Introduction
Agricultural expansion drives biodiversity loss in the tropics (Oakley & Bicknell, Reference Oakley and Bicknell2022). Protected areas are crucial for conserving biodiversity (Cazalis et al., Reference Cazalis, Princé, Mihoub, Kelly, Butchart and Rodrigues2020) but they have limited geographical coverage (Baldwin & Beazley, Reference Baldwin and Beazley2019). In the Afro-Asian tropics, protected areas are overrepresented at higher elevations (Elsen et al., Reference Elsen, Monahan and Merenlender2018), leaving low-elevation forests vulnerable to agricultural expansion and agroforestry (Namkhan et al., Reference Namkhan, Gale, Savini and Tantipisanuh2021). As a consequence, many forest-specialist species typically found in tropical low-elevation forests are more vulnerable to habitat loss than those at higher elevations (Mills et al., Reference Mills, Socolar, Edwards, Parra, Martínez-Revelo and Ochoa Quintero2023). This indicates the need to assess the potential of landscapes beyond protected areas (agroforestry plantations and privately- or government-owned forest remnants) for conserving biodiversity.
Biodiversity responses to different forms of land use have often been examined through taxonomic diversity, but this has limitations. For example, shifts in species prevalence and community composition, such as specialist species being replaced by generalist species, may not be reflected in taxonomic diversity metrics (Hillebrand et al., Reference Hillebrand, Blasius, Borer, Chase, Downing and Eriksson2018). This has spurred interest in other metrics such as functional and phylogenetic diversity (Cadotte et al., Reference Cadotte, Albert and Walker2013). Functional diversity reflects variation in functional traits, such as the loss of large-bodied animals (Coetzee & Chown, Reference Coetzee and Chown2016), and phylogenetic diversity reflects the persistence of evolutionarily distinct species (Turley & Brudvig, Reference Turley and Brudvig2016). Moreover, functional and phylogenetic overdispersion or clustering shed light on the putative roles of competition and environmental filtering in structuring communities (Webb et al., Reference Webb, Ackerly, McPeek and Donoghue2002; Cadotte et al., Reference Cadotte, Albert and Walker2013).
Although studies are beginning to examine land-use change impacts on taxonomic, functional and phylogenetic diversity, there is inconsistency in the strength and direction of impacts across studies. Land-use change has been shown to increase (Sreekar et al., Reference Sreekar, Si, Sam, Liu, Dayananda and Goodale2021), reduce (Bregman et al., Reference Bregman, Lees, MacGregor, Darski, de Moura and Aleixo2016) or not affect (Rurangwa et al., Reference Rurangwa, Aguirre-Gutiérrez, Matthews, Niyigaba, Wayman, Tobias and Whittaker2021) functional and phylogenetic diversity. This highlights the need for studies comparing a broader range of land-use types across different geographical and ecological contexts. Some studies have compared forests with land uses such as farms and urban environments (García-Navas & Thuiller, Reference García-Navas and Thuiller2020; Sreekar et al., Reference Sreekar, Si, Sam, Liu, Dayananda and Goodale2021), but few studies have compared forests with tree-based agroforestry systems (Jayathilake et al., Reference Jayathilake, Warren-Thomas, Nelson, Dolman, Bumrungsri and Juthong2021). This is pertinent in Asia, where agroforestry plantations threaten tropical forests (Namkhan et al., Reference Namkhan, Gale, Savini and Tantipisanuh2021).
Functional and phylogenetic diversity metrics are influenced by functional traits and evolutionary relationships between species, respectively. However, these community-level metrics can obscure the responses of individual species to land-use change, particularly in the tropics, where many species have narrow ranges, low densities and specialized niches (Stratford & Robinson, Reference Stratford and Robinson2005; Orme et al., Reference Orme, Davies, Olson, Thomas, Ding and Rasmussen2006). Species responses to land-use change are also shaped by functional traits (de Bello et al., Reference de Bello, Carmona, Dias, Götzenberger, Moretti and Berg2021), and thus examining which traits predict species-level declines or gains can explain the mechanisms driving community-level patterns. However, studies rarely investigate biodiversity responses to land-use change simultaneously at the community, species and trait levels (Leveau et al., Reference Leveau, Isla and Bellocq2020), leaving a gap in our understanding of how land-use change reshapes ecological communities.
