutter et al (2002) evaluated the role of capercaillie (Tetrao urogallus) as an umbrella for the conservation of alpine forest birds. However, more recent studies have evaluated the umbrella effect of different species across different taxonomic (Dunk et al. 2006;Breckheimer et al. 2014;Hawkes et al. 2019) and functional levels in the community (Sattler et al. 2013). Umbrella species and sets of species have been evaluated at different spatial scales and through different methodological approaches. Local studies typically involve comparison with different levels of occupancy by the umbrella species, based on results of exhaustive field surveys (Suter et al. 2002;Sattler et al. 2013;Copeland et al. 2014;Donnelly et al. 2017;Hawkes et al. 2019), while larger scale studies deal with spatial modelling of species requirements (Breckheimer et al. 2014).
Here, we used a single-species approach to assess the value of an endangered steppe bird linked to dry cereal farmland, the little bustard (Tetrax tetrax), as umbrella species for other endangered birds typical of cereal pseudo-steppes, which are extensive agricultural ecosystems and, therefore, particularly suitable for the assessment of potential umbrella species that may guide their management for conservation. Steppe and farmland birds are undergoing strong declines worldwide, mainly due to land-use change and agriculture intensification (Onrubia and Andrés 2005;Donald et al. 2006;Vorisek et al. 2010). They are considered a group of high conservation concern (BirdLife International 2018), since agricultural habitats host a high proportion of birds with threatened conservation status in Europe (Tucker and Heath 1994) and farmland birds show the steepest declines (Inger et al. 2015;Rosenberg et al. 2019). Our assessment was carried out at large geographical scale, encompassing an entire administrative region: Castilla-La Mancha (Spain). To this purpose, we used the favourability function (Real et al. 2006) on little bustard presence/ absence probability and evaluated whether the resulting favourability models could accurately predict the favourability of a set of other co-occurring steppe bird species. Favourability models allow direct comparison between species (Acevedo and Real 2012;309 Estrada et al. 2016) and have been successfully used to compare the habitat favourability of different species in different conservation and landscape planning contexts, including comparisons between ecologically close species (Acevedo et al. 2010), interacting species like predators and their prey (Real et al. 2009), and assessment of potential distribution range shifts under climate change scenarios (Estrada et al. 2016). However, to our knowledge, this is the first attempt to apply large-scale favourability models to evaluate a potential umbrella species.

Study species
The little bustard is a medium-sized, lekking steppe bird that inhabits natural and seminatural grasslands as well as dry cereal farmland across the Palearctic, from Portugal to western China (BirdLife International 2022; Morales and Bretagnolle 2022a). It has disappeared from large areas of its original distribution range during the twentieth century (Cramp and Simmons 1980;Schulz 1985;Collar 1996), and steeply declined in many others, including one of its former strongholds worldwide, the Iberian Peninsula, where the population has decreased by 50% in one decade (García de la Morena et al. 2018;Silva et al. 2018).

3
This critical situation has recently led the species to be officially classified as Endangered in Spain (https:// www. boe. es/ boe/ dias/ 2023/ 04/ 07/ pdfs/ BOE-A-2023-8751. pdf), which still may hold up to ca. 39% of the world's population (Morales and Bretagnolle 2022a), due its strong decline (ca. 13% annually, Traba et al. 2020).The species is classified as Vulnerable in Europe, as well as in the European Union (BirdLife International 2015) and listed in Annex I of EU's Birds Directive as well as in Annex II of the Bern Convention. Its current world conservation status is Near Threatened in the IUCN Red List (BirdLife Internationa 2022).
