Location-Based Damage Factors to Assess the Impact of Future Land-Use Changes on the Biodiversity Extinction Rate

Abstract

Rising warnings of a potential sixth mass extinction and ecosystem degradation, fueled by climate change and human activities, highlight the need for an accurate measurement of human impacts on biodiversity loss to support effective policymaking. This study incorporates the extinction rate metric with the life-cycle impact assessment (LCIA) model to develop a set of policy-relevant, closed-loop spatiotemporal damage factors (DFs). Here, biodiversity damage is expressed in E/MSY (expected extinctions per million species years), representing extinction rate-based impacts and intrinsic biodiversity value. This approach enabled quantification of the biodiversity extinction rate due to future land use changes. The total net extinction rate was estimated at 161.9 E/MSY, and the global weighted average DF was estimated at 1.73 × 10^–05 E/MSY/km² as a species richness- and area-weighted summary of spatially heterogeneous damage. The grid-scaled DF results were provided at a global resolution of 0.25 arc-degrees. DFs indicated marginal extinctions from land use changes within each grid cell, providing a means to customize the disaggregated-scale DF for various LCA inventories. These results improve the interpretability of location-based biodiversity assessments, making them more relevant to policy decisions.

1. Introduction

Biodiversity is vital for maintaining ecological stability, supporting economic activity, and providing essential ecosystem services. Currently, 0.01%–0.1% of species undergo extinction annually. The latest report from the World Wide Fund for Nature (WWF) revealed that, from 1970 to 2020, the Living Planet Index (LPI), an early indicator of ecosystem health, highlights increasing wildlife extinction and ecosystem degradation.

Existing policy-guiding frameworks, such as the planetary boundaries (PBs), are designed to guide the systematic, low-entropy progression of human activities and facilitate the restoration of natural ecosystems toward the “Holocene-state”. The extinction rate (measured in extinctions per million species per year, E/MSY) is a key metric for evaluating biodiversity extinction speed and offers notable advantages for operational feasibility and policy standardization. , Effective ecosystem management can help reduce anticipated extinction rates to targets recommended by PBs (<10 E/MSY) and to aspirational goals (∼1 E/MSY). Among the direct and indirect drivers of global biodiversity loss, land-use change is recognized as a key driver of habitat change and fragmentation. Anthropogenic-driven land conversions accelerate habitat and biodiversity loss, underscoring the need to clarify the mechanisms linking drivers to accelerated extinction rates.

Accordingly, various methodologies have been developed to quantify how drivers such as climate change and land use affect biodiversity loss. The Species Threat Abatement and Restoration (STAR) metric, which helps quantify the contributions of threat reduction and habitat restoration to global species conservation targets, is notable in this regard. However, STAR has limitations, particularly in its taxonomic coverage and its capacity to fully characterize the anthropogenic impacts on biodiversity loss.

Life cycle assessment (LCA) is a standardized method for quantifying the potential environmental impacts of products, processes, or services throughout their life cycles. Within LCA, life cycle impact assessment (LCIA) translates inventory flows into potential environmental indicators, including those related to ecosystem quality and human health. Recent advances include biodiversity-related LCIA indicators. In ReCiPe 2016, biodiversity impacts are expressed as species loss at the end point level, while midpoint indicators include land use and climate change-related pressures. The Potentially Disappeared Fraction (PDF) has been widely used in other LCIA frameworks such as LC-IMPACT to quantify potential biodiversity loss associated with anthropogenic activities. The Global Extinction Probability (GEP) method refines this by scaling regional extirpations to potential global extinctions. However, existing PDF-based indicators mainly describe biodiversity loss as the species disappearance. In contrast, extinction rate indicators reflect how quickly biodiversity is changing over time. Therefore, extinction rates (E/MSY) can provide additional information, especially for evaluating biodiversity changes against ecological limits and policy targets. However, biodiversity damage factors that directly quantify extinction rates are still largely missing from current LCIA frameworks.

The mean species abundance (MSA) index from GLOBIO, as recommended by the Convention on Biological Diversity, estimates biodiversity loss in terms of habitat intactness and describes ecosystem-level diversity. The LIME3 model is widely used in LCIA and enables translating biodiversity losses into economic terms, supporting initiatives like the Taskforce on Nature-related Financial Disclosures (TNFD). In LIME3, extinction risk (ER) estimates biodiversity loss and has been mapped under scenarios involving climate change and land-use impacts and is defined in terms of the number of extinct species, providing a deeper understanding of the loss. − However, these indicators do not explicitly quantify the extinction rates.