Birds are useful for examining the impacts of land-use change on biodiversity because they are good indicators of habitat quality (Canterbury et al., Reference Canterbury, Martin, Petit, Petit and Bradford2000), they perform diverse ecological functions (Sekercioglu, Reference Sekercioglu2006), and they are often widespread and occur across habitats, facilitating studies of the impacts of environmental change on biodiversity. Their morphological traits are useful for assessing functional diversity as they are strongly related to their niches and ecological function (Pigot et al., Reference Pigot, Sheard, Miller, Bregman, Freeman and Roll2020). Additionally, by 2100, land-use and climate change are projected to drive contractions of the ranges of > 900 bird species by 50%, and species from the tropics are considered especially vulnerable to land-use change (Jetz et al., Reference Jetz, Wilcove and Dobson2007). In India, citizen-science-driven assessments of Indian bird species have demonstrated long-term declines of at least 60% of bird species (SoIB, 2023).
Here, we compare multiple dimensions of bird diversity across land-use types in the Western Ghats Biodiversity Hotspot. We focus on protected areas, which are restricted to higher elevations, and low-elevation reserved forests (state-owned forests with permission for grazing and fuelwood collection for local communities), degraded private forests (privately-owned forests that are periodically clear-felled), and agroforestry monocultures (cashew, mango, rubber plantations). Specifically, we examined how land-use type and elevation influence (1) taxonomic, functional and phylogenetic diversity and (2) species occurrences, and (3) the role of functional traits in influencing species responses. We hypothesized that bird communities in human-modified land-use types would have lower species richness, with relatively well-protected forests having higher taxonomic diversity, overdispersed functional and phylogenetic diversity, and higher occurrence probabilities of specialist and conservation-concern species than degraded forests and agroforestry lands. Additionally, as protected areas are located at higher elevations, comparing protected areas with other low-elevation land-use types would allow us to examine the role of elevation in structuring bird communities. The Western Ghats, along with Sri Lanka, comprise a global biodiversity hotspot because of the high endemism and extensive threats (Myers et al., Reference Myers, Mittermeier, Mittermeier, da Fonseca and Kent2000). Over 35% of forests in the Western Ghats have been lost in the last 100 years, and forest conversion to agroforestry plantations is one of the primary drivers of this loss (Reddy et al., Reference Reddy, Jha and Dadhwal2016). Research on land-use change and birds has focused on taxonomic diversity and has been concentrated in the southern and central regions of the Western Ghats (Shankar Raman & Sukumar, Reference Shankar Raman and Sukumar2002; Karanth et al., Reference Karanth, Sankararaman, Dalvi, Srivathsa, Parameshwaran and Sharma2016). The northern Western Ghats are an ecologically important and distinctive region but remain understudied (but see Munje & Kumar, Reference Munje and Kumar2022). Forests here are being converted to monoculture plantations, and those that are not converted are periodically clear-felled for fuelwood (Kulkarni & Mehta, Reference Kulkarni and Mehta2013; Rege & Lee, Reference Rege and Lee2022).
Study area
We conducted the study during January–May 2022 in the northern Western Ghats of Maharashtra state, India (Supplementary Table 1, Supplementary Fig. 1). Whereas forests in the southern Western Ghats have been mostly converted to tea, coffee and rubber plantations (Anand et al., Reference Anand, Krishnaswamy, Kumar and Bali2010), privately-owned forests in the northern Western Ghats are being converted to mango, cashew and rubber plantations (Rege & Lee, Reference Rege and Lee2022). The northern Western Ghats have dry periods of 7–8 months, making the wet forest species more vulnerable to land-use change. The total annual precipitation and mean annual temperature range are 2,150–7,450 mm and 11–38 °C, respectively. Terrain is undulating, and the elevation of our survey sites is 12–1,110 m. The area originally harboured evergreen and semi-evergreen forests, but they have been degraded or lost as a result of repeated clearing or conversion to agroforestry plantations (Biswas et al., Reference Biswas, Sadekar, Biniwale, Taneja, Osuri and Page2024).
We surveyed six types of habitats: protected areas, reserved forests, private forests, and cashew, rubber and mango monoculture plantations (Supplementary Table 1, Supplementary Fig. 1). Protected areas are owned and protected by the state and are free from tree-felling and other forms of extraction. Reserved forests are also state-owned and protected from tree-felling, but some resource extraction (e.g. grazing, fuelwood collection) is permitted. Private forests are typically clear-felled in a 5–10 year cycle for fuelwood. All the study sites in the reserved forests, private forests and agroforestry plantations are at lower elevations, and the protected area sites are at higher elevations (Supplementary Table 1). Our design allowed us to assess bird diversity and species responses to land use by comparing reserved forests with private forests and plantations, and bird diversity and species responses to elevation by comparing low-elevation reserved forests with high-elevation protected areas.