In agricultural landscapes, the little bustard behaves as a permanent or semi-permanent grassland specialist Traba and Morales 2019) that depends on agricultural habitats such as fallows, grazed pastures, mowed meadows and dry alfalfa fields (Brotons et al. 2004;Silva et al. 2014;Morales et al. 2013;Bretagnolle et al. 2018;Traba and Morales 2019). These are typically interspersed in agricultural landscapes with other habitats, which little bustards use complementarily (e.g., while females select fallows for nesting, males may use ploughs or newly sown cereal crops for sexual display, Morales et al. 2008Morales et al. , 2013Devoucoux et al. 2019). In addition, the species social dynamics linked to its exploded lek mating system may lead some individuals to have large breeding home ranges (e.g., females seeking mates or forage grounds for chicks, or satellite males, Lapiedra et al. 2011;Ponjoan et al. 2012;Tarjuelo et al. 2013). Moreover, little bustards carry out significant seasonal movements, ranging from long-distance migration to localscale resource-track vagrancy (Villers et al. 2010;García de la Morena et al. 2015;Cuscó et al. 2018). Finally, the little bustard presents a varying degree of niche overlap with ecologically similar farmland birds and is well connected in the ecological interaction network of agricultural ecosystems (Tarjuelo et al. 2022). It has been shown to interact positively with some other steppe birds (e.g., pin-tailed sandgrouse Pterocles alchata, which benefits from the little bustard's greater capacity for antipredator surveyance in winter flocks, Martín et al. 2010; Garcia de la Morena 2015), or negatively with others (e.g., it segregates at small spatial scale with with the great bustard Otis tarda, due to competition for certain habitats, Tarjuelo et al. 2014Tarjuelo et al. , 2017. Further, as an omnivorous species (Bretagnolle et al. 2022a), the little bustard is associated to rich arthropod and plant communities (weeds and grassland plants, Faria et al. 2012, Bravo et al. 2017González del Portillo et al. 2021). Therefore, the little bustard is a species with large spatial requirements and rich ecological interactions, and thus a good candidate to qualify as umbrella species and a suitable model to address under the single-species approach (Roberge and Angelstam 2004). Other traits making the species a potentially suitable umbrella are: (i) its still large distribution range in many regions (e.g., the species is present in ca.

Study area
Our study is based on species distribution data from Castilla-La Mancha (central Spain, Fig. 1). Castilla-La Mancha is the third largest administrative region of Spain, with 79,463 km 2 . Climate in this region is overall continental Mediterranean with cold winters and hot summers. Thermal variation is large, both daily and seasonally. Precipitation is concentrated in the fall and spring, and in the main central plains hardly reaches 400 annual mm (Agencia Estatal de Meteorología AEMET 2020). Except for the northern, southern, and south-western mountain and upland fringes, most of its territory comprises open and flat or gently undulated dry farmland dominated by cereal cultures but also with a significant cover of vineyards and olive groves. In fact, the region's traditional landscape was a mosaic of several cereal rotation stages and the two mentioned woody crops (see, for example Suárez et al. 1997). Besides its core mosaic cereal farmland, Castilla-La Mancha harbours other important habitats for steppe and grassland birds like extensive pasturelands and high-plateau shrub-steppes, located mainly in its western and north-eastern fringes, respectively. Nowadays, the region is experiencing a rapid process of agriculture intensification and intensive woody culture expansion, particularly of trellis-vineyards and super-intensive olive groves (Ruiz-Pulpón 2013; Pérez et al. 2023), unsuitable as habitat for most steppe birds (Casas et al. 2020;Guerrero-Casado et al. 2022), or potential ecological traps for other species (Cabodevilla et al. 2021). Nevertheless, Castilla-La Mancha still harbours the bulk of several steppe bird populations in Spain, with, for example ca. 65% of the Spanish little bustard population (García de la Morena et al. 2018), ca. 62 and 15% of the Spanish pin-tailed (Pterocles alchata) and Black-bellied (P. orientalis) sandgrouse populations, respectively , and 17% of the Spanish great bustard (Otis tarda) population (the world's largest, Alonso et al. 2003). Indeed, Castilla-La Mancha can be considered the most important region for the little bustard in Spain and the EU (García de la Morena et al. 2018;Morales and Bretagnolle 2022a), and one of the most relevant for steppe-birds as a whole, both at the Spanish and European levels (Traba et al. 2007).

Distribution data
For Castilla-La Mancha, we obtained species presence/absence maps at 10 × 10 km UTM cells based on the Spanish atlas of breeding birds (Martí and Del Moral 2003) Suárez et al. 2006). We updated and refined the maps obtained from these sources with an exhaustive review of published . This compilation over eight years allowed establishing 10 × 10 km cells where the different species were systematically present or absent with a reasonable level of certainty ( Fig. 1 and Supplementary Figure S1). The publications and reports reviewed are presented in Supplementary Annex 1.