Thus, the evolving utilization of LCIA requires clarification of indicator definitions and applications in three key areas. (1) When evaluating biodiversity, the specific aspect of diversity being measured should be clearly specified to ensure consistency in interpretation and policy application. (2) Clear distinctions between extinction and disappearance are important, as species can exhibit recovery within ecosystems, and disappearance generally indicates functional degradation rather than complete extinction. (3) Indicators should be aligned with policy needs, for example, the extinction rate serves as a metric for evaluating biodiversity change against policy targets.

Additional data are therefore required to refine policy formulation, underscoring the importance of standardized, policy-relevant indicators. The rapid growth of frameworks, such as the TNFD, has increased the demand for biodiversity indicators that support both assessment and decision making.

In response to these needs, our research aims to develop a grid-cell-based damage factor that incorporates the extinction rate as a metric within the LIME framework. The new damage factor quantifies marginal increases in extinction rate attributable to land-use changes driven by human activity. Thus, it offers a more precise definition of biodiversity extinction and improved interpretability, enabling businesses and governments to better understand how local species values relate to land use practices. Building upon the existing LIME3 model, our approach achieves grid-scale visualization and upscaling, facilitating spatially explicit assessments that inform site-specific conservation planning. Regional vision supports flexible integration of primary and secondary inventory data in LCA, addressing current challenges in aligning damage factors with available inventory data sets. By incorporation of extinction rates into the LIME3 modeling framework, results can be directly evaluated against established policy thresholds, thereby strengthening the linkage between biodiversity assessment and conservation action.

2. Materials and Methods

In this study, the LCIA damage factor was defined as the potential extinction rate per unit area of land-use change (e.g., per km²) and calculated at the grid-cell level based on the land-use change area, particularly for additional land-use increments between land-use states. This formulation links the damage factor to the functional unit (FU), which is quantified as the total area of land use change associated with a specific product or activity. The method involved three steps: (1) estimating grid cell extinction rates due to future land use changes, (2) preparing a land use change map at the same resolution, and (3) generating a 0.25 arc-degree biodiversity damage factor map.

2.1. Extinction Rate Caused by Land Use Change

The extinction risk used to calculate the extinction rate in this study was based on previous global assessments covering 6,569 species (birds: 3791, reptiles: 98, mammals: 937, plants: 1296, amphibians: 447). The models estimated how future land use changes would affect extinction rates, assuming climate impacts remained unchanged. We convert extinction risk calculations from Tang et al. and Liu et al. into standardized extinction rate indicators, moving beyond traditional damage factors as defined by LIME3. Then develop grid cell-based damage factors at 0.25° resolution. According to LIME3, extinction risk (ER) for each species (i) was defined as the inverse of extinction time (year^–1), enabling direct calculation of extinction rates. Species-specific lifespans were estimated using methodology described in Tang et al., which predicts the number of years until a species’ distribution area falls below a defined threshold, assuming a constant rate of decline over the assessment period. As this approach extrapolates the observed rate of habitat change to the point where the species reaches a critical threshold rather than complete disappearance, the estimated lifespans can exceed the 85-year assessment period.

To predict these trends, we used WorldClim v2 climate data, LUH2 land use data for 2015 and projected 2100 (SSP1-RCP2.6), and GBIF species occurrence records. We modeled potential habitats with MaxEnt, , incorporating environmental variables and species-specific migration distances. To isolate the impact of land-use change, climate variables were fixed at the baseline level (2015), while land-use variables were updated for the future scenario (2100). We ran simulations on a supercomputer to predict the habitats for all 6,569 species under current and future land-use scenarios. , The ER for each species (i) was distributed across changing habitat pixels (c), following the spatial allocation formulation described in the Supporting Information, producing spatial distribution maps of ER for each species (ER _c,i , year^–1). Subsequently, we aggregated the spatial extinction risk values across taxonomic groupsReptilia, Aves, Plantae, Amphibia, and Mammaliacalculating the spatial ER for each group (ER _c,g) by combining results from all modeled species within that group, as represented by eq

$${ER}_{c,g}=\frac{1}{S_{g,\mathrm{mod}el}}\sum _{i\in g}{ER}_{c,i}$$ 1

where S _{g,mode l} denotes the number of species in modeled group g.