Methods
We quantified taxonomic, functional and phylogenetic diversity of birds across six land-use types and used a hierarchical modelling of species communities framework, incorporating species, trait and phylogenetic data, to assess species- and community-level responses of birds to land-use change (Fig. 1).
Workflow for the hierarchical modelling of species communities analysis. Five input data types were used to model bird community responses to land-use change across six land-use types in the Western Ghats, India. Solid arrows indicate fixed effects; dashed arrows indicate hierarchical random effects accounting for spatial autocorrelation and phylogenetic dependence among species. Key results summarize the three main outputs of the analysis.
Bird surveys
We surveyed birds, using line transects, during January–May 2022, which coincides with the breeding season of resident birds and is a relatively dry period (Sreekar et al., Reference Sreekar, Si, Sam, Liu, Dayananda and Goodale2021). We surveyed 184 sites (one transect per site, each surveyed once) across all land-use categories, with a total survey effort of 137.5 km (Supplementary Table 1). Transect lengths varied (0.17–2.19 km), with length dependent on the area available (Supplementary Table 1). The total effort in each land-use type was similar (Supplementary Table 1). Transects were spaced at least 500 m apart. We walked them during 7.00–12.00 and 14.50–18.20, at an average pace of 0.94 ± SD 0.42 km/h. The relative proportion of transects walked in the morning was similar (55–65%) across land uses, making any potential biases relatively consistent. We identified bird species from sightings and calls and recorded the number of individuals, when seen. We followed Clements (Reference Clements2007) for taxonomy. For the analysis, we only considered terrestrial, forest bird species seen within 50 m either side of the transect. Any wetland species observed were not included in the analysis.
Functional and phylogenetic data
We collated functional trait data and phylogenetic data following established methods. To estimate overall functional diversity in each land-use type, we focused on combining morphological and ecological traits following Chapman et al. (Reference Chapman, Tobias, Edwards and Davies2018). Morphological traits were trophic (beak length (culmen), width and depth) and locomotory traits (tarsus length, wing length, Kipp’s distance (the distance between the tip of the first secondary feather and the tip of the longest primary feather) and tail length), and ecological traits were diet and foraging strata. Morphological traits were sourced from AVONET (Tobias et al., Reference Tobias, Sheard, Pigot, Devenish, Yang and Sayol2022), and ecological traits from EltonTraits 1.0 (Wilman et al., Reference Wilman, Belmaker, Simpson, Rosa, Rivadeneira and Jetz2014). Diet categories were per cent of invertebrates, vertebrates, fish, carrion, fruits, nectar, seeds and other plant materials in the diet. Foraging strata categories were the prevalence of foraging on the ground, in the understorey, at mid tree height, in the tree canopy and aerial (Wilman et al., Reference Wilman, Belmaker, Simpson, Rosa, Rivadeneira and Jetz2014). To reduce the dimensionality of the data, we performed a principal component analysis for each of the four trait categories (locomotory, trophic, diet and foraging strata) and extracted the first principal component from each, which we then scaled to construct species × trait matrices for each land-use category. To create a functional dendrogram of traits, we used the packages vegan (Oksanen et al., Reference Oksanen, Blanchet, Friendly, Kindt, Legendre and McGlinn2019) and ade4 (Dray & Dufour, Reference Dray and Dufour2007) in R 4.2.2.X (R Core Team, 2022).
To determine the effect of land-use change on the phylogenetic diversity of birds, we obtained 100 randomly generated phylogenetic trees from Jetz et al. (Reference Jetz, Thomas, Joy, Hartmann and Mooers2012), following the Ericson backbone (a set of 10,000 trees with 9,993 operational taxonomic units each). Using these trees, we obtained a maximum credibility tree to estimate phylogenetic diversity.
