Movement of avian predators points to biodiversity hotspots in agricultural landscape

Global agricultural landscapes are witnessing a concerning decline in biodiversity, and this trend is predicted to persist. To safeguard these biodiversity-rich areas, it's crucial to pinpoint hotspots effectively. In doing so, we utilized various species of birds of prey as suitable sentinel animals due to their mobility and dependence on prey diversity and abundance. Between 2019 and 2021, we tracked 62 individuals from four predator species using GPS loggers in Estonian farmland. Dividing the study area into 50-meter grids and overlaying them with tracked individuals' locations enabled us to differentiate between hotspots of their activity and control sites. We conducted surveys on amphibians, birds, small mammals, and plant diversity to determine if avian predator activity hotspots correlated with overall biodiversity. Our findings revealed significantly higher diversity and abundance in the surveyed groups within activity hotspots compared to control sites. These hotspots continued to be frequented by raptors in the subsequent year, albeit not two years later. In conclusion, multispecies GPS telemetry of avian predators emerges as an objective, dependable, and spatially accurate biodiversity indicator. With the accumulation of movement data, we anticipate increased interest and adoption of this approach in biodiversity monitoring.

Methods
GPS Telemetry Dataset

Starting from April 2019, we trapped individuals of four different species of birds of prey in their known breeding territories. We used large mistnets with stuffed specimens of large avian top predators (white-tailed eagle and eagle owl) to provoke the attack of the focal species near their nest sites and to catch them in the net. We managed to trap and track 62 individuals of four different species (Table S1). As much as 59 individuals were caught during this project, but we also used additional data from three lesser spotted eagles that were caught earlier in the same area and were still transmitting data (Väli et al., 2020).  

We used 15 – 30-g GPS-loggers with solar panels (Ornitela, Lithuania), which were selected according to the body mass of each studied bird so as not to exceed the 3% threshold that is currently considered as acceptable by many bioethics committees. GPS loggers were attached to birds as backpacks that were sewn at the sternum with Teflon ribbon. Handling time reached 30 – 60 minutes per individual. In any case there was no mortality in following months and none individuals deserted their brood. GPS data was collected at 3 to 60-minute intervals depending on the battery level of each individual’s logger. Marsh harriers and lesser spotted eagles flew extensively and charged their tags through sunlight, so they mostly were able to acquire data with 3 to 10-minute intervals. Common buzzards and northern goshawks spent around 90% of their time perching and mostly under canopies (Mirski & Väli, 2021). Therefore, their charging ability was much lower, and GPS fixes were usually acquired in 10 to 60-minute intervals, depending on the individual.

Hotspot designation

To test whether the raptors’ space use would lead us to biodiversity hotspots, we first had to designate the hotspots of raptors activity – i.e. sites of highest use by different species and different individuals followed by GPS telemetry. For comparison, we also chose control sites. To designate both, we divided the study area into 50-m × 50-m grids, drawn in QGIS 3.22 and subtracted the forest, waterbodies, and peatland layers. The chosen grid size is a few times higher than the GPS positioning accuracy and precise enough to register selected habitat patches with sufficient detail.

Activity hotspots and control sites were chosen in a few sets and progressed with our knowledge on space use by tracked birds. Data on the first half of birds marked with GPS tags in 2019 (n=35) was used to designate pairs of hotspots and control sites for the biodiversity survey (bird survey) carried out in 2020. Chosen activity hotspots were then completed with the subsequent years based on movement data from early spring 2020 (April) and late spring 2020 (May). Activity hotspots for biodiversity surveys in 2021 were chosen using data from all of 2020, early spring of 2021 (April), and late spring of 2021 (May) (see Table S2 for the number in each set). We considered a grid cell as an activity hotspot when at least three different individuals of two different species were using it in a given season (see Fig. 1C). If more such grid squares were located in the adjacent position, we chose the square with the highest number of species recorded inside. Squares with obvious perching posts (such as single trees or pylons) were discarded, as at a given spatial scale it may attract birds of prey more as suitable perching or roosting site than as foraging site. When choosing between squares with the same number of species, we chose one with a greater number of individuals (with the highest priority) or observations (with secondary priority). To avoid choosing sites that were too close to each other (bird prey survey involved detectability radius of at least 100-m from the grid centre), we discarded other activity hotspots or control site in grid cells closer than 450 m from the chosen activity hotspot.

