Current national nature reserves are insufficient to safeguard the long-term survival of birds and mammals in China

Protected areas are central to biodiversity conservation, but their effectiveness depends on design, location, management, and functional connectivity. Many assessments emphasize area, representation, and governance while assuming high habitat quality within protected areas. However, animal movement and population persistence depend critically on functional connectivity, which can be reduced by natural heterogeneity and especially by human-driven habitat loss and fragmentation. Given global and national policy calls (e.g., Kunming-Montreal Global Biodiversity Framework) to enhance connectivity, there remains a lack of quantitative evaluation at large scales of how protected areas sustain viable wildlife populations. This study evaluates how China’s National Nature Reserves (NNRs) affect long-term survival of birds and terrestrial mammals by modeling functional connectivity within individual NNRs. It asks: (1) How effective are NNRs at conserving internal populations for long-term survival? (2) What natural and anthropogenic factors regulate NNR effectiveness? (3) To what extent do artificial landscapes (e.g., roads, built-up land, agriculture) modify NNR effectiveness?

Prior work has often focused on protected area extent, representativeness, and governance, with fewer studies explicitly quantifying functional connectivity and its implications for population viability at regional to national scales. Landscape fragmentation from infrastructure and human land use is known to reduce gene flow and increase extinction risk. Several studies modeled connectivity within PA networks and individual PAs, generally advocating more or new PAs to enhance connectivity, but often at single spatial scales and with limited taxonomic breadth or comparability. Roads and linear infrastructure are a major threat to connectivity, even within PAs. Biases in PA placement can overrepresent remote, less-threatened areas. Consequently, there is a need for multi-scale, multi-taxa approaches integrating habitat availability, ecosystem diversity, and human pressure to assess functional connectivity and conservation effectiveness.

The study modeled functional connectivity for wildlife populations in all 474 National Nature Reserves (NNRs) in mainland China (excluding Taiwan, Hong Kong, and Macau). A graph-theoretic approach implemented in Graphab was used, where habitat patches are nodes and least-cost paths between patches define potential links. The analysis required (1) a landscape resistance surface and (2) species movement ability. Resistance surfaces combined land cover (ChinaCover 2010 updated to 2020 at 30 m resolution via Landsat interpretation and OSM overlays for roads, railways, waterways, built-up and farmland), slope (ASTER GDEM v2, slope-weighting with coefficient c=10), and human activity. High resistance values were assigned to artificial hard-surface landscapes, especially major roads, highways, and railways, for mammals and birds, to reflect substantial barriers and additional impacts (traffic, pollution, noise, poaching). Movement abilities were standardized using six virtual species groups representing birds (broadleaf forest, coniferous forest, wetlands: BBF, BCF, BW) and terrestrial mammals (broadleaf forest, coniferous forest, grassland: MBF, MCF, MG). Two movement categories were used: low motility (median dispersal ≈10 km) and high motility (≈50 km), based on the 25th and 75th percentiles of minimum enclosing circle diameters of NNR boundaries and validated against literature on real species dispersal. Allometric relationships linked dispersal distances and diet (herbivore vs carnivore proxies) to daily movement distances, key patch areas for meta-populations, and isolated population range areas. Functional connectivity was quantified using the Probability of Connectivity (PC) index, which estimates the probability that two randomly chosen points in the study area are connected via habitat. Sensitivity of functional connectivity (SFC) to artificial landscape removal was assessed as the relative gain in PC after removing artificial landscapes. NNR effectiveness was defined as the ability to sustain the minimum meta-population size or maintain adequate inter-population genetic exchange given species’ ecological traits and movement. For low-mobility species, effectiveness required sufficient key patch area (MinAk) within an NNR; for high-mobility species whose ranges exceed NNR boundaries, a criterion based on connected habitat patches within daily movement distances and isolated population area thresholds (Ag) was used. Proportion of NNRs effectively protecting populations was computed per NNR and aggregated to biome and national scales. Predictors of functional connectivity included proportion of habitat area (HA), artificial landscape intensity (ALI), NNR area (AREA), and ecosystem diversity (ED). Generalized Additive Models (GAMs, negative binomial, mgcv) with interactions and restricted maximum likelihood were used to model nonlinear relationships, with model selection via AIC. Random forest with tenfold cross-validation (rfPermute) assessed variable importance and regional contrasts (east vs near vs west of the Hu-line). Human footprint datasets (HumanFootprint v2, Global Human Modification) were used to analyze anthropogenic pressure patterns and relate them to effectiveness. Sensitivity analyses confirmed robustness of connectivity and effectiveness results across alternative dispersal distances and diet assignments (high correlations, R² > 0.97 for connectivity; R² > 0.85 for effectiveness).