Taxonomic, functional and phylogenetic diversity
We used survey coverage (the probability of not finding a new species with the next surveyed individual) to determine sampling completeness (Roswell et al., Reference Roswell, Dushoff and Winfree2021). In our case, the survey coverage was high (> 90%). To compare species richness across land uses, we used the Hill–Shannon Diversity measure, standardized to 92% sampling coverage, which was the lowest coverage value recorded in rubber plantations, following Roswell et al. (Reference Roswell, Dushoff and Winfree2021). Hill–Shannon Diversity is e H’ , where H’ is the conventional Shannon index. This diversity measure places less emphasis on rare species than species richness, is useful for diversity comparisons and is robust to standardization (Roswell et al., Reference Roswell, Dushoff and Winfree2021). For sample-coverage-based rarefaction, we used the package iNEXT (Hsieh et al., Reference Hsieh, Ma and Chao2016) in R. We bootstrapped the data 50 times to estimate 95% confidence intervals, and inferred significant differences if the 95% CI did not overlap, following Cumming et al. (Reference Cumming, Fidler and Vaux2007).
We used mean pairwise distance for the estimation of both phylogenetic and functional diversity, as it allows direct comparison between the two metrics by applying the same tree-based calculation to both the phylogenetic and functional dendrograms. Mean pairwise distance is the average pairwise (phylogenetic/functional) distance between species pairs in a community (for additional details, see Supplementary Material 1). With the conversion of forests to agroforestry plantations, we expected the selective exclusion of species with particular traits and unique evolutionary lineages to result in a decline in functional and phylogenetic diversity. We expected the general trend in standardized effect size values to show overdispersion (as a result of competition) in forests and underdispersion (as a result of environmental filtering, which is the process by which altered or harsh environmental conditions selectively exclude species, resulting in communities of more similar species than expected by chance) in plantations.
Species and trait responses to land-use change
To determine how bird species respond to different land uses and whether these responses are influenced by functional traits or evolutionary relationships, we used the hierarchical modelling of species communities framework (Ovaskainen & Abrego, Reference Ovaskainen and Abrego2020). This approach allowed us to identify bird species whose occurrence declined in plantations and degraded forests and the traits associated with these declines. Additionally, it enabled us to assess whether closely related species exhibited similar responses to different land uses. Understanding the relationships between functional traits and land uses allows us to generalize our findings to other regions. We excluded species detected in < 5% of transects as they provided little information on community assembly and posed challenges for Markov chain Monte Carlo convergence (Ovaskainen & Abrego, Reference Ovaskainen and Abrego2020). Thus, we modelled the occurrence of 59 bird species as a function of land-use categories, the elevation of the transect and transect length. We included transect length as a fixed effect in the model as it is variable. We included the location of each transect as a random effect, to account for spatial autocorrelation. As habitat modification, especially conversion to agroforestry plantations, can alter vegetation structure and resource availability for birds, we expected traits linked to foraging strata, locomotion, trophic level and diet to be impacted by land-use change (for additional information on traits, see Supplementary Material 2). To determine the effect of land-use change on bird phylogenetic diversity, we downloaded 10,000 trees from the posterior distribution of the global bird phylogeny (Jetz et al., Reference Jetz, Thomas, Joy, Hartmann and Mooers2012), using the Ericson backbone topology and restricting the tree to our 59 focal species. From these trees, we randomly selected 100 trees and generated a maximum clade credibility tree as our representative phylogeny.
We used the package Hmsc in R to fit the model with default prior distributions (Tikhonov et al., Reference Tikhonov, Opedal, Abrego, Lehikoinen, de Jonge, Oksanen and Ovaskainen2020). We sampled a posterior distribution with three Markov chain Monte Carlo chains, each run for 375,000 iterations. We removed the first 125,000 iterations as burn-in. The chains were thinned by 1,000 to yield 750 posterior samples for three chains. We assessed model convergence through potential scale reduction factor and effective sample size for all the beta and gamma coefficients. We used Tjur’s R 2 to assess model fit (Tjur, Reference Tjur2009) and variance partitioning to assess the proportion of variation explained by the three types of fixed effects (land use, elevation and trail length), apart from the random effect. We evaluated species’ responses to land-use change by examining beta coefficients. To determine whether traits explain species responses to different land-use types, we examined gamma coefficients, which indicate the strength of effects of land-use types on traits. We estimated the variation in species responses to the predictor variables, using the rho estimate to determine the phylogenetic signal in species responses. We considered the influence of coefficients significant if the 95% credible intervals for beta, gamma and rho parameters did not overlap zero.
