Species composition, relative abundance, and habitat association of birds in Dodola dry evergreen afro-montane forest and sub-afro-alpine scrubland vegetation, southeast Ethiopia

Description of the study area

Location

The Dodola natural forest habitat is part of the Adaba Dodola Jalo forest which is one of the 61 National Forest Priority Areas (NFPA) of the country that covers approximately 530 km² (Gelashe, 2017). The Ericaceous sub-afro alpine habitat is found at higher elevations to the natural forest, while the plantation forest below the dry evergreen afro-montane forest. Dodola forest is located West Arsi zone of the Oromia regional state, southeastern Ethiopia (Fig. 1). The study area is adjacent the Bale mountains massif and occurs at 325 km from Addis Ababa towards the southeast, 70 km from Shashemene. The area is bordered by the Kofale district to the west, the Adaba district to the east, the Nensabo and Kokossa districts to the south, and the Asasa district to the north. The geographical location ranges between 6°39′E38°57′N and 7°0′E39°24′N. The altitude range varies from 2,400–3,712 m.a.s.l. The area is a part of tropical forest and tropical shrub land that consists of natural forest (Dry evergreen Afromontane Forest), Ericaceous vegetation (sub-afro alpine habitat) and community plantation forest of a total of 738.30.24 km².

Methods

Preliminary survey

Preliminary assessment was carried out for identification of key habitats during September 2018. To observe habitat type, age effect, topography, and climatic factors for survey design preconditions. During this period, waypoints were collected using GPS in each habitat type (QGIS.org, 2018). A pilot survey was also conducted for sample size information.

Sampling design

A point transect sampling method was used to investigate bird species composition, relative abundance, and habitat association (Buckland et al., 1993). Based on the preliminary survey, the study area was stratified into three dominant habitat types: the sub-afro alpine Ericaceous scrubland habitat; dry evergreen Afromontane Forest; and mixed plantation forest using QGIS. In each habitat type, systematic sampling design was employed. There are eleven blocks: five Erica, five forest and one plantation. The total block area was 128.839 km² area, which is 17.5% of the study area. A systematic point grid of a 1.5-kilometer fixed dimension was randomly superimposed (Fig. 2), and rotation onto the survey region employed proportionally in each habitat type (Buckland et al., 1993). The required number of sample points in the survey region calculated as (1)${\displaystyle \frac{b}{{\left(cv\left(D\right)\right)}^2}\ast \frac{k0}{n0}}$ where k₀ and n₀ are roughly estimated in a pilot survey, and the value of b = 3 (Buckland et al., 1993). In the pilot survey there were five points and 54 individual bird observations. The required number of points was 111 points; 42 points in Erica, 64 in forest and five in plantation (Fig. 2). In cluster: (2)${\displaystyle k=\frac{k0\left\{b+{\left[sd\left(s\right)/s\right]}^2\right\}}{nocv{t}^2}.}$

Data collection

Field guidebooks were tools for identification of the type of bird species exist in the area (birds of the horn of Africa, birds of the East of Africa, birds of Lake Tana, and important bird areas of Ethiopia) (Redman, Stevenson & Fanshawe, 2016). The data collected was carried out for two seasons during the months of July and August for the wet season and December and January for the dry season. Per season, data collection was conducted in two sessions/visits. Data collection was carried out early in the afternoon and late in the afternoon. Detection distances was measured from the point to detected object (Buckland et al., 1993). All observation beyond 70 m sighting distance were truncated. Birds’ songs were used for most elusive forest birds (Buckland et al., 2001). Identification and counting of most bird species were assisted by binoculars. Points taken 200 m distance inside from edge to avoid edge effect. Duration a point count lasts from 2 min to 20 min (Bibby, Jones & Marsden, 1998).

Data analysis

Lists of information about habitat type, season, visit, block, point, cluster size and species code were organized in a single data frame. With the help of R software data organizing functions, for similarity and diversity analysis, the data was organized in form of data frame where rows as species list, and columns as the presence and absence data. One column for a single habitat, and one column for a sample point to similarity and species accumulative curve data analysis respectively.