Among the 31 species reviewed in García de la Morena (2013) for Castilla-La Mancha, we selected 9 species to test the little bustard's suitability as umbrella species: great bustard, pin-tailed and black-bellied sandgrouse, stone curlew, red-legged partridge (Alectoris rufa), Montagu's harrier, skylark, calandra lark (Melanocorypha calandra), and blackeared wheatear (Oenanthe hispanica). In order to test the umbrella effect over a wide array of ecological niches, this list comprises species of different landscape preferences (cereal pseudo-steppes, shrub-steppes and pasturelands), level of habitat-specialisation, diet types (herbivorous, seed-eating, insect-eating, omnivorous) and size (large non-passerines vs passerines). This information is summarized in Table 1, which also includes the number of the remaining 21 species not analysed here to provide a more complete idea of the potential umbrella effect of the little bustard for the steppe bird community.

Environmental variables
In order to predict specie's habitat favourability in Castilla-La Mancha at 10 × 10 km UTM cells, we collected data at that scale on the following environmental components (Table 2), relevant for the prediction of steppe bird distribution at large spatial scale (e.g., Suárez-Seoane et al. 2002Benítez-López et al. 2014). (a) Climate: we obtained raster data from WorldClim Version2 (http:// www. world clim. org/), available for the 1970-2000 period (Fick and Hijmans 2017), with a 30 s spatial resolution (which basically equates to 1 km 2 resolution, see https:// www. ngdc. noaa. gov/ mgg/ topo/ report/ s6/ s6A. html) on average annual precipitation, mean annual temperature and mean annual solar radiation. To match this resolution with our 10 × 10 km UTM cell grid, we averaged 30'-resolution values for each 10 × 10 km UTM cell (see Delgado et al. 2011 for a similar approach). In addition, we calculated average thermal range from annual minimum and maximum temperature data. Despite the temporal mismatch between the climate data available and the bird distribution data reviewed (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013), our approach is adequate because (i) the 30 yearslong period of climate data used guarantees robust mean estimates for all variables and (ii) any potential deviation from those means is the same for all species being compared (see 'favourability models' below). (b) Topography: we calculated a digital elevation model (DEM) for Castilla-La Mancha by merging five raster layers obtained from the Spanish National Plan of Aerial Ortho-photography (http:// pnoa. ign. es/). Each layer corresponds to a DEM for each of the five provinces of the region at a 200 m resolution. We used the DEM to calculate the corresponding slope values. Again, we averaged 200 m-resolution values per 10 × 10 km UTM cell. (c) Land use: to better adjust land-use data to the time period covered by bird distribution data, we used CORINE Land Cover 2006 (https:// www. eea. europa. eu/ publi catio ns/ COR0-landc over), which is the CORINE version most closely matching the period of bird data compilation. We obtained raster data on the surface percentages of each land-use type in each region at a 1 km resolution, which we averaged for each 10 × 10 km UTM cell. The land-use types considered are presented in Table 2. (d) Vegetation productivity: finally, we obtained normalised-difference vegetation index (NDVI) values for each cell as a measure of primary production levels. Data were provided Table 1 Landscape preferences, habitat-specialisation (generalist vs specialist), and diet of the studied steppe birds Based on Díaz et al. (1996) and Tellería et al. (1999). The column "Overlap with other species" provides the number of species among those 21 (out of the total 31 whose distribution was reviewed by

Favourability models
We fitted a binomial logistic regression model on little bustard presence/absence data at 10 × 10 km cells using a forward step-wise procedure. We selected logistic regression and a stepwise approach in order to know which variables explained the broad-scale distribution of the species (i.e., those first entering the model) and which of them are related to finerscale distribution patterns (i.e., those entering last) (Márquez et al. 2011). In addition, the stepwise approach has been described as one of the best methods to fit spatial distribution models combining different sets of explanatory variables (Romero et al. 2015), as done here. To avoid collinearity in models, we examined pair-wise correlation of explanatory variables based on Pearson's correlation coefficient and eliminated one from each pair of variables with r > 0.7. DEM and NDVI were inter-correlated, as well as mean-precipitation and temperature. Thus, we deleted DEM from the analyses in order to keep a measure of primary productivity as the NDVI; we also deleted average temperature, since we already had a non-correlated thermal variable such as thermal range. In addition, we used the variance inflation factor (VIF) to assess collinearity in models, which was non-significant.