For the 59 focal species, we extracted the following attributes: habitat affiliation in terms of known preference for evergreen versus deciduous forest, based on Ali & Ripley (Reference Ali and Ripley1999), and national population trend (stable/declining/uncertain), as assessed by The State of India’s Birds report (SoIB, 2023). We compared the beta coefficients from the hierarchical modelling of species communities analysis across the six land-use categories for each species based on their habitat preferences and long-term population trends, using the 95% bootstrapped confidence intervals (Cumming et al., Reference Cumming, Fidler and Vaux2007).
Results
Taxonomic, functional and phylogenetic diversity across land-use types
We detected 3,797 individuals of 138 bird species across land-use types (Supplementary Table 1). This included three species categorized on the IUCN Red List (IUCN, 2025) as Vulnerable (Buceros bicornis, Ocyceros griseus and Columba elphinstonii) and two as Near Threatened (Anthracoceros coronatus and Microtarsus priocephalus). The reserved forests, our reference sites for low elevations, had 1.4–1.8× higher Hill–Shannon diversity than private forests, the three kinds of agroforestry plantations, and the high-elevation protected areas (Fig. 2a).
(a) Taxonomic, (b) functional and (c) phylogenetic diversity of birds across land-use categories. The forested habitats tend towards overdispersion, whereas the agroforestry plantations tend towards functional and phylogenetic clustering. MPD, mean pairwise distance.
The per cent of variance explained by the first principal components for trophic, locomotory, diet and strata traits were 98.4%, 91.1%, 62.3% and 54.1%, respectively (for details on trait loadings on the first principal component, see Supplementary Table 2). The functional and phylogenetic diversity of birds tended to have positive standardized effect size values for less disturbed forested habitats (protected areas and reserved forests), indicating the tendency of the communities to be overdispersed; i.e. the observed functional and phylogenetic measures were greater than expected by random (Fig. 2b,c).
Species responses to land-use change
The average explained variation in occurrence for all species combined was 20%. The hierarchical modelling of species communities analysis revealed that the total variation explained was > 30% for several range-restricted and endemic species found in peninsular India (e.g. Acritillas indica, Ocyceros griseus; Fig. 3). Among the fixed effects, land-use type explained > 50% of the variation in species occurrence, followed by elevation (20.5%; Supplementary Fig. 2). The spatial random effect played a relatively minor role in explaining variation in species occurrence compared to land-use type and elevation (Supplementary Fig. 2). The spatial scale parameter α was estimated to be zero, indicating a lack of evidence of a spatial signal in the data.
Species-specific responses of the 59 bird species to different land-use types. Dark and light grey indicate negative and positive associations, respectively, with posterior probability ≥ 0.95. The probability of occurrence of many species was lower in modified landscapes than in the reserved forests.
The probability of occurrence of 38 of the 59 species (65%) was lower in at least one of the modified habitats (private forests or agroforestry plantations) than in the reference reserved forest sites (Fig. 3). The coefficients of overall bird species responses to agroforestry plantations were lower than for forested habitats, as inferred from non-overlapping 95% bootstrapped CIs (Supplementary Fig. 3), indicating negative impacts of agroforestry plantations on birds. Compared to low-elevation reserved forests, the probability of occurrence of nine species (e.g. Ocyceros griseus, Dicrurus aeneus) was lower in all modified habitats (Fig. 3). Compared to reserved forests, the probability of occurrence of 16 species (e.g. Anthracoceros coronatus, Leptocoma minima, Spilornis cheela) was lower in all three types of agroforestry plantations. Some species (e.g. O. griseus, M. priocephalus, M. horsfieldii, L. minima) are range-restricted (found only in peninsular India) or are endemic to the Western Ghats. The occurrence of several open habitat species (e.g. Pycnonotus cafer, Cinnyris asiaticus) was higher in agroforestry plantations than in the reference reserved forest sites (Fig. 3). Probability of occurrence of only nine species (11%; e.g. Columba elphinstonii, Hypsipetes ganeesa) was greater in high-elevation protected areas than in the low-elevation reserved forests (Fig. 3).
Responses of evergreen forest specialists and species with declining populations to land-use types
The probability of occurrence of species that preferred evergreen forests was lower in the agroforestry plantations than in forested habitats, as inferred from non-overlapping bootstrapped 95% CI (Fig. 4a). There was no discernible pattern for species that preferred the open deciduous forests (Fig. 4b). The probability of occurrence of species showing declining trends as per the report on the State of India’s birds (2023) was lower in agroforestry plantations (Fig. 4c). The pattern was not evident for species having uncertain or stable population trends (Fig. 4d,e). Among agroforestry plantations, the probability of occurrence of birds with uncertain trends was lower in rubber plantations than the benchmark low-elevation reserved forests (Fig. 4d).