- Overall effectiveness: NNRs effectively protected about 57% of bird populations and 42% of terrestrial mammal populations for long-term survival; combined, roughly 50% across birds and mammals. - High-mobility species: Only <25% of bird populations and 13% of terrestrial mammal populations with high motility were effectively protected. - Spatial patterns: Eastern China’s PAs provide poor protection for high-mobility birds and mammals. For high-mobility species, effective NNRs had, on average, 5.6 times (birds) and 13.5 times (mammals) larger habitat area than ineffective NNRs. About 48% of NNRs had >50% habitat area (HA) but were still too small to support viable populations. - Biome differences: Temperate coniferous forests and temperate broadleaf and mixed forests had the lowest effectiveness for high-mobility species (10.8% and 15.9%). These biomes have <3% NNR coverage in China (vs global averages of ~10% and ~16%), very small average NNR sizes (<4% of China’s national average PA size), high natural vegetation conversion rates, and coincide with densely populated, economically developed regions. - Drivers of connectivity: Random forest identified proportion of habitat area (HA) as the most important predictor of functional connectivity. Effects varied geographically: ALI dominated in NNRs east of the Hu-line; NNR area constrained connectivity near the Hu-line; ecosystem diversity (ED) was more influential in the west. GAMs showed connectivity increases with HA, declines with ALI, exhibits a quadratic peak at medium ED, and has a size-dependent AREA effect (decreasing for small PAs, increasing for large PAs), with stronger AREA effects at higher HA and ED. - Human activities: Artificial landscapes comprise <2% of total NNR area yet account for nearly 40% of functional connectivity loss for highly mobile terrestrial mammals. NNR effectiveness declines with increasing human footprint; for high-mobility taxa, effectiveness can reach ~50% when HF < 4 but drops below 30% under higher human activity intensity. - Taxonomic differences: Birds are generally less vulnerable than mammals to road-induced fragmentation at the same dispersal ability, potentially due to greater reliance on flight and differing resource requirements, though birds are not necessarily at lower extinction risk overall. Agricultural areas within PAs may provide resources for some birds, but associated infrastructure and disturbances undermine conservation outcomes.

The findings show that many NNRs, particularly smaller ones and those under higher human pressure, are insufficient to ensure long-term population survival, especially for high-mobility and large-bodied species. Functional connectivity within PAs is as critical as connectivity among PAs. The overlap between small PA size, high human activity, and low effectiveness indicates that size constraints and anthropogenic fragmentation jointly limit conservation outcomes. The study informs the debate on single large versus several small protected areas by emphasizing population-level viability and connectivity requirements rather than species richness alone. Multiple interacting drivers—habitat amount, artificial landscape intensity, ecosystem diversity, and PA size—shape connectivity, with region-specific dominant constraints across the Hu-line. Infrastructure notably offsets conservation gains; in high human footprint contexts, reducing artificial landscape intensity can yield larger connectivity benefits than merely increasing habitat proportion. Management should adopt common but differentiated strategies across regions, prioritize increasing habitat diversity to ensure key resources (e.g., water, wetlands), reduce human activities within PAs, optimize and aggregate small units (e.g., via national parks), and improve inter-PA connectivity through corridors and networks to bolster resilience under climate change and ongoing development.

China’s NNR network currently secures long-term survival for roughly half of modeled bird and mammal populations, but performs poorly for high-mobility species. Small PA size, low habitat coverage, and high artificial landscape intensity substantially diminish functional connectivity, with artificial landscapes causing disproportionate connectivity loss. Maintaining high functional connectivity within PAs is as important as enhancing connectivity across national and global PA networks. The study provides a multi-scale, multi-taxa framework to evaluate PA effectiveness based on population viability and connectivity. Policy and management should prioritize: (1) reducing artificial landscape intensity within PAs, especially in high human footprint regions; (2) increasing habitat amount and diversity; (3) optimizing PA configurations by integrating small units and establishing larger reserves where feasible; and (4) linking PAs via ecological corridors and cross-regional networks to enhance gene flow and adaptability. Future research and practice should refine species-specific movement and demographic data, better map and mitigate infrastructure impacts, and incorporate socio-ecological heterogeneity into targeted, region-specific conservation strategies, including complementary biodiversity measures in farmlands outside NNRs.

- Species representation: The study uses virtual species groups based on habitat preference and dispersal traits, validated against literature and selected real species, but does not model all species-specific behaviors or demographics. - Data constraints: Road and infrastructure data from OSM may underestimate true extent and barrier effects; fences and right-of-way widths may be greater than mapped. Human footprint datasets are from specific years (HF v2, 2009; GHM v1, 2016), potentially mismatching current conditions. - Resistance parameterization: High resistance values for artificial landscapes and slope weighting (c=10) are assumptions that, while conservative, may not capture all species’ responses. - Scope: Taiwan, Hong Kong, and Macau are excluded. Marine-focused NNRs were treated only in single-PA effectiveness tallies and excluded from biome/national aggregations. Some small, linear watershed NNRs required 10 km buffers, potentially affecting area estimates. - Scale emphasis: Analyses focus on movements within single NNRs and do not fully capture broader metapopulation dynamics across multiple PAs beyond the defined criteria.