Data analyzed based on distance sampling method distance 7.3 software (Thomas et al., 2010), and the mark recapture distance sampling (MRDS) analysis engine supplemented by R software (R Core Team, 2019). R software was used to analyze ANOVA test using the Car package, and similarity and diversity indicies were analyzed with the Simba and Vegan package (R Core Team, 2019). AIC and the chi-square statistical test were applied to obtain the best-fitted models (Buckland et al., 2001; Buckland et al., 1993). The result was analyzed based on the data recorded on 111 sample points and 222 total efforts of two replication or visit during both seasons. The analysis of distance was based on the formula described by Eqs. (1)–(4) (Buckland et al., 2001; Buckland et al., 1993).

For point transects analyses MRDS always uses the P3 estimator for encounter rate variance. (3)${\displaystyle v{a}^{\wedge }{r}_{p3}\frac{n}{T}=\frac{1}{T\left(k-1\right)}=\sum _{i=1}^{k}ti\left(\frac{ni}{ti}-\frac{n}{T}\right)}$ where ti is the number of times point i was visited, $T={\sum }_{I=1}^{K}ti,ni={\sum }_{j=1}^{t}nij$ andis $n={\sum }_{i=1}^{k}ninj$ the number of objects detected at point i on visit j.

Relative abundance of avian species determined using encounter rates calculated for each species by dividing the number of birds recorded(n) by the number points (k) multiply time of visit or effort (t) (Buckland et al., 2001).

The encounter rate (ER) was estimated as: (4)${\displaystyle ER=\frac{n}{kt}OR\frac{n}{K}.}$ Encounter rate data was classified into crude ordinal categories of abundance (e.g., abundant, common, frequent, uncommon, and rare) (Table 1).

The number of individuals per total effort were ≤0.01, 0.01–0.2, 0.2–1,1-4 and >4. For each interval, the following abundance labels is given rare, uncommon, frequent, common, and abundant, respectively. Therefore, the relative abundance of each bird species was determined by Excel if function of rare, uncommon, frequent, common, and abundant. For example, if the encounter rate is ≤0.01, the species is considered as rare. Analysis were prepared for two type of data selection steps in multispecies analysis options. the first is setting individual species analysis using data filter, the second was not based on individual species; thus, birds as one taxonomic categories of class of Aves as compared to species taxa. In both steps, habitats were stratum whereas seasons were analyzed by using data filter separately.

A two-way ANOVA was used to analyze density and number of individual observation effect of three factors through season, habitat, and species. The ANOVA type III error to investigate interaction effect (Model 1). The ANOVA type II error was used for incasing of non-interaction effect (Model 2) (Model 1)${\displaystyle {\mathrm{\mu }}_{ijk}=\mathrm{\mu }+{\alpha }_i+{\beta }_j+{\gamma }_k+{\delta }_{ij}\dots \dots \dots \dots .}$ (Model 2)${\displaystyle {\mathrm{\mu }}_{ijk}=\mathrm{\mu }+{\alpha }_i+{\beta }_j+{\gamma }_k\dots \dots \dots \dots .}$ where, μ = the overall mean of species observed, α_i, β_j and γ_j are the i^th, j^th and k^th habitat, season and species effects, respectively. where δ_ij is interaction term (Searle, Speed & Milliken, 1980). Post-hoc test used for separate group analysis for interaction effect results. Estimated marginal means (emmeans) was used for non-interaction effect pairwise comparison of groups. Differences were considered statistically significant at 5% (Chambers & Hastie, 1992). Unbiased sim was calculated as $\tau ={\sum }_{i=1}^S\left(\frac{n_i\left(n_i-1\right)}{N_i\left(N_i-1\right)}\right)$ , Simpson’s index D = $\sum \left({\frac{ni}{N}}^2\right)$ Simpson’s Simpson returns 1-D and inv Simpson returns 1/D (Hurlbert, 1971) $H=-\sum \left(\frac{ni}{N}\right)log\frac{ni}{N}$, $E=\frac{H}{logS}$ where n_i denotes number of individuals in the i ^th species (n_i = 1,2,3…., n and n1 + n2…n = N), S = total number of species (Shannon, 2001). In Fisher’s logarithmic series the expected number of species f with n observed individuals is $f_n=\alpha \frac{x^n}{n}$ The parameter α is used as a diversity index. The parameter x is taken as a nuisance parameter which is not estimated separately but taken to be n/(n+ α) (Fisher, Corbet & Williams, 1943). The species discovery curve was used species richness/number of species discovered across each sample points based on the sample-based rarefaction formula for adequate sample size for a multi-species survey. ${\displaystyle \bar{S}=S_n-{\left(\genfrac{}{}{0.0pt}{}{n-nk}{i}\right)}^{-1}\sum _{k\in G}\left(\genfrac{}{}{0.0pt}{}{n-nk}{i}\right),i=1,\dots ,n}$ A collection on n samples, the rarefaction curve is the plot of $\overline{Si}$ against i (i = 1, …, n), where S_i indicates the arithmetic mean, S_n denotes the total number of observed species, nk denotes the number of samples containing at least one individual species k ∈ G (Chiarucci et al., 2008).