Untransformed mean beta coefficients and bootstrapped 95% CI for each land-use type, from hierarchical modelling of species communities, for bird species that prefer (a) evergreen and (b) deciduous forests, and (c) for species that have exhibited population declines over time, (d) for species with uncertain population trends, and (e) for species that have remained stable (as determined by the report State of India’s Birds, 2023). Note that the mean coefficient for species that prefer evergreen forests in modified landscapes (including private forests) is lower than in forested ecosystems, and populations that have exhibited long-term declines in India are negatively affected by habitat modification.
Traits and species response to land-use change
Functional traits explained 20% of the variation in bird species responses to the predictor variables. Frugivorous diet was negatively associated with plantations and degraded forests (Fig. 5), suggesting frugivorous birds are vulnerable to habitat degradation and conversion. The trait indicating the vertical foraging strata utilized by birds showed distinct associations. High strata values were associated with birds that forage on the ground or in lower strata (e.g. Galloperdix spadicea, Copsychus saularis) and low values with canopy-dwelling birds (e.g. hornbills). Stratum was negatively associated with the reference reserved forests, indicating a greater prevalence of canopy-dwelling birds in the less-disturbed reserved forests (Fig. 5). We did not detect a phylogenetic signal in species response to predictor variables (ρ = 0, 95% CI 0.0–0.4), suggesting that closely related species did not respond more similarly than unrelated species to predictors beyond what was expected from their functional traits.
Mean posterior gamma parameter values indicating how species traits influence responses to predictor variables. Cells are shaded based on the sign and strength of the trait–predictor relationship, with dark grey indicating a significant negative association and light grey a significant positive association (posterior probability ≥ 0.95). For example, a dark grey cell in the fruit–cashew plantation cell suggests that fruit-eating species respond negatively to cashew land use. Negative intercepts (dark grey) imply that species with high trait values have a lower baseline probability of occurrence across sites.
Discussion
We examined community-, trait- and species-level responses of birds to land-use change by comparing three forested ecosystems (varying in protection level and elevation) and three agroforestry plantations, using spatially replicated transect surveys in the Western Ghats biodiversity hotspot. We found that the probability of occurrence of several evergreen-forest-affiliated bird species and frugivores was lower in periodically disturbed forests and agroforestry plantations and the probability of occurrence of several deciduous-forest-affiliated species increased, reflecting a shift in composition as a result of habitat modification. However, this shift in composition was still insufficient to maintain diversity, as modified landscapes tended towards functional and phylogenetic clustering, and the two less disturbed forested ecosystems (protected areas and reserved forests) tended towards overdispersion, with species having diverse traits and lineages. The low-elevation reserved forests had a higher diversity of birds than the high-elevation protected areas. Collectively, our results highlight the value of securing remnant forest patches in lower elevations outside protected areas for lowland tropical avifauna, and suggest the need for collaborative conservation and restoration to secure these forests.
Previous studies have mostly focused on determining land-use change impacts on taxonomic, functional and phylogenetic diversity (Chapman et al., Reference Chapman, Tobias, Edwards and Davies2018; Penjor et al., Reference Penjor, Cushman, Kaszta, Sherub and Macdonald2022) and, on occasion, functional traits (Sreekar et al., Reference Sreekar, Si, Sam, Liu, Dayananda and Goodale2021). In our study, along with community metrics and functional traits, we also determined individual species responses, responses of forest specialists and species with large-scale population declines to land-use change. The value of integrating data at multiple levels of organization (community, traits and species) is evident when examining the impacts of rubber plantations. At the community level, rubber appears to harbour more bird species than other agroforestry plantations, but species detections and occurrence probabilities are far lower in rubber than in other plantations. Thus, not all agroforestry plantation types are similar, and rubber plantations may be more detrimental for birds in this region. This is corroborated by another study from the Western Ghats that recorded a lower density of birds across foraging and habitat guilds in rubber plantations than in coffee and areca (Karanth et al., Reference Karanth, Sankararaman, Dalvi, Srivathsa, Parameshwaran and Sharma2016). Compared to mango and cashew plantations, rubber cultivation has expanded relatively recently in the landscape, largely on lands that were previously forested (Munje & Kumar, Reference Munje and Kumar2022). Legacy effects from the previous land use may contribute to the relatively high functional and phylogenetic diversity in rubber plantations (García-Navas & Thuiller, Reference García-Navas and Thuiller2020). Investigating only community-level metrics may not have revealed the finer details of the impacts of rubber plantations on birds. We recommend that future research should focus on simultaneously examining species-, trait- and community-level responses to land-use change to ensure that negative impacts are not masked because of examination at a single level.