To extrapolate the ER from the assessed species to the global species pool, we calculated taxonomic group-level extinction risk values. The ER values were first calculated from our 6,569 modeled species (birds: 3,791, reptiles: 98, mammals: 937, plants: 1,296, and amphibians: 447). This extrapolation addressed the taxonomic imbalance in our data set, where plants comprised over 99% of IUCN recorded species (425,321 out of 464,123) but represented only 20% of the assessed species (1,296 out of 6,569). Scaling to the full species pool helped capture the overall contribution of plant diversity to global ERs. We assumed that extinction patterns within taxonomic groups were generally comparable due to shared ecological traits, allowing the diverse habitat- and threat-level sampling to represent broader patterns. This extrapolation does not imply that extinction risk is absent for nonmodeled species.

The ER represents the number of extinctions per species per year; it is therefore a per-species extinction intensity that can be expressed as an extinction rate by multiplying by one million, consistent with the standard definition (E/MSY). To address the scale difference between the modeled species and the broader global species pool, we extrapolated the taxonomic group-level ER derived from the modeled species to the corresponding global species pool within each group. Here we considered the world total of described species (S _global ; 464,123) from the IUCN Red List record. The number of species in each taxonomic group is denoted as S _g,global (birds: 11,195, reptiles: 12,162, mammals: 6,701, plants: 425,321, amphibians: 8,744). This extrapolation assumes that the 6,569 modeled species (S _{g,mode l} ) are representative of extinction patterns within their taxonomic groups. The calculated extinction rates were then mapped spatially based on taxonomic distribution patterns. Therefore, when computing the overall distribution of global extinction rates (Extinction Rate _c,global ) was then calculated as a taxonomic group-weighted average and expressed in E/MSY by multiplying by one million, as shown in Equation ). The extinction rate was calculated by aggregating taxonomic group–level ER values and weighting them by the proportion of global species richness represented by each taxonomic group.

$${ExtinctionRate}_{c,global}=(\sum _{g=1}^5{ER}_{c,g}\times \frac{S_{g,global}}{S_{global}})\times {10}^6$$ 2

2.2. Land-Use Change Data

As a crucial component in developing the damage factor, we applied the same land-use data (SSP1-RCP2.6) for 2015 and 2100 as variables in the model sourced from LUH2. Accordingly, we reclassified the land-use data into seven major categories based on type: cropland, pastureland, urban land, primary forested land, primary nonforested land, secondary forested land, and secondary nonforested land. Specifically, for nonprimary land, our comparison of land-use changes from 2015 to 2100 revealed increases and decreases in land share, with typical examples being urban land and cropland. Meanwhile, for primary land, specifically primary forested land and nonforest areas, a decreasing trend was observed when comparing current and future land-use change scenarios.

To determine net changes in land-use status between two time points, we compared the proportions of land-use types within each grid cell at both time points. Rather than tracking transformations between categories or conducting a cross-analysis of land-use changes for each grid cell, our approach focused on summarizing changes by type.

To avoid double-counting land-use changes within a single grid cell, we calculated the damage factor unidirectionally based on increases in each land-use type resulting from human activities (cropland, pastureland, urban land, secondary forested land, and secondary nonforested land). This approach accounts for compensatory area shifts among land-use types that occur during land-use transitions within a grid cell. Despite accounting for seven land-use types in our model, primary natural land did not experience increases in future changes. Any potential decrease in primary natural land was reflected as an increase associated with the other five land-use types. Therefore, we calculated the total area of the five non-native land types (Area _{land–use type,c} ) that increased within each grid cell (Grid cell unit: km²) that increased within each grid cell between 2015 and 2100.

$${Area\_inc}_c=\sum _{land-usetype}\Delta {Area}_{land-usetype,c}$$ 3

The scenario outlined the impact of increased dynamic land compensation per km² within the grid rather than focusing on the effects of cross transformation. This approach was flexible and more straightforward for users to interpret.

