Given recent global endeavors to increase protected area coverage, it is crucial to comprehensively evaluate the efficacy of various area-based conservation strategies in effectively reducing biodiversity loss. Here, we investigated responses of wildlife populations to different protection levels and environmental variables at the landscape scale in the Katavi-Rukwa Ecosystem, western Tanzania. To this end, we conducted line distance sampling surveys and counted dung of six target mammal species (elephant, giraffe, buffalo, zebra, topi, hartebeest) along foot transects within areas differing in protection levels (from strict to less-strictly protected: national park, game reserve, forest reserve, game-controlled area, and unprotected areas). Based on these dung counts, we modelled the spatial distribution of these six mammal species using a species-specific density surface modelling framework. We, found consistent effects of protection level and land-use variables on the spatial distribution of the target mammal species: dung densities were highest in the national park and game reserves, intermediate in less-strictly protected areas and lowest in un-protected areas. Beyond species-specific environmental predictors for dung densities, our results highlight consistent negative associations between dung densities of the target species and distance to cropland and avoidance of areas in proximity to houses. Our findings underpin differences in ecological effectiveness of protected areas within one ecosystem. Protection level and land use play crucial roles in moderating the spatial distribution of all considered mammal species. Our findings suggest that a landscape approach needs to guide effective conservation across the entire protection gradient of the Katavi-Rukwa Ecosystem.

We established a 35 km buffer around the boundary of the KNP to include areas with different protection levels, ranging from unprotected to strictly protected. To capture an even coverage of transects across different protection levels, we divided the study area into 5 km by 5 km grids, so that transects were separated by 5 km to minimize spatial autocorrelation of the data. We randomly selected 105 grids (21 grids within each protection level) and placed triangular-shaped transects of 3 km total length in the centre of each selected grid. To ease logistics in the field, we opted for 1 km segment length. We surveyed each transect once during the dry season between July and September 2021. Three people (one each primarily responsible for navigating, observing, and recording) walked along the transect. We used a handheld GPS and compass to navigate between segments (i.e. we moved towards 90° E in the first segment, 330° NW in the second segment, and 210° SW in the third segment). In each transect, we counted and recorded the number of individual dung piles. Upon detection, we identified each dung pile to species level. To measure perpendicular distances from the centre of a dung pile to the centre line of a transect, we used a tape measure. To define the centre line of the transect (and avoid rounding of distances near the line to zero), we placed a walking stick in the centre of our path (Marques et al. 2001). Before the formal survey, we conducted a literature review and a pilot survey to establish species-specific dung pile definitions based on the shape of pellets and quantity of pellets per dung pile. We analysed the data in a density surface modelling framework, a two-stage method which first accounts for uncertain detectability (primarily as a function of distance between transect and observations) and a spatial model of the density of the target population.

Microsoft office excel and R