Contrary to some previous studies that have not consistently found lower functional and phylogenetic diversity of forest birds in agroforestry landscapes than in forests (Rurangwa et al., Reference Rurangwa, Aguirre-Gutiérrez, Matthews, Niyigaba, Wayman, Tobias and Whittaker2021; Sreekar et al., Reference Sreekar, Si, Sam, Liu, Dayananda and Goodale2021; Penjor et al., Reference Penjor, Cushman, Kaszta, Sherub and Macdonald2022), we found that modified landscapes had lower taxonomic diversity, and functionally and phylogenetically clustered communities. Conversion of forests to open farmland landscapes has resulted in the loss of forest birds and the addition of wetland birds (Rurangwa et al., Reference Rurangwa, Aguirre-Gutiérrez, Matthews, Niyigaba, Wayman, Tobias and Whittaker2021; Sreekar et al., Reference Sreekar, Si, Sam, Liu, Dayananda and Goodale2021). The addition of wetland birds may compensate for the reduction in functional diversity from the loss of forest birds. This likely results in inconsistent patterns in functional diversity with drastic land-use change, as observed in earlier studies (Rurangwa et al., Reference Rurangwa, Aguirre-Gutiérrez, Matthews, Niyigaba, Wayman, Tobias and Whittaker2021; Sreekar et al., Reference Sreekar, Si, Sam, Liu, Dayananda and Goodale2021). We focused on forest birds and compared tree-crop landscapes (and not open agricultural landscapes) with degraded and less disturbed forests. Therefore, we may have found reduced functional and phylogenetic diversity in modified landscapes.
Forest disturbance and land-use change resulted in declines of functionally important groups such as frugivores. The occurrence of key bird groups, such as hornbills Ocyceros griseus and Buceros bicornis, wood pigeon Columba elphinstonii, fairy bluebirds Irena spp. and multiple endemic bulbuls, which play important roles in seed dispersal (Naniwadekar et al., Reference Naniwadekar, Chaplod, Datta, Rathore and Sridhar2019), was negatively associated with degraded private forests and agroforestry plantations (Supplementary Fig. 4). Large frugivores (e.g. hornbills) play functionally unique roles. They are critical for the seed dispersal of large-seeded plants in the Asian tropics (Naniwadekar et al., Reference Naniwadekar, Chaplod, Datta, Rathore and Sridhar2019, Reference Naniwadekar, Gopal, Page, Ghuman, Ramachandran and Joshi2021) and are thus critical for forest regeneration; their loss can hinder the recovery of degraded forests.
Agricultural landscapes and monocultures typically have lower phylogenetic diversity than forests, suggesting phylogenetic filtering and loss of evolutionarily distinct lineages, and retention of closely-related species (Morelli et al., Reference Morelli, Benedetti, Perna and Santolini2018). The occurrence of members of several distinct lineages (e.g. hornbills, fairy bluebirds, leafbirds Chloropsis spp., leaf warblers Phylloscopus spp.) was lower in disturbed and modified land uses (Supplementary Fig. 4). In other cases, the absence of distinct genera within a lineage (e.g. Rubigula, Acritillas, Microtarsus among bulbuls) in disturbed and modified land uses was associated with the replacement by fewer closely related species (e.g. Pycononotus spp.). Thus, the observed reduction in phylogenetic diversity and associated phylogenetic clustering of the community resulted from the loss of evolutionarily distinct lineages and increased relatedness between the species that persist in altered landscapes.
Broadly, bird diversity declines with increasing elevation, with functionally and phylogenetically clustered communities at higher elevations on tall mountains (Montaño-Centellas et al., Reference Montaño-Centellas, McCain and Loiselle2020), as low temperatures may impose physiological and energetic barriers to persistence (Jankowski et al., Reference Jankowski, Londoño, Robinson and Chappell2013). Although some species appear to respond negatively to elevation gradients, and we detected a reduction in species richness, the cumulative effect did not translate into a signal of phylogenetic or functional clustering of the communities. However, the 1,100 m elevation gradient we examined in the northern Western Ghats is narrower relative to gradients in the Himalaya or Andes, over which functional and phylogenetic filtering have been observed (Penjor et al., Reference Penjor, Cushman, Kaszta, Sherub and Macdonald2022).