We employed a random sample with 70% of the data for model parametrization and left the remaining 30% for validation. Validation was carried out by the assessment of model sensitivity (proportion of presences correctly classified) and specificity (proportion of absences correctly classified), as well as the correct classification rate (CCR, total proportion of cases correctly classified). We used the specie's prevalence (total proportion of presences in the 10 × 10 km grid) as threshold value for the assessment of these model performance parameters. Nagelkerke's R 2 was used as a measure of model explained deviance (Hosmer and Lemeshow 2000). We then applied the favourability function (Real et al. 2006) to presence/absence probability values calculated with logistic regression to obtain favourable areas for the species in Castilla-La Mancha. We selected this function because, unlike probability values calculated with logistic regression, favourability is not affected by the species prevalence (ratio of presences/absences), and thus unbalanced samples, and reflects only environmentally suitable conditions (Acevedo and Real 2012). Therefore, favourability allows direct comparison between species, which is of great use in conservation management (Acevedo and Real 2012;Estrada et al. 2016), and its use is encouraged in predictive modelling like the one presented here (Real et al. 2006). Favourability can be calculated from logistic regression probabilities as follows (Real et al. 2006): where P is the probability value in each 10 × 10 km UTM cell, n 1 is the total number of presences and n 0 is the total number of absences. Favourability values vary from 0 to 1. Thus, we calculated a favourability value for the species in each 10 × 10 km UTM cell.
We used the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) (Fielding and Bell 1997;Manel et al. 2001;Lobo et al. 2008) to assess favourability 1 3 model robustness. This index is particularly useful since it is independent from any favourability threshold value (Manel et al. 2001). AUC values of 0.5-0.7 indicate poor model predictive ability, values of 0.7-0.9 indicate good model predictive ability and values > 0.9 correspond to very good predictive capacity (Swets 1988). Besides the little bustard, we used the same procedure to obtain favourability models for the co-occurring species in the study region.
In order to evaluate the little bustard's suitability as umbrella for co-occurring species, we employed the variables selected by the forward step-wise procedure in the little bustard favourability model to predict the favourability of the 9 selected species in Castilla-La Mancha at the 10 × 10 km scale, following the analytical approach described above. We assessed the fit and robustness of these models based on their explained deviance (Nagelkerke's R 2 ) and AUC values. Finally, in order to assess differences between little bustard and co-occurring species in the use of the environmental factors measured by our explanatory variables, we compared the values of explanatory variables included in the little bustard model in favourable cells (F > 0.5) with those of favourable cells for each co-occurring species. Since this analysis involved multiple probability tests, we applied Bonferroni's correction to reduce the probability of Type I error. All statistical analyses were carried out in SPSS Statistics 23.

Results
The environmental variables selected by the little bustard model in Castilla-La Mancha were slope, annual mean thermal range, annual precipitation, percentage of rain-fed cereal cultures in the landscape and annual mean solar radiation (Table 3). Slope, annual precipitation, and annual mean solar radiation were negatively associated to favourability, while percentage of dry cereal cultures and thermal range were positively associated. The model was robust, presenting very good performance (AUC = 0.905) and high explanatory capacity (Nagekerke's R 2 = 0.60, see Table 3 for remaining performance indices). In other words, favourable areas for the little bustard are, in the first place, topographically flat and subjected to significant seasonal temperature contrast (an indicator of climate continentality), but also relatively dry, and dominated by dry cereal farmland, though not excessively insolated.
The spatial variation of little bustard habitat favourability in Castilla-La Mancha is shown in Fig. 2A, where a core favourable area running from north-west to south and south-east can be appreciated. Fifty two percent of Castilla-La Mancha's 10 × 10 UTM cells were favourable for the little bustard (F > 0.5), and 38% were very favourable for the species (i.e., they were in the highest quartile of favourability: F > 0.75).
Favourability models for the remaining species are presented in Supplementary  Table S1). Models fitted with the variables explaining little bustard favourability and applied on the nine co-occurring species from Castilla-La Mancha yielded varying performance results (See Table 4).
Models for pin-tailed sandgrouse, great bustard, Montagu's harrier and calandra lark showed good or very good performance (AUC = 0.82-0.95) and high explanatory power (Nagekerke's R 2 = 0.38-070). According to these models, most favourable cells for the little bustard were also favourable (F > 0.5) for these species (94% for Montagu's harrier, 81% for Great bustard, 75% for pin-tailed sandgrouse and 73% for calandra lark). The pattern was similar for cells of high favourability for the little bustard (F > 0.75), since 76% of them were also very favourable for pin-tailed sandgrouse and calandra lark, 65% for Montagu's harrier and 61% for great bustard.