2.3. Spatial Damage Factor

The damage factor was defined by analyzing the increase in extinction rates resulting from the marginal effects of land-use change, following the LCIA framework, which shows a relatively linear relationship. The damage factor for each grid cell, DF _c , was determined by dividing the calculated total extinction rate (Extinction Rate _c,g ) by the total area of land-use change within that cell Area _c . The extinction rate represents an integrated scenario-period intensity consistent with the temporal definition of land-use change. (eq )­

$${DF}_c=\frac{{ExtinctionRate}_c}{{{Area}_c}^{*}}$$ 4

where DF _c is the damage factor for grid cell c (unit: E/MSY/km²); Extinction Rate _c is the taxonomic-group-based extinction rate for grid cell c (unit: E/MSY); and Area _c is the total increased area of land-use change within grid cell c (unit: km²).

When performing a spatial analysis of the damage factor, the target grid cells in both data sets must be aligned. Thus, we binarized each raster for both data sets (extinction rate: E(c); land-use change: L(c)), resulting in binary indicators I _E (c) and I _L (c), where 1 denoted the presence of a nonzero value, and 0 otherwise.

Based on these indicators, we defined a consistency mask M _c = 1­(I _E (c) = I _L (c)), where M _c served solely as a selection criterion to identify valid target cells and was not used as a multiplicative factor in subsequent calculations. This revealed three possible scenarios. (1) I _E (c) ≠ 0, I _L (c) = 0: Grid cells where the extinction rate was nonzero while land-use change was zero. These were excluded from our final target range. Since we considered only five of the seven land-use change categories in the inventory data set, the grid with a land-use change value of zero was primarily attributable to unaccounted primary land change. (2) I _E (c) = I _L (c): Grid cells where the extinction rate and land-use change data were simultaneously present (both = 1) or simultaneously absent (both = 0). These scenarios were selected as our target grid cells for calculating the damage factor. (3) I _E (c) = 0, I _L (c) ≠ 0: Cells where extinction risk was estimated as zero despite the presence of land-use change were also excluded. In these cases, the absence of detected extinction risk cannot be confidently interpreted as a true absence of impact. To avoid potential misinterpretation, we adopted a conservative approach and excluded these cells from the target range.

$$\begin{array}{l}{I}_E(c)=1(E(c)\ne 0),I_L(c)=1(L(c)\ne 0)\\ M(c)=1(I_E(c)=I_L(c))\\ C_{valid}=\{c\in \mathrm{\Omega }/L(c)\ne Nodata\wedge E(c)\ne NoData\}\end{array}$$ 5

Subsequently, we applied the mask to both data sets, excluding the grid cells outside the calculation target. This yielded the target extinction rate (Extinction Rate _c,global ) and land-use inventory (Area_inc _c ), which can be used to calculate the ultimate damage factor:

$${{Area}_c=Area\_inc}_c·M_c$$ 6

$${ExtinctionRate}_c={ExtinctionRate}_{c,global}·M_c$$ 7

where M _c is a binary consistency mask (0 or 1) applied to each grid cell c, used to select valid cells for damage factor calculation.

However, when spatial data were used for calculations, extreme damage factors could occur when the land-use change quantity was minimal or approached zero. Therefore, to address these outliers, we established a threshold L _threshold (w) for land-use change: if the inventory value Area _c,g (w) fell below this threshold, the minimum land-use was set to 0.01 km² per grid cell to avoid unstable estimates.

$$\begin{array}{l}{}^{*}{\mathrm{\Omega }}_{adjust}=\{\overline{)\omega \in \mathrm{\Omega }}|Are{a}_c(\omega )|<0.01\}\\ {}^{*}{L}_{threshold}=\{\begin{array}{l}0.01,\omega \in {\mathrm{\Omega }}_{adjust}\\ {Area}_c(\omega ),\omega \notin {\mathrm{\Omega }}_{adjust}\end{array}\end{array}$$ 8

Finally, using these models and calculations, we obtained the global spatial distribution of the required damage factors.

To obtain representative damage factors at regional and national scales (DF _reg ), we aggregated grid-level damage factors using biodiversity-weighted averaging, as reflected in eq . This approach weighted each grid cell by its species richness (S _c ) and land-use change area (Area _c ), ensuring that the regional damage factor reflected the cumulative biodiversity impacts of land-use transformation.

$${DF}_{reg}=\frac{\sum _{c=1}^n({DF}_c\times S_c\times {Area}_c)}{\sum _{c=1}^n(S_c\times {Area}_c)}$$ 9

2.4. Land Use Type-Specific Damage Factor

To derive land-use-type-specific damage factors, damage factors (DFs) were first calculated at the grid cell level as the ratio between the total increase in the extinction rate caused by human activities and the total area of land-use expansion within each grid cell. In this framework, all land use types within a grid cell share the same damage factor.