The diversity and relative abundance presented by tables, qq plot and detection function plot. Statistical difference presented through ggplot2 supported by narrative descriptions. Habitat association of number of species were computed for Sorenson’s similarity index (SI) among habitats under two seasons by using the following formula. SI = 2a/2a+b+c; where 2a = number of species common to two habitats, b = number of species in first habitat, c = number of species in the second habitat (Sorensen, 1948).

Species composition

Over the course of two distinct climatic periods (dry and wet), a total of 78 species of birds were recorded. Within the recorded species, the Abyssinian Catbird (Parophasma galinieri), Ethiopian Siskin (Serinus nigriceps), and Yellow Fronted Parrot (Poicephalus flavifrons) have been identified as endemic. Furthermore, there exists a subset of ten species, inclusive of the Wattled Ibis (Bostrychia carunculate), the black-winged lovebird (Agapornis taranta) and Rouget’s Rail (Rouget‘s rougetii), which are recognized as endemic to both Ethiopia and Eritrea (Appendix S1). Based on the lowest AIC value of MRDS analysis engine, the fitted model was single observer distance model and half-normal key function with model for scale parameters is a constant (CDS).

Figure 3 shows the species discovery curve and Fig. 4 shows the extrapolation curve with increasing number of species in the y axis with sample points in the x axis; the curve turns as asymptote shape indicates that the species discover is adequate. The asymptote predicts 86 species to be discovered, which means that over 90% of the species in the area were discovered with a slope 2.62 (the more the slope close to zero, a few or none of species in the area are left detected) (Fig. 3).

The Quantile-Quantile (QQ) plot, which shows the fitted cumulative distribution function (cdf) against the empirical distribution function (edf), represents the number of observed bird species. The dots on the plot correspond to these observations. The line in the QQ plot represents the expected distribution if the model fit was perfect. The proximity of the dots to the line indicates the fit of the model. In this case, the dots surrounding the line suggest that the model is well-fitted (Figs. 5 and 6). The detection function plots illustrate the expected probability density function of frequencies divided by distance. The curve in these plots represents the expected distribution, while the histograms display the number of observations. The unweighted Cramer-von Mises tests a p-value was less than 0.001 in both seasons (Figs. 7 and 8). It is important to note that the detection function depicted in Figs. 6 and 8 represents the overall class Aves. This means it does not account for individual bird species observed in the study.

The species composition of birds during the wet and dry seasons was not significantly different (F, Season = 0.004, p > 0.05) which was 0.95. On the other side, there was a significant difference among habitats (F, Habitat = 12.78, p < 0.05) which was 7.466e−06 ***. There was no season and habitat interaction effect (F2, Habitat: Season = 2.28, p > 0.05) which was 0.11. The estimated marginal means, also known as least-squares means, revealed a significant difference in the mean number of species across two habitat types: Erica and forest. The mean number of species in the Erica habitat was 24 (±3.16 SE), while in the forest habitat it was 22 (±2.33 SE). However, in the plantation habitat, the estimated marginal mean value was −0.8 (±4.3 SE) (Fig. 9).

According to a Tukey pairwise comparison test with a 95% confidence interval, there was no significant difference in the mean number of species between the Erica and forest habitats. Plantation had the least mean number of species. The P value for Erica vs forest was >0.05. The P value for Erica vs plantation and forest vs plantation was <0.01.