However, models for stone curlew, skylark, black-bellied sandgrouse and black-eared wheatear performed poorly (AUC = 0.53-0.57) and explained a very small proportion of deviance (Nagekerke's R 2 = 0.004-0.02). The red-legged partridge model occupied a somewhat intermediate position, with acceptable performance and discrete explanatory power (AUC = 0.72, Nagekerke's R 2 = 0.06). As shown in Fig. 2, the habitat favourability of these species across Castilla-La Mancha is homogeneously intermediate, except for the redlegged partridge, whose favourable values tend to concentrate in the north of the region (Fig. 2H). Consistently, while the degree of coincidence between favourable cells (F > 0.5) for the little bustard with favourable cells yielded by these models was still high (89% for red-legged partridge, 87% for black-bellied sandgrouse, 71% for stone curlew, 68% for skylark and 53% for black-eared wheatear), it was almost non-existent when cells of high favourability (F > 0.75) were considered (only 16% for red-legged partridge and 0% for stone curlew, skylark, black-bellied sandgrouse and black-eared wheatear). The difference in variable use by the little bustard and the species co-occurring with it in Castilla-La Mancha is presented in Table 5, which shows that, overall, Montagu's harrier, great bustard and both sandgrouse species tend be closer to little bustard in variable use, while the remaining five species tend to significantly differ on this respect.

Discussion
After applying the results of little bustard favourability models to nine co-occurring species, two different groups can be distinguished among them. On one hand, there are species, namely pin-tailed sandgrouse, great bustard, Montagu's harrier and calandra lark, whose favourability is closely predicted by variables associated to little bustard favourability. At large scale (i.e., the one considered in this study) these species are typically 1 3 associated to these extensive cereal farmland landscapes, although at smaller scales they can be linked to, for example, particular landscape features like fallows, meadows or field edges (e.g., Delgado and Moreira 2000;Morales et al. 2013, Tarjuelo et al. 2014 or even vegetation structure characteristics (Traba et al. 2015). On the other hand, there are species like stone curlew, skylark, black-bellied sandgrouse, black-eared wheatear and red-legged partridge, whose favourability is weakly predicted by little bustard favourability models, yielding poor model performance results. These species tend to show intermediate favourability values homogeneously across the study region, probably because they can thrive in a larger variety of landscapes (see Tellería et al. 1999 andDíaz et al. 1996 for general habitat preferences of these species) and thus, at the large scale considered, they behave as generalists. Moreover, some land-use changes associated to agriculture intensification and occurring over large extents (e.g., expansion of super-intensive olive groves or trellis vineyards; Cabodevilla et al. 2021;Guerrero-Casado et al. 2022), may have reduced habitat favourability for little bustard in certain areas, but not necessarily for some others from this group like the red-legged partridge or the stone curlew which seem more tolerant to these changes (Moreno-Mateos et al. 2011;Cabodevilla et al. 2021).
Although the umbrella species concept is not free from criticism (Seddon and Leech 2008;Wang et al. 2021), it has never ceased to be applied in different conservation contexts (Fleishman et al. 2001;Lambeck 2003;Breckheimer et al. 2014). Umbrella species are those whose requirements cover those of other organisms in the community (Roberge and Angelstam 2004;Caro 2010). These requirements can be evaluated at different spatial scales, from fine habitat structure to large geographical scale (Sutter et al. 2002;Roberge and Angelstam 2004;Breckheimer et al. 2014). As we discuss below for the case of the little bustard and coexisting steppe and grassland birds associated to cereal farmland, the umbrella effect of a given species (single-species approach, Roberge and Anglestam 2004) at large geographical scale can be tested using favourability models.