This assumption is consistent with the MaxEnt, in which habitat probabilities are determined by the overall environmental structure of each grid cell. This environmental structure results from the combined configuration of all land-use types, rather than from any single type. Therefore, habitat changes and associated extinction risks within a grid cell were considered the integrated outcome of all land-use changes occurring in that cell, and individual land-use types were assumed to contribute equally to the grid-level biodiversity impact.

$$\begin{array}{l}{\mathrm{\Omega }}_k=\{\overline{)\mathrm{c}\in \mathrm{\Omega }}{L}_k(c)>0\}\\ {DF}_k=\{{DF}_c|c\in {\mathrm{\Omega }}_k\}\end{array}$$ 10

where Ω denotes the set of all grid cells, and L _k (c) represents the area of land-use type k in grid cell c. Ω _k , therefore defines the subset of grid cells where land-use type k occurs. The distribution of DF _c within Ω _k was used to characterize land-use type-specific damage factors.

Land-use-type-specific damage factors were then derived by analyzing the distribution of grid-level factors across all grid cells where a given land-use type occurred. This captured the spatial heterogeneity of biodiversity impacts associated with each land-use type across different environmental and biodiversity contexts at the global scale. These maps are provided in the Supporting Information.

Variability in land use type-specific damage factors was characterized based on the distribution of grid-level values. Bootstrap resampling across all valid grid cells (1000 iterations) was applied to estimate the weighted average damage factors and their 95% confidence intervals for each land use type. This analysis was intended to characterize the variability of the damage factors across different spatial contexts.

3. Results

3.1. Global Extinction Rate

Initially, we estimated species-specific extinction risks based on habitat changes projected using MaxEnt species distribution models. Model performance was validated using a stratified random sample of 175 species (35 per taxonomic group, representing 2.7% of the full data set), yielding a mean area under the curve (AUC) of 0.886 ± 0.090 (SD range: 0.561–0.993). Extinction rates were subsequently calculated by assessing extinction risk within each grid cell, which may reflect species richness patterns, as species-level extinction intensities are aggregated within each cell (Figure ). This pattern is partly related to regional species richness because extinction intensity aggregates species-level risks within each grid cell, while the E/MSY metric reduces but does not completely remove the influence of species richness.

1.

Distribution of estimated extinction rates on a spatial scale. (a) Net extinction rate (unit: E/MSY) reflecting an 85-year extinction period encompassing both positive (habitat loss) and negative (habitat gain) values. (b) Extinction rate attributed solely to habitat loss (positively valued occasion). (c) Extinction rate attributed solely to habitat gain (negatively valued occasion). The ribbon plot illustrates the spatial characterization distribution corresponding to the latitude scale.

This primary contribution emphasizes fine scale analyses at both species and grid cell levels. At the species level, certain species experienced considerable habitat loss attributable to anticipated land-use alterations and environmental context changes, leading to considerably high positive extinction rates. Conversely, some species experienced habitat gains from land-use changes, manifesting as negative extinction rates. Under the SSP1-RCP2.6 scenario, the gross extinction rateaccounting solely for positive extinction rateswas measured at 165.29 E/MSY, whereas the total gross negative extinction impact, reflective of habitat gain, was −3.36 E/MSY. Combined, these results produce a net impact of approximately 161.9 E/MSY. Notably, the negative values indicate habitat expansion predicted by changes in habitat suitability; they do not represent realized population recovery. Additionally, these outcomes are linked to model range shift capacities in response to environmental change, a trend particularly notable among bird species due to their generally higher dispersal capabilities and greater potential for broad range adjustments.

3.2. Location-Based Damage Factors

A location-based damage factor map was generated using grid units with an approximate resolution of 0.25 arc degrees (Figure ). Each grid cell indicates the potential marginal extinction rate associated with an incremental increase of 1 km² in the anthropogenic land use. This data set is well-suited for biodiversity assessments utilizing finer-scale primary inventory data.

2.