The highest species diversity (D) during the wet and dry seasons were observed in Forest habitat (dry evergreen afromontane forest), followed by the Erica (sub-afroalpine) habitat with (0.951 & 0.949) and (0.929 & 0.926) respectively, while the mixed plantation habitat had the least with (0.905 & 0.887). The highest species evenness was observed in the Erica habitat. For the entire season, the forest habitat had the highest species diversity (0.943), while Erica habitat had the highest species evenness (0.85) (Table 2).

Species relative abundance

In the dry season, a total of 2,639 individual birds were recorded, while in the wet season, 2,410 individual birds of 78 species were observed (Table 3). In the 2018 IUCN red list categories, six species faced global threats, three species neared the threat status, and a total of 69 species were classified as least concern.

The mixed plantation forest habitat recorded the highest relative abundance of Aves, with 15.7 and 16.7 during the dry and wet seasons respectively. This was followed by Erica in the dry season with 13.14, and the forest in the wet season with 10.67. The dry season exhibited a higher overall seasonal relative abundance of 11.89 (Table 3).

The relative abundance of individual species in stratified habitat is shown in Tables 4, 5 and 6 in Erica, forest and plantation habitat, respectively. In the Erica (sub-afroalpine habitat), the encounter rate was calculated as number of individual observations in the Erica per Erica point samples times number of a point visit (n/84). In the Erica habitat, the Red-wing Starling had the highest relative abundance during the dry season (1.95), while the Scare Swift had the highest relative abundance in the wet season (1.02). The Chestnut-napped Francolin and Common Buzzard were not recorded in the dry season, and similarly, the Yellow-billed Kite and White-headed Vulture were not recorded in the wet season. During the dry season, four, 10 and 21 species were classified as common, frequent, and uncommon, respectively. During the wet season, one, 17 and 18 species were common, frequent, and uncommon, respectively. Rare and abundance species were not recorded under the two seasons (Table 4). No species were recorded as rare or abundant in either of the two seasons (Table 4).

In the dry afromontane forest habitat, the encounter rate was calculated as number of individual observation per the dry afromontane forest habitat effort (n/128). Montane White-eye was the highest relative abundance during both seasons (1.88 and 1.57). Mouse-colored Penduline Tit, Variable Sunbird, Abyssinian owl, African Stonechat and Common Buzzard were not recorded in the dry season, while in the wet season, Yellow-billed Kite and White-headed Vulture were not recorded. During the dry season two, 14, 39 and two species were common, frequent, uncommon, and rare respectively. During the wet season, one, 19 and 34 species were common, frequent and uncommon respectively. Rare and abundant species were not recorded under the wet season (Table 5). In the plantation forest habitat, the encounter rate was calculated as number of individual observations per plantation forest habitat Effort (n/10). Ground Scarper Thrush was the highest relative abundance during dry seasons (3.1). In the wet season, the yellow crown canary was the highest (2.00). During the dry season two, 10 and 7 species were common, Frequent and Uncommon, respectively. During the wet season, two, four and 14 species were common, Frequent and Uncommon respectively. Rare and abundance species were not recorded under two seasons. six species in the dry season and five species in the wet season were isolated record in the season (Table 6).

Habitat association of bird species

Not all species were distributed in all habitat type. Of 39 different bird species, 15 species specific to Ericaceous sub-afro alpine scrubland vegetation (Table 4), while in Afromontane forest habitat, a number of 59 different bird species were founded, about twenty six species were specific to the habitat (Table 5), In the plantation forest habitat 26 bird species were recorded, there three species specific to community plantation forest habitat (Table 6) and the rest 34 species were recorded either in three or in only the two habitats (Appendix S1). A forest habitat accounts for a high number of bird species and high specific species. The Erica and forest habitats share more species that are common in the dry season (Fig. 10 and Table 7). Plantation and forest share more species that are common in the wet season (Fig. 11 and Table 7). Plantation and Erica share the lowest common species (Table 7).

