The core of the “regional-landscape-site” cross-scale framework lies in its multi-scale systemic integration, overcoming the limitations of traditional single-scale conservation approaches to develop comprehensive solutions for climate change complexities (Figure 3). This framework emphasizes vertical interactions and interdependencies, responding to climate change uncertainties through both spatial cascades and policy implementation.

In terms of spatial scales, vertical integration manifests through a species-landscape-region planning hierarchy. While site-scale species protection yields local benefits, it faces challenges in achieving broader ecosystem functional adaptation goals. Conversely, landscape-scale approaches achieve functional integration through ecological networks (corridors/refugia), yet face land-use conflicts (Mendonça et al., 2021), governance fragmentation (Dorst et al., 2022), and multi-stakeholder coordination challenges (Kauark-Fontes et al., 2023). Large-scale interventions require coordination at a broader regional scale. Scientific understanding of cross-scale adaptation challenges helps avoid maladaptive practices (Schuldt et al., 2023).

The primary advantages of vertical interactions and inter-dependencies are reflected in the ecological linkages across spatial scales and the hierarchical transmission and information feedback mechanisms in implementation.

Ecological interdependencies enable cross-scale conservation coordination, forming functional ecological networks that enhance the overall efficacy of biodiversity adaptation strategies. For instance, at the regional scale, climate models can identify refugia, providing a scientific basis for the design of corridors at the landscape scale. In turn, the construction of ecological networks at the landscape scale creates migration pathways for species conservation at the site scale.

The critical gaps in biodiversity adaptation strategies across spatial scales are as follows:

Assessment of problem-governance scale alignment. Effective cross-scale management first requires a clear distinction between the spatio-temporal scale at which problems occur (problem scale) and the institutional scale at which governance is implemented (governance scale), and an assessment of the degree of alignment between the two (Padt et al., 2014). Many countries face horizontal mismatches between internal governance scales and the scale of climate change impacts. For example, there is inconsistency between the boundaries of river basins such as the Rhine, Meuse, or Scheldt (problem scale) and the administrative boundaries of member states (governance scale) in the European Union region (Dewulf et al., 2015).

According to the IPCC (2007), the vulnerability of species to climate change encompasses three dimensions: exposure, susceptibility, and adaptive capacity (IPCC, 2007). (i) Exposure refers to the degree to which species are exposed to significant climate change, such as the proportion of species in regions experiencing rapid climate change. Generally, the faster the rate of climate change and the greater its intensity and frequency, the higher the exposure of species in that area. (ii) Susceptibility indicates the degree to which species are affected by climate hazards, determined by their intrinsic biological characteristics and emphasizing the outcomes of climate change impacts, such as species extinction or reduced abundance due to climate influences. (iii) Adaptive capacity denotes the ability of organisms to adjust to, exploit, and respond to potential damages, opportunities, or consequences (McCarthy et al., 2001). The intersection of high susceptibility, high exposure, and low adaptive capacity represents the greatest vulnerability (Hole et al., 2011).

Understanding species vulnerability under a range of potential future scenarios is important due to the uncertainty in climate change projections and in species, ecosystem and human responses. On one hand, direct threats to biodiversity from climate change may include: changes in phenology, changes in species distribution shifts, community composition alterations, ecosystem function changes and loss of living space. On the other hand, human responses to climate change also have an impact on biodiversity, mainly in the agriculture, water, health and energy sectors. For example, upslope shifts in cultivation due to climate (Warner et al., 2009). Glacial retreat due to climate change is significantly reducing dry season flows in glacial rivers, prompting the construction of upstream reservoirs to ensure adequate flows for hydroelectric power generation and downstream agricultural needs, which may reduce the diversity and abundance of organisms along the way (Vergara et al., 2007). In addition, using biofuels as an alternative resource is seen as a way to reduce greenhouse gas emissions, while crop-based biofuel production leads to the conversion of rainforests, savannas, grasslands and other natural ecosystems to agricultural land, which generates significant carbon debt and causes widespread degradation of natural ecosystems. These indirect threats can be as serious as, or even exceed in scale and scop the direct threats, while further affecting policy feasibility (Fargione et al., 2008).