Control sites were designated as farmland areas with similar characteristics to the hotspots and were physically accessible to the individuals we tracked. We selected them as pairs with an activity hotspot, always located  500 meters from their centers. The procedure began with the extraction of centroids of activity hotspots and the creation of 500-meter buffers around them. We initiated the search for a control site starting from the first grid square intersected by the buffer at a 0-degree azimuth from an activity hotspot, moving clockwise. The control site was the first suitable cell found, with the following exclusions: 1) cells that had more than one observation of tracked raptors in a given spring (to avoid similarities with hotspots), 2) cells that were closer than 500 meters to already chosen activity hotspots and other control sites in the same season, and 3) cells containing landcover other than farmland.

Hotspots and control sites were designated using part of the tracking data in a given season or even utilizing data from the previous season. To verify the disparities in raptor activity at these sites, once the breeding season data was completed, we tallied the number of all observations, the count of different individuals and species, and the number of different days these sites were used by the tracked birds during the season, within each grid square (all done in QGIS 3.22). To accomplish this, we didn't employ the raw dataset. Instead, we resampled the data into 15-minute intervals using the amt package in R (Signer et al., 2019). This approach allowed us to objectively assess the accuracy of activity hotspot selections, irrespective of variations in data collection among the best-charging individuals.

Biodiversity surveys 

We measured the species richness and abundance of amphibians, birds, and small mammals, as well as the species richness of vascular plants. The sample sizes differed between the studied groups because of large differences of labour in inventories of various taxa from 15-min bird counts to 48-h small-mammal trappings. However, the ratio between activity hotspots and control sites was always drawn in equal proportions.

We conducted the inventory of vascular plants in July–September 2020 – 2021. In total, we studied 120 50-m × 50-m plots (68 in 2020, 52 in 2021; Fig. S1). In each plot, we assessed the species composition and the coverage of each species in five 1-m2 squares and estimated species richness as a total species list of all five. We also calculated the Jaccard index as a diversity measure to estimate the species similarity between the squares (Maguard 2004). The plant species list of five 1-m2 squares were compared in pairs, and the average Jaccard index was calculated for one plot. The smaller the average value of the Jaccard index, the greater the difference was in the species composition between the squares, and the greater the species diversity was in the 50-m × 50-m plot.

We conducted bird inventories in 74 plots in the first year and in the same number in the next year following the methodology of point counts (Bibby et al., 2000). We counted birds twice in a season (in the first half of May and in the first half of June) and averaged the numbers. The observer stayed at the centre of a 50-m × 50-m plot for 15 minutes and recorded all birds within a 100-m radius (Fig. S2). Hence, the actual studied area was larger than the 50-m × 50-m plot because we expected that the observer may scare birds away from the plot.

We studied small mammals in 92 plots (46 plots per study year). The main dataset was collected by trapping small mammals in August and September. In each 50-m × 50-m plot, we conducted trapping in five 1-m2 squares (Fig. S2). In each 1-m2 square, we placed a set of traps consisting of two snap-traps, two box-traps (Sherman large; 23×9×8 cm3), and one trap-hole (a cone made of a bottomless plastic bottle with a depth of about 25 cm and an upper diameter of about 10 cm). Each trap was baited with a piece of black bread. We trapped small mammals for two consecutive nights in each plot. We set the traps up in the evening and checked them in the next morning. We counted amphibians in the course of other inventories (birds, mammals, vascular plants) in a total of 92 plots (46 annually). Observers were walking along five 50-m long transects, in each covering 10-m stripe of the ground in their eyesight.