Previous works have already shown that little bustards are associated to flat and cereal farmland-dominated landscapes at large geographical scales (see review in Traba et al. 2022), which in Castilla-La Mancha and the Iberian Peninsula in general are also characterized by dry continental climates with high insolation (Suárez-Seoane et al. 2002Delgado et al. 2011). On the other hand, the little bustard is a permanent or semi-permanent grassland specialist associated to agricultural habitats such as fallows, grazed pastures, mown meadows and dry alfalfa fields (Brotons et al. 2004;Silva et al. 2014; Morales et al. Table 4 Performance of habitat favourability models for nine species of steppe birds using explanatory variables selected in the little bustard favourability model (see Table 3 Bretagnolle et al. 2018;Traba and Morales 2019). These habitats are found mainly in open cereal plains like those comprised in our study region, which further contributes to explain the favourability results obtained. Interestingly, a recent study based on point counts has found that little bustard habitat quality at this finer scale decreased with urban encroachment (Arroyo et al. 2022), which supports this species dependency on undisturbed cereal steppe landscapes. The species adequately predicted by the little bustard's favourability models are also linked to those landscapes, consistent with the high performance of their favourability models when variables defining little bustard favourability are applied (Table 4). Moreover, at finer spatial scale, some of these species are known to interact closely with the little bustard, which further supports this consistency. For example, pintailed sandgrouse and little bustards tend to form mixed-species winter flocks where mutualistic and/or commensalist interactions occur (Martín et al. 2010;García de la Morena 2015). Also, great and little bustard segregate within the mosaic farmland landscape where they use specific habitats differently (Tarjuelo et al. 2014(Tarjuelo et al. , 2017, and Montagu's harrier partly relies on similar insect prey as little bustards (i.e., orthopterans, Bretagnolle et al. 2022a, b) or may prey on little bustard eggs and chicks (Corbacho et al. 2005). Finally, the range of variation of the different environmental variables selected in the little bustard model in cells with favourability greater than 0.5 in relation to those of co-occurring specie's cells with favourability greater than 0.5 was largely coincident (Table 5), with very few significant differences. Contrarily, the coincidence with landscape-scale generalist species is much weaker: most of them show significant differences in most variables (Table 5). Therefore, our results are consistent with the niche-based concept of umbrella species proposed by authors like Lambeck (2003) and Roberge and Angelstam (2004), and support the use of the little bustard as an umbrella species for the conservation of a relevant number of poorly known and declining species strongly associated to extensive cereal farmland landscapes (Morales and Traba 2016), but not for all the steppe birds species considered at regional scale. In this line of evidence, the high level of spatial overlap between little bustard favourable 10 × 10 km cells and the favourable cells of these co-occurring species allows defining areas of general favourability for these cereal farmland specialists. Such areas could be used in large-scale conservation planning to concentrate land-use management efforts aimed at the conservation of these declining species, which are generally more costly to monitor than the little bustard, both in terms of population size and spatial distribution. Similarly, the large-scale response of these coexisting farmland specialists to conservation measures is more difficult to predict than that of the little bustard due to the lower level of existing knowledge on their basic ecology . Overall, the use of the relatively widely distributed (in this case, present in 442 10 × 10 km UTM cells), easy-to-monitor and well-studied little bustard as umbrella species for rapidly declining cereal farmland birds seems to be a cost-effective conservation strategy. It is worth mentioning in this context that little bustard surveys and monitoring usually rely only on male observations, since females are much less detectable (e.g., García de la Morena et al. 2018), but female local abundance tends to be positively correlated with male local abundance (Devoucoux et al. 2019;Arroyo et al. 2022), and thus habitat favourability models based on male data can be assumed correct for the species as whole, at least at the scale considered here.
The little bustard favourable area in Castilla-La Mancha identified here is highly consistent with the steppe-bird 10 × 10 km hot-spot cells defined in this region by Traba et al. (2007): 90.4% hot-spot cells in Castilla-La Mancha fall within the little bustard favourable area. In spite of that, this area is poorly protected, with very little surface covered by protected areas (ca. 800 km 2 with favourability greater than 0.5 included in the Natura 2000 1 3 network, i.e., 1.71%). This illustrates that our approach can also be used to assess the level of protection of species assemblages across large geographical regions.
In this study we have used the favourability function to show that the little bustard can be considered as a suitable umbrella species for a set of different declining bird species associated to cereal farmland at large spatial scale. Nevertheless, the fact that several other steppe bird species were poorly covered by the little bustard favourability model highlights the limitations of the single-species approach in applying the umbrella species concept, as pointed out by previous authors (Roberge and Anglestam 2004), and the convenience to test the performance as umbrella of species using distant ranges of the niche space (here modelled as habitat favourability) in order to find a suit of species that may offer maximum coverage to other organisms at the adequate scale (Tellería 2012). In the case of Iberian steppe birds, species strictly linked to flat and open landscapes, but yielding intermediate favourability values throughout the study region, such as the black-bellied sandgrouse or the stone curlew (see Fig. 2) may be good candidates. Lastly, favourability models can potentially be applied to higher spatial resolutions to account for habitat factors that can be directly managed to improve habitat quality. We therefore encourage their application within the umbrella species framework at more detailed scales in order to identify landscape features that may contribute to increase habitat favourability for an as-large-as-possible species assemblage, and thus as a tool to manage landscapes and habitats harbouring large numbers of species of conservation concern, as it is the case of cereal farmland.