Assessing species vulnerability based on this understanding of vulnerability components is an important element of adaptation planning. For understanding and reconciling the expected interactions of the various elements, Williams et al. (2008) have integrated a working framework for species vulnerability assessments, which associates the interactions between vulnerability, exposure and adaptive capacity to guide biodiversity adaptation to climate change (Williams et al., 2008). Guided by the theoretical framework, identifying interactions between vulnerability, exposure and adaptive capacity and estimating vulnerability are the next important task (Pacifici et al., 2015). Based on the consideration of species distributional changes, population changes and extinction potential (Pacifici et al., 2015), methods for assessing species vulnerability are categorized into correlative, mechanistic, trait-based and combined approaches (Table 2).

(a)Correlative approaches are the most widely used, mainly based on observed species distributional changes in relation to climate change to predict the possible future suitable distribution areas of species. Species distribution models (SDMs), in particular, emphasize data-driven projections of potential distributional shifts (Jakubska-Busse et al., 2024). Employing an ensemble of SDMs under the RCP8.5 scenario, Dawe and Boutin (2016) projected that the climatically suitable habitat of the white-tailed deer (Odocoileus virginianus) would expand northward across North America by 2050 (Dawe and Boutin, 2016). Dynamic vegetation models (DVMs) simulate climate-driven vegetation succession (Heffernan et al., 2024). Yu et al. (2014) coupled the LPJ-GUESS model with outputs from 19 GCMs under RCP8.5 and demonstrated that evergreen broad-leaved forests are projected to replace deciduous forests in eastern China by 2100 (Yu et al., 2014). Climate-trajectory models highlight species’ capacity to persist and disperse by comparing future climatic analogues with current conditions. Ohlemüller et al. (2006) quantified the spatial extent of analogous and non-analogous climates across Europe to evaluate species’ adaptive capacity under climate change (Ohlemüller et al., 2006). While correlative approaches efficiently predict future habitat suitability, their reliability is contingent upon robust model selection and high-resolution climatic data (Elith and Leathwick, 2009) and may overlook biotic interactions influencing adaptive capacity (Meier et al., 2011).

(b)Mechanistic approaches focus on quantifying the probability of extinction under climate change by explicitly incorporating species-specific traits, physiological tolerances, and habitat interactions into viability assessments. Population viability analysis (PVA) exemplifies this strategy; it integrates physiological thresholds and the stochasticity of climatic events to estimate extinction risk, with particular emphasis on endangered taxa. Using PVA, Vargas et al. (2007) demonstrated that an increase in El Niño frequency elevates the probability of population collapse for the Galápagos penguin (Spheniscus mendiculus) to 78% within the next 100 years (Vargas et al., 2007). Although mechanistic approaches are powerful in integrating physiological and behavioral mechanisms underlying extinction risk, they demand extensive data and remain challenging to couple with non-climatic degradation (Foden et al., 2016).

(c)Trait-based assessment (TBA) refers to assessing the potential climate change impacts on a species by identifying the population, ecological niche and habitat characteristics of individual species through literature surveys, data compilation and expert consultation (Aguirre-Gutiérrez et al., 2025). A species is considered to have limited adaptive capacity if its habitat is within a restricted elevational range, if it has low genetic diversity or if it is dependent on only a few prey or host species. Applying this framework to the world’s avifauna, Foden et al. (2013) identified alpine endemics such as the Himalayan Snowcock (Tetraogallus himalayensis) as highly vulnerable because of their restricted altitudinal range and weak dispersal ability (Foden et al., 2013). Although TBA translates functional attributes into quantitative vulnerability scores, the weighting of individual traits requires expert calibration to minimise subjectivity.

(d)Combined approaches integrate correlative, mechanistic and TBA in a complementary manner to meet empirical needs. Pearson et al. (2014) coupled SDMs with PVA and demonstrated that dispersal barriers can trigger local extinctions of European amphibians even within climatically suitable habitats; such integration reduces predictive uncertainty, yet it demands extensive cross-disciplinary data support (Pearson et al., 2014).

