Context: Globally, an estimated 1.3% of the bird population has gone extinct over the last millennia largely due to loss of habitat preferences.

Aims: This short-term study investigated disturbance-related drivers as predictors of bird habitat selection and assemblages in the Owabi Ramsar wetland.

Methods: The study was carried out over a five-month period (May to September 2019), in four habitat types (farmlands, Built-up, forest reserve, and open water area). Data was collected in 84 plots across four habitats, using a point-count cantered technique. The Gambin model, nMDS, Chao-1, and Hill numbers models were used to evaluate bird distribution, habitat preferences, and diversity, while the CCA ordination technique was performed to examine the influence of drivers on bird assemblages.

Key results: In all 1,260 individual birds, belonging to 81 species were identified. The Majority of the birds preferred the farmlands and built-up habitats in spite of the severe disturbance (e.g., crop farming, sand winning, and fire), accounting for 55.39% variability in their community structure. The open water was the least preferred habitat and was dominated by the White-faced whistling duck. Despite the drop in species similarity with increasing disturbance from the open water to the built-up, fewer than five species were widely distributed in all the four habitats (e.g., Bronze-mannikins and White-throated bee eater), indicating their broad range habitat preferences and ability to adapt to varied conditions. The forest reserve tended to be the most diverse, reflecting the spatial distribution of birds mediated by nesting microhabitats, varied food availability, less predation, and low disturbance.

Conclusions: The study adds to previous work on the impact of increasing disturbance on the Owabi Ramsar wetland ecosystem. This study particularly highlights the role of disturbance-related drivers and habitat type in bird habitat preference and the need to intensify conservation activities within the catchment of the Owabi Wildlife Sanctuary.

Implications: Giving the increasing level of disturbance, there is the likelihood that the forest and water-dependent bird population will decline sooner if managers of the wildlife sanctuary fail to halt dumped solid waste in the open water, logging and expansion of croplands into the forest reserve.

Study area and sites

Owabi Wildlife Sanctuary was established in 1971 and is located approximately 23 km northwest of Kumasi in the Ashanti Region of Ghana (Oduro and Aduse-Poku 2005). The study area is located in the Owabi Wildlife Sanctuary (6⁰ 47’ 3.32”- 6⁰ 41’ 52.31” N and 1⁰ 44’ 0.81”- 1⁰ 37’ 53.04” W) in the Atwima, Kwabre District, Ashanti Region – Ghana (Figure1). It is one of the smallest (13 km square where the study will be conducted) conservation areas in the Upper Guinea Rainforest of Ghana. It is also the only inland Ramsar site protected under the Ramsar Convention (Oduro and Aduse-Poku 2005). It was designated a Ramsar site in 1988 and managed by the Forestry Commission of Ghana (Garshong et al. 2013). The wetland, which is approx. 22 ft deep and covering an area of 2.03 ha, with seven streams draining into it (Nartey et al. 2019) is surrounded by a forest and a Wildlife Sanctuary. The wetland is one of the major supplies of potable water to the Kumasi Metropolitan Area (Nunoo et al. 2012). The Wildlife Sanctuary is rich with indigenous water birds and various species of wintering and staging birds during migration and supports about 199 vascular plants species (Garshong et al. 2013). The climate is semi- humid, with an average annual rainfall of 1,488 mm. Rainfall pattern if bimodal, with the main wet season between March and June and a minor wet season in September – October. The average annual temperature ranges between 24–34°C. Soil type is typical forest ochrosol. The surroundings of Owabi wetland and the wildlife sanctuary are subjected to a variety of human uses, including crop farming, sand winning, livestock grazing, logging, and fishing.

Site/plot selection and bird sampling procedure across the four habitats

Four habitats (Farmlands, Built-up, Forest reserve and Open water area) were selected on a current remote sensing image (e.g., TM Landsat 4 image using the stratified random approach. A field reconnaissance survey was then carried out to validate the locations selected for the bird survey. This was followed by another stratified random selection of the survey plots by overlaying 100 grid cells of 25 x 25 m on the TM Landsat 4 image, to cover all the four selected LULC habitats. We employed the fixed radius point count technique of 25 x 25 m dimensions (250 m²) (Hutto et al. 1986; Bibby et al. 1992; Sunderland 2006) and at least 150 m apart, to count birds in each of the four sites (i.e., built-up area = 25, forest reserve = 25, agricultural land = 25 and open water area = 9 P.C.C plots); bringing the total to 84 survey plots. Bird count was carried out twice a day; in the early morning hours (6:30 - 9:30 G.M.T.) when they are out feeding and late afternoons (16:30 - 18:30 G.M.T.) when they were most active at low ambient temperature. We slowly walked through each survey plot to visually identify birds at close range or used a pair of Bushnell Falcon binoculars (of 10 x 50 mm dimensions) for birds at distant locations up to 20 m from the centre of the survey plot. In using the pair of binoculars, we looked out for morphological features such as like colour and structure of the beak, colour of a tail feather, colour of feathers around the neck, colour of the head feathers and the presence of comb-like feathers. Birds that were not visually spotted, were identified by their vocalization, with the aid of a highly trained person (Mr. Ossom - Ornithologist) for recognizing of bird species. We avoid counting fly over birds because we assumed that their habitat preference did not strongly relate to the land use types found within the study area.

During windy days, bird count was avoided, as sighting birds might be difficult. Bird nest was not counted, since it was practically impossible to establish the species and number of birds co-habitating a nest. Bird count was done on a weekly basis (12 weeks in total), between August and November 2019. All birds were identified in situ with the aid of Birds of Ghana keys (Borrow and Demey 2010; Birds of Ghana Galleries, 2010, http://www.pbase.com/)

Assessment of disturbance-related drivers

Eight disturbance-related drivers such as erosion, farming, fire, logging, grazing, bare ground, sand winning, and solid waste were first identified on a ground truth assessment and subsequently categorized by adopting Salafsky et al. (2003) and Battisti et al. (2009). This assessment was done to evaluate their effect on bird assemblages across the four habitats. These disturbance-related drivers were assessed weekly for their magnitude (i.e., scope + severity), and in the same plots where birds were sampled.

The scope and severity of identified drivers were scored from 1 - 4, with 1 representing low scope and magnitude and 4 represent high scope and magnitude. The scope was assessed according to the percentage of the study plot affected by a given driver within the preceding five years (2014 -2019) (Battisti et al. 2009). To verify whether the identified activities had indeed persisted for at least the preceding five years, one-on-one field interviews were held with managers and users of the wetland (35 persons in all). The scores for scope were assigned as follows: 4: found throughout (50%) the sample plots at the sites; 3: in 15–50% of the plots; 2: found in 5–15% of the plots and 1: found in < 5% of the plots (after Battisti et al. 2009).

For ‘‘severity’’ we referred to the degree or intensity to which a driver has physically damaged each plot per site in all four sites. A score of 4 was assigned if the driver led to a severe loss of ground cover and eroded surface; 3 = if the impact induced severe damage; 2 = if the impact led to moderate damage; 1 = if impact induced little or no damage. The magnitude was computed as the sum of the average scope and severity of an activity per plot and ranked (Salafsky et al. 2003).