Habitat variables influencing bird assemblage patterns across maize and legume fields before and after harvest in West Africa
Data files
Jul 30, 2024 version files 267.13 KB
-
EcounterrateWLB.csv
194.58 KB
-
README.md
3.22 KB
-
RichnessWLB.csv
57.83 KB
-
RscriptWLB.R
11.50 KB
Abstract
Though agriculture has been linked to the decline in bird populations due to the associated changes in vegetation structure and composition, its potential to sustain birds has been explored over time. A sustainable agricultural landscape should be one that has the capacity to support bird species all year. To this end, we investigated the factors influencing pre- and post-harvest farm conditions on bird abundance, richness, and feeding guilds in three different crop fields in Jos-East and North, central Nigeria. We used line transects to survey birds and vegetation variables from 30 fields each of legumes, maize, and mixed (maize and soya bean) during the pre- and post-harvest periods in central Nigeria. We modeled the effects of field conditions and vegetation parameters on bird species richness and encounter rates at overall and feeding-guild levels. Our results showed that bird species richness was significantly higher pre-harvest than post-harvest. Bird encounter rate and feeding guild encounter rate were not affected by pre- and post-harvest conditions. Bird species richness, encounter rate, and guild encounter rate were significantly higher in legume and mixed crop fields than in maize fields. In addition, in-field tree density had a significant positive influence on bird and guild encounter rates, and species richness. Bird encounter rate and richness were significantly higher when the nearest noncrop vegetation was either a gallery forest or rocky outcrop. A similar trend was observed for both insectivore and granivore encounter rates. Bird encounter rate declined with increasing distance from water sources and noncrop vegetation. This study shows that pre- and post-harvest conditions of crop fields can moderate the number and richness of birds on farmlands while the retention of trees on farmlands contributes to higher bird assemblages.
https://doi.org/10.5061/dryad.g4f4qrfzw
The dataset contains variables recorded during the study. Birds were recorded during the pre- and post-harvest periods using a line transect. Each site had three crop types of 10 replicates and was visited four times in each period. Birds seen or heard during each visit were identified and numbers recorded.
Description of the data and file structure
Two separate data files contain numerical, continuous, and categorical variables; the EncounterrateWLB and RichnessWLB files.
Unique variables in the EncounterrateWLB file are:
-
number: refers to the count of birds recorded in each field
-
encounter_rate: refer to the calculated encounter rate of birds using the formula “number of birds/length of the transect”
-
guild: refers to the feeding guild of individual bird species
-
species: contains the scientific name of each bird recorded
Unique variables in the RichnessWLB file is Richness which is calculated bird species richness using the vegan package in R.
Common variables in both files are:
-
N-trees: refers to the number of trees in each field.
-
crop height: refers to the average height of crops in centimeters recorded within a 1X1m quadrat in each field.
-
position: shows if a bird was sighted at the edge or within the field.
-
D_vegetation: indicates the distance from the field boundary to the nearest non-cultivated vegetation. It is recorded in a distance band where 1 = distance within 100 m; 2 = distance between 101 m and 300 m; 3 = 301 m – 500 m; and 4 = > 500 m.
-
vegetation type: shows the nearest non-cultivated vegetation type closest to each field.
-
matrix habitat: refers to the type of habitat that borders the field on all four sides. It is categorized based on the dominant vegetation type: a monotype habitat, where at least two similar habitat types border the field; cultivated habitat, where at least two fields border the field; and mixed habitat, where three or more different habitats border the field on all sides.
-
Water: refers to the distance from the field boundary to the nearest water source. It is recorded in distance bands where 1 = distance within 100 m; 2 = distance between 101 m and 300 m; 3 = 301 m – 500 m; and 4 = > 500 m.
-
D_trees: refers to the distance from the field boundary to the nearest tree in meters.
-
Tree.density: refers to the density of trees recorded within each field. It was computed as the “number of trees/area of field”. Unit is the number of trees per square meter.
-
Area: refers to the size of the field and was calculated as length * breadth in meters
-
farm code: refers to the unique identification for the 10 replicates of each field type at each study site.
-
before/after: refers to the period of data collection, i.e. whether data was collected pre-harvest or post-harvest- categorical
-
session: number of times each field was visited during pre- and post-harvest periods of data collection.