Monitoring serves as a critical component of adaptive planning implementation, providing early warnings for emergent climate risks and establishing empirical bases for conservation effectiveness evaluation, with its outcomes requiring real-time integration into the planning revision process (Corelli et al., 2024). Specifically, monitoring objectives can be operationalized through three key dimensions: (i) Understanding how ecosystems, habitats, and species respond to climate change while identifying compounding stressors that may exacerbate these responses; (ii) Generating data for model development and validation to enhance predictive capacity for climate adaptation scenarios; (iii) Assessing the effectiveness of policy and management interventions (Bongaarts, 2019). As a representative case study, the European Biodiversity Observation Network (EU BON) project (2012–2017) established a continent-scale monitoring infrastructure through standardized permanent plots (accessible via https://monitoring.europabon.org), systematically recording species abundance, phenology, and microclimate data to analyze climate-land use interactions. The longitudinal datasets enabled refinement of species distribution models and evaluation of Natura 2000 PAs’ capacity to accommodate climate-induced species range shifts (EU BON, 2017).

Operationalizing sustained monitoring requires the iterative execution of three core tasks: (i) Standardization of indicator systems. Standardized monitoring based on the Essential Biodiversity Variables (EBVs) enables comparable metrics across genetic, species, and ecosystem levels, facilitating cross-project data integration and trend analysis (Geijzendorffer et al., 2016). For instance, the SoilBON network employs microbial diversity and soil organic carbon EBVs to assess policy impacts on subsurface biota, with springtail (Collembola) abundance serving as a rapid indicator of soil quality (Guerra et al., 2021). Likewise, the Global Coral Reef Monitoring Network adopts hard coral cover, macroalgal canopy cover, and fish diversity and abundance as three robust EBVs for reef health assessment (Obura et al., 2019). (ii) Multi-scale data integration and model validation. Species, ecosystem, climatic and remotely sensed data collected at nested scales are assimilated into analytical models to enhance the resolution of interaction effects (Jetz et al., 2019). Rogers et al. (2025) integrated multi-scale data via zero-inflated Bayesian regression to quantify the joint influence of climate and land use on freshwater fish assemblages in the northeastern United States (Rogers et al., 2025). (iii) Data sharing and attribution analysis. Scientifically rigorous designs facilitate seamless data exchange, enabling the timely detection of natural trends and extreme events while disentangling causal pathways. The Colorado Parks and Wildlife agency exemplifies this approach through the Colorado Beaver Activity Mapper, which fuses: GPS-mapped beaver dam, citizen science activity reports submitted via the Engage CPW platform, and ecological process data from interagency portals to identify the drivers of beaver–human conflict dynamics (CPW, 2025; Longwell, 2025).

Moreover, the protracted nature of climate change and the persistence of statistical noise necessitate long-term monitoring programs (Leung and Gonzalez, 2024). Most environmental time series require extended periods before underlying trends or variable relationships achieve statistical significance. Many ecological responses—including species migration and community succession—as well as cyclical climatic phenomena such as the El Niño, unfold over decades. For instance, the long-lived trees and limited seed dispersal mean that forest communities may require centuries to complete demographic turnover, whereas contemporary anthropogenic warming exceeds historical natural variability, resulting in marked lags between climatic forcing and forest response (Fastovich et al., 2025). Consequently, the informational value of biodiversity data increases exponentially with the length of time series (Robinson et al., 2005). Nevertheless, sustained biodiversity monitoring has been documented as a high-cost endeavor, with cumulative expenditures reaching the millions to billions of dollars—an issue repeatedly identified as a critical financing challenge in recent international assessments (CBD, 2024; NPWS, 2021; UNEP-MAP, 2022).

