Assessing the effectiveness of a national protected area network for carnivore conservation

Protected areas (PAs) are widely recognized as vital tools for biodiversity conservation, aiming to mitigate the impacts of increasing human pressures on ecosystems and species. Despite their importance in safeguarding biodiversity, maintaining ecosystem function, and contributing to human well-being, a comprehensive understanding of their effectiveness remains limited. While past research has focused on PA efficiency (e.g., the extent to which species ranges are covered by PAs), there's a growing need to assess their ecological effectiveness. Recent studies have explored the effectiveness of PA management and their impact on reducing specific threats, such as deforestation. Methodological advancements, particularly the use of matching tools, have improved assessments of PA impact on deforestation by accounting for confounding factors that might influence deforestation rates independently of PA status. However, extending these methodologies to evaluate the impact of PAs on wildlife populations presents significant challenges. This is due to the complexities of wildlife ecology and the limited availability of comprehensive datasets suitable for comparison of protected and unprotected sites with similar environmental characteristics. Large carnivores, facing considerable threats from anthropogenic pressures such as habitat loss and human-wildlife conflict, are particularly important to study. This research focuses on evaluating the effectiveness of Finland's PA network in supporting the populations of four large carnivore species: Eurasian lynx, gray wolf, wolverine, and brown bear. The unique long-term dataset from the Finnish Wildlife Triangle Scheme (FWTS) provides an opportunity to apply innovative statistical methods, combining matching methods and hurdle-mixed effects models, to strengthen inferences about PA effectiveness on a national scale. The hypotheses considered were that PA impact varies between species, regions (with lower effectiveness predicted in southern Finland and Lapland), and over time (potentially declining effectiveness due to increased human-carnivore conflict).

Previous research has yielded mixed results on the effectiveness of PAs in protecting wildlife. Some studies have reported more positive wildlife population trends within PAs compared to outside, while others have found stable trends or a significant contribution of PA networks to megafauna habitat protection. However, concerns remain about potential overestimation of PA impacts due to the omission of counterfactuals in some earlier studies. The application of matching approaches to wildlife data has lagged, primarily due to a lack of extensive datasets allowing for comparison of protected and unprotected sites with similar ecological characteristics. While matching methods have shown promise in assessing the effects of PAs on deforestation, their application to wildlife populations, which often exhibit complex spatial and temporal dynamics, poses additional challenges.

This study leverages a unique long-term dataset from the Finnish Wildlife Triangle Scheme (FWTS), which has collected wildlife abundance data across Finland since 1989 using approximately 2000 12-km transects. The dataset includes yearly estimates of the distribution and relative abundance of four large carnivore species: Eurasian lynx, gray wolf, wolverine, and brown bear. The study employs a two-pronged approach: 1) Matching methods to control for confounding factors and compare carnivore densities inside and outside PAs; and 2) Hurdle-mixed effects models to assess the influence of land protection on carnivore densities, along with other environmental covariates. For the matching analysis, one-to-one nearest-neighbor matching with replacement was employed using a generalized Mahalanobis distance metric and a caliper of 0.2 standard deviations of the propensity scores. The quality of matches was assessed by checking covariate balance. The absolute PA effect (the difference in densities between paired protected and unprotected units) was calculated at the national and regional levels, with iterative random sampling used to account for differences in the number of repeated observations. The hurdle-mixed effects models accounted for zero-inflation and right-skewed data using the GLMMadaptive package in R. These models assessed the relationship between carnivore density and several covariates, including land protection status, year, and other environmental factors, with interaction terms included to assess spatial and temporal variations in PA effectiveness. Model selection was based on the Akaike Information Criterion (AIC). Marginalized coefficients and their standard errors were extracted to account for the conditional interpretation of fixed effects in mixed models with nonlinear link functions. Model goodness-of-fit was assessed using residual diagnostics. Additional models were built to assess the potential confounding effect of PA size.

Matching analyses revealed no significant effect of PAs on the densities of three of the four carnivore species (lynx, wolf, and wolverine) at the national level. However, brown bear densities were significantly lower inside PAs than outside. No significant differences in densities were found between PAs and non-PAs for individual years for any species. Regional matching analyses did not reveal contrasted spatial patterns in PA performance. Hurdle-mixed effects models largely corroborated the matching results, showing no overall effect of protected areas on lynx, wolf, and wolverine densities. However, these models indicated higher brown bear densities inside PAs than outside—a contrasting result compared to the matching analysis. This discrepancy might be due to structural differences in the datasets used in both approaches and differing hierarchical scales. The hurdle-mixed models revealed species-specific spatiotemporal variations in PA effectiveness. An interaction between protection status and longitude was found for lynx, with higher densities outside PAs in western Finland but higher densities inside in eastern Finland. For wolverines, the hurdle-mixed model indicated a declining trend in densities within PAs over the study period, especially in Lapland, potentially due to increased poaching and human-wildlife conflict related to reindeer depredation. PA size did not significantly affect the densities of any of the four carnivore species.

The findings suggest a complex interplay of factors influencing the effectiveness of Finland's PA network for carnivore conservation. The lack of a consistent positive effect of PAs on carnivore densities does not necessarily imply ineffectiveness. Other factors, such as prey availability outside PAs or higher anthropogenic mortality inside PAs, may offset the potential benefits of protection. The contrasting results for brown bears between the two analytical approaches highlight the importance of considering both overall and site-specific effects. Regional and temporal variations in PA effectiveness underscore the need for context-specific conservation strategies. The declining wolverine densities within PAs in Lapland, despite the seemingly protective effect of the PAs, highlight the ongoing challenges of human-wildlife conflict in managing large carnivores in regions with extensive reindeer husbandry. This situation underlines the urgent need for improved co-management strategies to alleviate conflicts.

This study provides valuable insights into the complex relationship between protected areas and carnivore conservation in Finland. While the national PA network does not consistently support higher carnivore densities than unprotected habitats for many species, the results highlight crucial spatiotemporal interactions and the need for adaptive management strategies. Future research should focus on addressing human-wildlife conflicts, particularly in areas with high reindeer husbandry, improving enforcement of anti-poaching measures, and further investigating seasonal variations in carnivore habitat use within PAs. This research also makes a significant contribution by introducing and validating matching methods for wildlife conservation impact assessment, providing a valuable tool for improving future conservation planning and achieving global biodiversity goals.

The study acknowledges several limitations. The lack of difference between densities in PA and non-PA sites does not definitively indicate PA ineffectiveness without accounting for separate pressures. Benefits from higher habitat quality inside PAs could be counteracted by higher mortality. Differences in data collection seasons for bears (late summer) versus other carnivores (mid-winter) could influence findings. The use of indirect density indices (snow track counts) might introduce some uncertainty. Furthermore, the study mainly focused on Finland's protected area network, so results may not be directly generalizable to other regions with differing ecological contexts, human-wildlife conflict levels, and PA management practices.