Most protected areas in the northern Western Ghats are restricted to higher elevations, with poor representation at lower elevations. We have demonstrated here that low-elevation forests that are not within the protected area network support higher diversity. They also harbour significant populations of threatened species (Mudappa & Shankar Raman, Reference Mudappa and Shankar Raman2009). The mean expected probability of occurrence of the Vulnerable Malabar grey hornbill Ocyceros griseus and Near Threatened Malabar pied hornbill Anthracoceros coronatus was higher in the low-elevation reserved forests than in the high-elevation protected areas. This highlights the importance of remnant patches of reserved forests. The private forests were degraded as a result of periodical clearing. Unlike the Sindhudurg district, where 10% of the district’s geographical area is within reserved forests, Ratnagiri’s forests (60% of the district) are largely privately owned, with just 1.3% within reserved forests (Kulkarni & Mehta, Reference Kulkarni and Mehta2013). The state-governed reserved forests are vulnerable to diversion for non-forestry use and infrastructure development (LIFE, 2020). Given the importance of the reserved forests in this area for avian biodiversity and the vulnerability of private forests to land-use change, the existing reserved forests should be declared conservation reserves, and their denotification should be avoided. Declaration of some of the reserved forests as conservation reserves has already been achieved in the region, setting a precedent for the remaining patches of reserved forests (Bhalerao, Reference Bhalerao2020).
Given the differences in community-, species- and trait- level responses to land-use change, investigation is required of the impacts of such change on biodiversity at multiple levels. This will provide information on responses of evolutionarily distinct and conservation-priority species to land-use change, informing management decisions. Although conversion to different kinds of agroforestry plantations may result in similar structural changes to forests and may negatively impact endemic species and those of conservation concern, the impacts on communities and species vary by plantation type. Studies often pool biodiversity responses to different forms of agriculture (farms, plantations). However, community and species responses must be examined separately for each kind of agricultural practice, as management regimes differ in their effect on biodiversity. More than 100 km2 of forests are converted annually for non-forestry use in India, mostly at low elevations. Linear infrastructure and mining contribute significantly to this loss (LIFE, 2020). Relatively less protected reserved forests are more vulnerable to conversion than protected areas. Given that most of the land in the region we studied is privately owned, the remnant patches of reserved forests are vital reservoirs of biodiversity. Suitable alternatives to the conversion of reserved forests are vital in this region, as these low-elevation forests are critical habitats for threatened birds. Systematic efforts are needed to identify remnant low-elevation forests and accord them better protection. Collaboration between landowners and conservation practitioners is vital for protecting and restoring degraded forests on private lands.
Supplementary material
The supplementary material for this article is available at doi.org/10.1017/S0030605325102184
Author contributions
Study design: RN, AO, KS; fieldwork: SB, NB, VS; data analysis: RN, NB, YVT; writing: RN, NB, YVT, AO, SB, VS, KS.
Acknowledgements
We thank the Maharashtra Forest Department, particularly Chief Wildlife Warden Sunil Limaye, Divisional Forest Officers Nanasaheb Ladkat, Uttam Sawant and Vishal Mali of Sawantwadi and Chiplun Forest Division, and Suhas Patil and Ajit Mali for providing permissions and support during fieldwork; On the Edge Conservation (UK) for funding support; Ninad Gosavi, Himanshu Lad and Siddhi Damle for help with fieldwork; Suri Venkatachalam, Madhura Niphadkar, Atul Joshi, Aparna Watve, Praveen Desai, Milind Patil, Parag Rangnekar, Hemant Ogale, Bhau Katdare, Girish Punjabi, Shashank Dalvi, Kaka Bhise, Prasad Gavde, Mahesh Mhangore, Gajanan Shetye, Vinayak Sapkal, Pooja Pawar and Anushka Rege for support and discussions; and all villagers and households who hosted us during our travels.
Conflicts of interest
None.
Ethical standards
Our research abided by the Oryx guidelines on ethical standards. This observational study did not involve capture or handling of animals, and did not involve interviewing human subjects. The Maharashtra Forest Department (No: Desk-22(8)/WL/Research/CR-53(20-21)/3361/22-22) provided research permits.
Data availability
The data are available from the corresponding author upon reasonable request.
Open access