In most cases, continuous and intact native habitats are the best solution for biodiversity conservation. However, in reality, a large number of economically or socially significant land-use types have separated PAs into ecological islands. Reconfiguring PA networks can help populations move along ecological corridors and increase population size, thus increasing adaptive resilience and improving resistance to climate change impacts. Corridors, stepping stones and matrices collectively transform scattered PAs into an interconnected network that retains a natural vegetation-like structure, forming PA networks. Linear corridors directly connect PAs through habitat patches of habitat, facilitating species dispersal (Stralberg et al., 2020b). For example, the tri-national Great Limpopo Transfrontier Conservation Area (GLTFCA) elephant-movement corridor network, completed in 2024 by South Africa, Mozambique, and Zimbabwe, provides approximately 15,000 African elephants and wildebeest with 3,500 km² of continuous dry-season migration habitat (Bakari, 2025). Appropriately sized stepping stone patches shorten the spatial distance between suitable habitats and lower the energetic cost of cross-landscape movement (Schüßler et al., 2020). In the Atlantic Forest of Brazil, the Portal de Paranapanema restoration project established 90 agroforestry stepping stones (5–20 ha each) that reconnected forest fragments, restoring 1,800 ha of contiguous forest within a decade (Hilty et al., 2020). As the largest, most homogeneous, and most connected component of the landscape, the matrix plays a pivotal role in mitigating edge effects when its ecological quality is improved (Ruffell et al., 2017). Colombia’s Caribbean Ecological Connectivity Initiative converted 1.5 million ha of agricultural matrix (oil palm, pasture, and urban zones) into multifunctional sustainable-use zones through zonation, integrating them with core reserves and corridors to prevent protected-area isolation (FAO and UNEP, 2020).

Corridors and stepping stones constitute essential supplements to the matrix, offering structural guarantees for species movement and the continuity of key ecological processes. However, it must be noted that:

Corridor effectiveness is highly contingent upon the dispersal capacity of the focal taxa: volant birds may benefit, whereas amphibians and invertebrates often fail to use corridors that are too narrow or environmentally unsuitable. Lynch (2019) observed that contemporary urban greenway designs disproportionately cater to mammals and birds, neglecting the requirements of low-mobility species (Lynch, 2019). Moreover, Linear corridors are prone to induce edge effects, leading to abrupt changes in microenvironment parameters such as light and wind speed, and may become a diffusion pathway for invasive species (Bennett and Bennett, 2003). In experimental corridors at Savannah River Site, South Carolina, increased edge illumination significantly elevated densities of the invasive fire ant (Solenopsis invicta), resulting in a marked decline in native ant diversity (Resasco et al., 2023).

Research and practice in restoring landscape connectivity can be divided into two categories: focal species-based and network structure-based approaches.

Focal species-based connectivity, the more traditional approach, involves planning by predicting species’ future distribution based on their exposure, sensitivity, and resilience under climate change (Krosby et al., 2015). Key applications includ: (i) finding habitats or corridors that will remain valuable for certain priority species even under climate change (Fan et al., 2017); (ii) predicting species distributions based on climate change, and thus determining how to protected area or promote landscape connectivity (Choe et al., 2017). Published researches have also shown passionate concerns about ecological corridor planning for flagship species such as Asian elephants and giant pandas (Mandal and Das Chatterjee, 2023).

Restoring landscape connectivity based on network structures has been incorporated into biodiversity conservation in multiple countries. Since 1992, to ensure ecological connectivity and habitat quality, The EU’s Natura 2000 network (established in 1992) covers nearly 28,000 sites (18% of land area), which is at the heart of the EU’s ecological conservation and climate change adaptation program (BISE, 2022). Mitigation banks in the U.A. adopt a range of strategies to conserve, manage and restore degraded habitats, connect fragmented habitats, create buffers and habitats for adaptive conservation of biodiversity (EPA, 2002). The Cape Floristic region of South Africa has a protected area plan that incorporates river corridors across mountains (Pressey et al., 2007). Australia has initiated large-scale landscape restoration and connectivity projects to combat climate change (Taylor and Figgis, 2007). In addition, countries have been developing green infrastructures to restore the connectivity between cities and nature. Green infrastructure is defined as an interconnected network of natural areas (e.g., rivers, wetlands, forests and wildlife habitats), and human-made environments (e.g., green spaces, parks, farmlands and pastures) (Canzonieri et al., 2007), the connectivity of which is essential for the survival of natural species, air and water quality, and human health and quality of life. In 2022, the EU launched biodiversity strategy for 2030, which designated “the improvement and restoration of ecosystems and their services through the development of green infrastructure” as one of its six headline targets (EU, 2022).

