Materials and methods
Study area
Tigray is one of the nine autonomous National Regional States of the Federal Democratic Republic of Ethiopia (FDRE, Fig. 1A). It is divided into seven administrative zones namely: Western zone, North-western zone (NW), Central zone, Eastern zone, South-eastern zone (SE), Southern zone and Mekelle special zone which are further divided into more 47 administrative districts (Fig. 1B). The region has several agro-ecological zones that fall under three major biomes namely: Sudan-Guinea Savanna biomes; Afrotropical Highlands biomes and Somali-Masai biomes (Fishpool & Evans 2001; Haileselasie & Teferi 2012). It has four traditional agro-ecological classifications: The “Qhola” (below 1500 m a.s.l), “Weyna Degu’a” (1500-2300 m a.s.l), “Degu’a” (2300-3200 m a.s.l), and “Wurch/Alelama” (above 3200 m a.s.l). The highest mountain range in the region is the “Tsibet Sky Island” with its highest peak reaching 3960 m a.s.l, located in the Southern Zone of Tigray National Regional State and the lowest point is around 500 m a.s.l in the Tekeze valley, western Tigray.
Here, 35 reservoirs of varying age were included in a study of avifauna of the limnetic aquatic ecosystem of the region (see Appendix S1 in supporting information). These reservoirs were purposively selected because of previous experience to each site (Haileselasie & Teferi 2012; Teferi et al. 2013; Haileselasie et al. 2018). The reservoirs have different age from the youngest Mihtsab Azmati reservoir (MAZ; 5 years) to oldest Bokoro reservoir (BOK; 44 years). They also vary in elevation (range: 1512– 2747), water surface (range: 1.78– 45.41 ha) and water depth (range: 1– 13.7 m). To evaluate the relationship between limnological characteristics of reservoirs and patterns of bird species richness and distribution, we selected nine variables: reservoir area, water depth, elevation, pH, nutrient concentration (Total Nitrogen & Total Phosphate), water turbidity, water temperature and electrical conductivity that are known to influence ecological processes and species abundance in freshwater wetlands and lakes (Hoyer & Canfield 1994; Seymour & Simmons 2008; Rajpar & Zakaria 2010; van der Valk 2012). Ecological variables such as: presence and absence of fish, emergent vegetation, presence and absence of forest edge and/or downstream wetland were recorded and coded into an appropriate format for analysis. Limnological characteristics of each reservoir: reservoir’s morphometry (altitude, area and maximum depth) and water chemistry variables: water temperature (°C), pH, electrical conductivity (EC, µS/cm) were measured in the field using portable pH/EC/ multi-meters. Whereas integrated water samples were collected and brought to Aquatic Ecology Laboratory (Mekelle University, Tigray) for determination of Total Phosphorus (TP, μg/l) and Total Nitrogen (TN, μg/l) in the laboratory using standard methods stated in Nelson and Sommers (1975).
Bird surveys
Data for this study were obtained by counting birds that were observed during a survey of 35 reservoirs (see Fig. 1). Birds observed utilizing limnetic ecosystems were recorded by observers who motored around each reservoir in a small boat and/or by walking along the perimeters of the reservoir depending on the size of the water body. Birds were identified to a species level based on bird’s field guide for East Africa; birds of sub-Sahara and country specific checklists (Urban & Brown 1971; Sinclair & Ryan 2003; Ash & Atkins 2009). Species richness in this article’s context is defined as the total number of bird species observed throughout the entire sampling period in the region (as gamma diversity) and number of bird species recorded in each reservoir (as alpha diversity). Here no attempt has been made to calculate annual bird abundance (birds/area) for each reservoir. English names and Taxonomy of birds reported here, follows the International Ornithological Congress (IOC) standard format (Gill & Donsker 2020).
Data analysis
Differences in limnological variables between reservoirs are visualized by Principal Component Analyses on residuals of full limnological data set. Principal components were extracted from covariance matrices using the function rda in vegan package of the R software (R Development Core 2014). The Eigenvalues and % of variance for each axis were used to retain number of significant PC axis for further analysis. And the Euclidean distance after standardizing the variables, followed by Ward clustering is used to display plot of the first two PC axis of limnological components.
To partition gamma diversity into its alpha (α) and beta (β) components; gamma (γ) diversity of birds as species richness with q=0, is equated as multiplicative (i.e. α*β= γ) relationship. As a result beta (β) diversity is calculated as gamma diversity divided by mean alpha diversity, with all samples being equally weighted as applied in the R-software package vegetarian (Jurasinski, Retzer & Beierkuhnlein 2009).
To explore relationships among environmental variables (ENV) and geographic location (SPACE) of the reservoirs and bird species richness, redundancy analysis ordination (RDA) is performed. Species richness data was submitted to a multiple regression analysis at limnological variables (ENV), biological variables (BV) and age of the reservoirs in order to investigate the most important explanatory factors influencing avian species richness and their distribution. The Monte Carlo Permutation test of 999 permutations is used to test statistical significance of the relationship. Pearson’s correlation coefficients is used to examine correlations between the variables and to reduce the number of explanatory factors.
Besides, the scores of species (alpha diversity) and environmental variables resulting from the ordination is used to build a bi-plot that illustrates the relationships between environment and bird species richness. To describe the environmental preferences of particular species, Redundancy Analysis ordination (RDA) in R software (R Development Core 2014) was applied. The function partitions the variation-varpart in vegan R package (Oksanen et al. 2013) using adjusted R-squared (R2adj) in redundancy analysis ordination (RDA) is used to disentangle the effect of these variables: in species - environment - space - age variation partitioning by partial regression.