-
site: refers to the study location where data was collected
Bird Sampling
Bird species were surveyed using the line transect method. A transect was laid along the longest side of the field such that it was positioned between the field and a 10 m matrix habitat (Figure 2A and B). All fields used in the study had a nearly rectangular shape. The open nature of the area made it possible to detect birds ahead of time and the small size of the field allowed birds to be detected from one end of the field to the other while walking a transect. However, in instances where the observer was uncertain if a sighted bird was within 10 m of the field edge, the observer noted the bird's perch location relative to the closest landmark. After walking the transect, the distance from the field boundary to where the bird was perched was measured using a GPS. Additionally, to address challenges in counting birds at the edge on the far end of the field, the observer strategically positioned himself upon reaching the field to scan through the edge without disturbing birds, recording any sightings. Afterward, he proceeds to walk the transect. The length of the transect varied due to differences in the size of fields selected for the study and it ranged from 49 m to 274 m. This includes a two 10 m extension of the matrix habitat for collecting edge data. The matrix habitat of a crop field was defined as a length of 10 m extensions from all sides (each crop field had four sides) of the field boundary (Figure 2A and B). The width of the fields also ranged from a minimum of 20 m to a maximum of 100 m and included a 10 m extension of the matrix habitat. Birds were viewed with the aid of a binocular (NATURE-TREK®). A GPS (Garmin, eTrex® 10) was used to mark the length of each transect and to measure the area of the field. The area of the field was thereafter computed using the formula “Length × Breath” of each field. There was a minimum distance of 180 m between the fields except for two fields that had a distance of 100 m between them. Each site was visited four times before harvest and four times after harvest between 06:30 to 11:30 hours on each sampling day. During each sampling session, the transect was walked at a steady pace and all birds seen and heard, including migrants, in the field or at the defined matrix habitat of the field were recorded. Migrants refer to Palaearctic passerine species wintering in the West African region between the months of September and April. Birds sighted within a field were categorized as “in-field” birds while the birds sighted at the defined matrix habitat were assigned “edge” birds. Birds that were seen flying toward the direction of the next study field were noted and excluded if found in the next field to minimize the possibility of double counting. Data collection was conducted on days when the weather was clear and visible. We classified all recorded bird species into six mutually exclusive dietary guilds based on their predominant diet which are carnivores (feed predominantly on vertebrates), frugivores (feed predominantly on fleshy fruits), granivores (feeds predominantly on seeds), insectivores (feeds predominantly on insects) and nectarivores (feed predominantly on nectar) (Kissling et al. 2012; Sulemana et al. 2022).
Vegetation Survey
Distance from the field to the nearest natural vegetation was measured in distance band where: 1 = distance within 100 m; 2 = distance between 101 m and 300 m; 3 = 301 m – 500 m; and 4 = > 500 m. The natural vegetation in the study area was either a gallery forest, savanna, rocky outcrop, or woodlot. The number of trees within the defined matrix habitat on every side was counted. A tree here refers to a plant that has a diameter at breast height of ≥ 10 cm (Ivande & Cresswell, 2016). The distance from the field boundary to the nearest tree on any side of the field was recorded using a GPS. Furthermore, the matrix habitats on the four sides of the study fields were a mix of gallery forest, savanna, rocky outcrop, woodlot, or different crop types. These were categorized based on the dominant vegetation type: a monotype habitat, where at least two similar habitat types border the field; cultivated habitat, where at least two fields border the field; and mixed habitat, where three or more different habitats border the field on all sides. We estimated in-field tree density using the number of trees in the field based on the formula:
In-field tree density = number of trees/ area of field(m2)
In addition, we estimated the distance to the nearest water source categorized as follows: 1 = distance within 100 m; 2 = distance between 101 m and 300 m; 3 = distance between 301 m and 500 m; and 4 = > 500 m. Furthermore, we measured crop height from three points in the field using a measuring tape and a wooden rod marked at a 2 cm interval. Within each field, a 1 × 1 m quadrat was laid 5 m from the diagonal ends into the field and at the centre of the field. The average height of crops that fall within this quadrat was recorded. A full list of variables measured can be found in Table S1.
Data Analysis
We calculated the encounter rates of birds using the formula,
RE=n/Lt
where RE = encounter rate; n = total number of birds recorded in each field; and Lt = length of transect in meters.
We further computed the encounter rates for the feeding guild level. Bird species richness was estimated as the total number of species in each field type. We consider the effect of detectability to be minimal because the transects were laid along the boundary of the field such that it was possible to scan through the field for bird species. Thus, we have used the total number of species encountered as a measure of species richness. All birds recorded, including migrants, were included in the analysis. We used a generalised linear model fitted with quasi-Poisson distribution and log link function to analyse the data due to over-dispersion. We checked for over-dispersion using the “performance” package in R (Ludecke et al. 2021). Preliminary analysis indicated that the in-field number of trees positively correlated with in-field tree density, while the size of the field correlated positively with transect length (n = 90, r = 0.85, p < 0.001 and n = 90, r = 0.78, p < 0.001, respectively). We dropped the in-field number of trees and size of the field, which were the only significantly correlated variables in our data set. For each response variable, i.e. bird encounter rates, bird species richness, and feeding guild level encounter rates, we built a global model to predict how they are each influenced by the dependent variables. The dependent variables (with interactions) are tree density, distance to nearest water source, distance to nearest natural vegetation, season, position, crop type, natural vegetation type, matrix habitat, site, and crop height. For each model built, transect length was fixed to eliminate its potential confounding effect. A complete list of the global model for encounter rate, species richness, and feeding guild level encounter rate, as well as the best models, are shown in Tables S3 and S4, respectively. We arrived at the best model by manually removing one at a time, thus those variables with insignificant p-values. Histogram plot of residuals was used to check the model fit. Means are reported as mean ± standard error. Plots of marginal effects were created, with values generated using the 'ggeffects' package (Lüdecke 2018), except for the plots of tree density, which were plotted directly from the data because the predicted values consisted of only one observation. All analyses were performed using R statistical software (R Core Team, 2022).