Escalating frequency, intensity, and duration of extreme climatic events impose substantial physiological and demographic stress on biota (Murali et al., 2023). Climate change refugia emerge as pivotal sanctuaries for species persistence and population recovery. Unlike landscape connectivity initiatives that design migration pathways based on species traits and dispersal capacity, refugia-oriented strategies focus on safeguarding residual populations to maintain genetic diversity (Ackerly et al., 2020). Refugia are defined as spatial habitats into which populations contract when confronted with climatic stress, providing critical buffering when necessary (Morelli et al., 2020). For taxa constrained within such refugia, migration or dispersal to more suitable habitats is often precluded by intrinsic or extrinsic barriers (Poulos et al., 2013). Consequently, facilitating population persistence within refugia is pivotal for both post-extinction recolonization and long-term adaptation under climate change, rendering the identification of refugia a priority for biodiversity conservation planning.

Refugia are species- and stressor-specific, shaped by the dynamic interplay between stressors and organisms (Greiser et al., 2020; Stewart, 2010).

Climate change as the dominant stressor. The velocity and magnitude of climatic shifts constitute the primary criteria for refugium identification (Szcodronski et al., 2024). If global warming is constrained to 2 °C, integrating climate-change refugia into an expanded protected-area network remains feasible (Saunders et al., 2023). For example, the U.S. National Park Service designated the meadow complex of Devils Postpile National Monument as a climate-change refugium and implemented invasive-tree removal to preserve its ecological function (Morelli et al., 2016). When warming exceeds 2 °C, however, most refugia will be restricted to high latitudes and elevations (Lawler et al., 2020).

Technically, climate change refuge identification depends on the accumulation of data across multiple species, such as relying on phylogenetic comparisons (Keppel et al., 2018), phylogenetic diversity metrics (Costion et al., 2015) and phylogenetic geographical analysis (Liu et al., 2025) to collect genomic data of certain species. Based on the collected species data, SDM is used for large-scale identification. Specifically, the habitat requirements, population dynamics and dispersal of a species are incorporated into bioclimatic models to predict the potential future distribution of a species at local, regional or larger spatial scales, and to identify refuge locations based on current distribution versus modelled future distribution projections (Briscoe et al., 2016). However, refugium identification confronts significant methodological, data and ecological challenges due to the complex interactions among diverse stressors, landscape contexts and refugial taxa.

Uncertainty in climate models. SDMs serve as a critical tool for identifying refugia (Georges et al., 2024), yet most models primarily correlate species distributions with macroclimatic variables, failing to comprehensively incorporate key ecological factors such as soil properties, vegetation structure, hydrology, land use, invasive species, and behavioral buffering. When projecting future ranges and identifying refugia, due to truncating all acceptable conditions for species, it may encounter niche truncation challenges (Anselmetto et al., 2025), which may either over or underestimate the true location of refugia. Furthermore, SDMs exhibit delayed responses to dynamic changes. Most current approaches rely on steady-state climate assumptions, rendering them inadequate for capturing the immediate impacts of extreme events (e.g., wildfires, droughts) on refugia. For instance, studies on post-glacial oak refugia demonstrate that species migration is strongly climate-driven (Hao et al., 2023), yet existing models struggle to integrate abrupt disturbances (e.g., flash droughts) that may disrupt refugial habitats.

Spatial blind spots emerge from reliance on high-resolution environmental data. Refugia identification proves particularly sensitive to spatial resolution and thermal buffering thresholds. Coarse-scale global climate datasets (e.g., 1-km resolution) systematically underestimate fine-scale temperature variations in topographically complex terrain, causing omission of critical microrefugia (e.g., ravines, caves) (Rosauer et al., 2013). Furthermore, the paucity of high-resolution environmental data in tropical, deep-sea, and polar regions biases identification efforts toward well-studied areas, leaving many potential refugia undetected (Morelli et al., 2020).

Stressor-species interaction gaps. Most studies focus on flagship species while neglecting community-level interaction dynamics, including competitive interactions among plant species, plant–animal relationships (herbivory, pollination, seed dispersal) (Ackerly et al., 2020); anthropogenic pressures (e.g., trawling impacts on marine refugia, deep-sea mining effects) (Zelli et al., 2025). Climate-human pressure interactions may drive species reassembly and loss, potentially altering ecosystem functioning and potentially unbalancing refugium network design (González-Trujillo et al., 2024).