Low detection frequencies and detection probabilities

Some species, including the African Black Swift, Pied Crow, and Common Buzzard, exhibited low detection frequencies, falling below the standard recommended threshold of 60–80 observations. This could introduce bias into results, as insufficient data for these species may hinder robust analyses (Buckland et al., 1993). Additionally, certain species, particularly those specialized in woodland habitats, displayed lower detectability. This lower detectability, especially for species in closed habitats, may be influenced by various factors, including habitat structure and observer bias (Johnston et al., 2014). For multispecies surveys, it is crucial to account for local habitat effects on all species, not just those with abundant data (Zipkin et al., 2010).

Seasonal variability

The research unveiled substantial seasonal fluctuations in bird abundance, with marked differences between the dry and wet seasons. While data collection was successful in both seasons, some species exhibited stronger presence during the wet season. This observation aligns with existing research emphasizing the influence of seasonal changes in resource availability and weather conditions on avian populations (French & Rockwell, 2011; Li et al., 2022). However, the effect of seasonality on avian species composition may be less pronounced in tropical regions. Seasonal changes in bird populations, feeding habits, and migration patterns are more prominent in temperate regions (Ward, 1969; White, Warren & Baines, 2015). Many birds in the study area are resident breeders, with limited migration during seasonal shifts, possibly contributing to the insignificant effect of seasons on bird species composition (Appendix S1). Migratory birds in Ethiopia are primarily associated with aquatic, wetland, and riverine habitats (Brooks, 2009).

Habitat influence on bird composition and structure

Habitat strongly influenced bird species composition, with distinct preferences observed among different species. Some species displayed specific habitat associations, such as the White-backed Black Tit in forest habitat, Thekla Lark in Erica, and Semi-colored Flycatcher in plantation habitat (Tables 4, 5 and 6). This suggests that avian community composition in the Ethiopian Highlands is intricately linked to habitat types. These findings align with previous studies highlighting the importance of habitat characteristics in shaping avian communities (Aynalem & Bekele, 2008). This suggests that various bird species have adapted to distinct ecological niches within the Ethiopian Highlands. The high species abundance in natural forest habitats further underscores their significance in avian biodiversity conservation. These findings resonate with studies by Hendershot et al. (2020), which underscored the importance of preserving diverse habitat types for effective avian biodiversity conservation. The heightened encounter rate observed within plantation forests (Table 6) suggests the presence of edge effects, particularly as influenced by adjacent agricultural areas. These edge effects are known to attract generalist bird species (Khamcha et al., 2018), which typically exploit transitional zones. However, it is noteworthy that the elevated encounter rate is primarily attributed to a select few bird species that have specifically adapted to the plantation forest habitat.

Microclimate and habitat structure emerged as major drivers influencing avian community composition within specific habitats (Rajpar & Zakaria, 2015). The relationship between habitat and species composition was further evident in the similarity index results, which indicated higher similarity between neighboring habitats (Table 5). Microclimate, habitat structure, and environmental gradients likely contribute to species distribution patterns and preferences within the Ethiopian Highlands.

Altitudinal gradients played a role in avian diversity, with the highest species composition recorded in middle elevation zones, primarily within forest habitats (Quintero & Jetz, 2018). Decreases in diversity at higher altitudes may be attributed to factors such as lower speciation or higher extinction rates, potentially influenced by smaller areas or lower temperatures (Quintero & Jetz, 2018).

The size of the habitat patch and edge effects may also influence avian species composition. Edge-sensitive, neutral, and preferring species respond differently to habitat edges (Brand & George, 2001). Forest species exhibit sensitivity to the contrast between natural and anthropogenic habitats (Zurita et al., 2012). Plantation habitats, characterized by smaller areas, displayed lower species composition estimates (Quintero & Jetz, 2018). As habitat destruction is a significant concern, particularly in forested areas, preserving diverse habitats and their associated bird species should be a conservation priority (Girma et al., 2017; Wang et al., 2017).

Future research in this region should address the limitations of our study. Long-term monitoring with extended survey periods, including intermediate seasons, can provide a more comprehensive understanding of avian population dynamics. In addition, expanding taxonomic coverage and accounting for external factors such as climate change and invasive species will enhance our understanding of avian biodiversity in the Ethiopian Highlands.