For keystone species, in situ conservation is one of the most effective ways to reduce the biodiversity loss on a global scale. The specific implementation is divided into three steps:

Using data on species richness, endemism, species threatened, taxonomic distinctiveness and habitat uniqueness to varying degrees to identify keystone species based on vulnerability and irreplaceability (Cottee-Jones and Whittaker, 2012).

Thanks to international efforts, 14% of forests (Schmitt et al., 2009) and 88% of vertebrate species (Rodrigues et al., 2004) have been protected in 34 biodiversity hotspots around the world, and in situ conservation of species has been actively promoted. However, half of the world’s KBAs are still unprotected (Butchart et al., 2012). In addition, in situ conservation on the site scale has difficulty maintaining good ecological processes because of inadequate management effectiveness and the isolation of PAs from each other, and the loss of biodiversity at the population level is often prone to far-reaching ecological and evolutionary consequences, such as the loss of top predators. Under land use pressure, in situ conservation faces the enormous challenges of feeding 9 billion people by 2050 and biodiversity conservation.

Endangered species are more susceptible to climate change and there are climate thresholds beyond which the probability of extinction increases dramatically, thus making ex situ conservation an important initiative for the adaptive management of endangered species (Solomon et al., 2007). Ex situ conservation refers to the relocation of plants, animals and other organisms from areas that have become unsuitable to other areas that are suitable for survival in form of assisted dispersal (Liu et al., 2014), assisted migration (Guinan et al., 2025) and assisted colonization (Gallagher et al., 2015). Ex situ conservation requires three criteria to be met: (i) Seed to seed, requiring relocated plants and animals to be able to grow freely and survive through sexual reproduction; (ii) Representation requires that relocated species maintain genetic integrity and represent the genetic diversity of the population; (iii) Maintaining the population gene frequencies of genes after translocation to avoid outbreeding depression, genetic assimilation and intragression (Engelmann and Engels, 2002; Quinlan et al., 2025). Targets 12 and 13 of the Aichi Biodiversity Targets call for the conservation of biological genetic diversity through ex situ conservation projects (Geijzendorffer et al., 2017), and countries around the world are actively exploring ex situ conservation systems, which have been extended to cover wildlife, crops, domesticated animals and microbial strains.

While significant achievements have been made in ex situ conservation, it is crucial to recognize its distinctive characteristics. (i) Unlike climate change refugia that emphasize population relocation and long-term retention in natural systems, ex situ conservation focuses on species under human intervention. For example, wild plant conservation primarily utilizes botanical gardens and seed banks; food and agricultural plant preservation employs germplasm repositories and nurseries; wild animal conservation relies on zoos, safari parks, aquaria, and other captive breeding programs; while microbial conservation is accomplished through strain conservation (FAO, 2007). (ii) Ex situ conservation attempts may fail and even accelerate species extinction. A tragic example is the 1981 capture and captive breeding of Japan’s last five crested ibises (Nipponia nippon), which ultimately failed to prevent the species’ extinction in 2003 (Biodiversity Center of Japan, 2024). (iii) Under climate change scenarios, ecosystem transformations may become so profound that species reintroduction becomes unfeasible, potentially turning ex situ conserved populations into evolutionary relicts (Minteer, 2014). (iv) Technical constraints include: (a) climate model uncertainty and error may cause mis-translocation risk, as evidenced by the Poweshiek skipperling (Oarisma poweshiek) conservation program in the U. S., where climate model biases led to dramatic post-release survival declines (Runquist et al., 2025); (b) small population sampling reduces genetic diversity (Mclachlan et al., 2007). A global meta-analysis confirmed that inadequate wild population sampling resulted in significantly lower genetic diversity in restored populations compared to reference groups over 50 years, impairing long-term adaptive capacity (Wei et al., 2023); (c) microhabitat mismatches increase post-release mortality; and (d) translocated individuals can introduce or encounter novel pathogens. In Canada, the ex situ conservation of whitebark pine (Pinus albicaulis) faced dual threats: the absence of Clark’s nutcracker (Nucifraga columbiana), its natural seed disperser in native habitats, and the concurrent spread of white pine blister rust (Cronartium ribicola) to recipient sites (Sáenz-Romero et al., 2021).