(a) Damage factor distribution map in a 0.25arc-degree resolution (unit: E/MSY/km²). Further disaggregated net damage factors at the taxonomic group-based level: (b) birds, (c) plants, (d) amphibians, (e) reptiles, and (f) mammals. The more continuous pattern observed for birds is mainly due to their larger sample size, broader spatial coverage of occurrence data, and higher dispersal ability, while other taxonomic groups show more fragmented patterns due to more heterogeneous distributions.

The global averaged net DF was 1.73 × 10^–05 (95% confidence intervals [CI]: 1.71 × 10^–05, 1.75 × 10^–05) E/MSY/km². Analysis of spatial distribution characteristics revealed that negative DF values were predominantly concentrated in inland regions, including Central Asia, northwestern China through Mongolia and Siberia, and the Arabian Peninsula. Notably, the mean negative DF was −7.26 × 10^–05 (95% CI: −7.62 × 10^–05 to −6.89 × 10^–05) E/MSY/km², whereas the mean positive DF was 1.17 × 10^–03 (95% CI: 1.09 × 10^–03–1.25 × 10^–03) E/MSY/km². Considerable overlap was observed between these regions and global biodiversity hotspots. The high biodiversity richness in these regions correlated with abundant natural resources; however, the loss and extinction of endemic species rendered these habitats particularly vulnerable to environmental changes. , The global top 10% of DF values ranged from 2.79 × 10^–4 to 1.2 E/MSY/km², while the top 5% ranged from 1.42 × 10^–3 to 1.2 E/MSY/km². This suggests that only a small fraction of spatial pixels exhibits extreme DF intensities. Geospatial analysis highlighted two area in the top 10% of high damage factors: southern Chile and Arnhem Land in northern Australia. In southern Chile, land-use projections anticipate secondary forest cover will increase by 30–40% by 2100. This expansion correlated with reduced damage factor values for bird species due to enhanced habitat connectivity along a critical South American migratory corridor. However, secondary forests negatively affected the habitat suitability for reptiles and mammals, raising damage factor values, which underscore the ecosystem-specific trade-offs associated with forest regeneration. While birds benefit from less fragmented habitats, secondary forests lack structural complexity, exacerbating the pressures on other terrestrial vertebrates.

In northern Australia, primary nonforested lands were extensively distributed but projected to decrease by approximately 30% by 2100 due to conversion to cropland. This land-use change could lead to significant biodiversity losses among native reptiles, mammals, amphibians, and vascular plants, with the highest extinction rates projected for the vascular plant species.

3.3. Regional-Based Damage Factor

By aggregating the grid cell-level DF values, we derived representative regional DF metrics that served as generalized indicators, especially enhancing biodiversity assessments based on secondary inventory data within LCA. Considering the typically non-normal distribution of values across regionsincluding positive and negative outliersand the large sample sizes per region, the weighted average was employed to mitigate the influence of these factors on regional assessments. As a representative value, the weighted average offers enhanced stability and interpretability. Additionally, each representative damage factor is accompanied by corresponding 95% CIs for a clearer understanding.

More than half of the top 10 countries with the highest regional damage factor values were official global biodiversity-rich regions, including New Zealandwhich ranks first, with a representative damage factor estimated at 2.64 × 10^–04 E/MSY/km² (95% CI: 1.85 × 10^–04 to 3.44 × 10^–04; Figure b). Finland followed in second place with a damage factor of 1.95 × 10^–04 E/MSY/km² (95% CI: 1.37 × 10^–04 to 2.54 × 10^–04). Additionally, Norway was estimated at 8.40 × 10^–05 E/MSY/km² (95% CI: 5.88 × 10^–05 to 1.09 × 10^–04). Species inhabiting cold northern climates remain particularly vulnerable and sensitive to habitat changes driven by environmental shifts. While many species adapt to warming climates by migrating poleward, this strategy is constrained in Arctic nations due to the absence of suitable habitats beyond the polar limit.

3.

(a) Disaggregated regional damage factors that use the weighted average for the extraction and analysis of each region. (b) Top 10 countries with the highest damage factors globally; areas with less than 1 grid cell were excluded from the ranking due to uncertainty. Error bars represent the bootstrap-derived 95% confidence intervals estimated from grid-level damage factors variability within each region. (unit: E/MSY/km²).

In the East Himalayas to Southeast Asian biodiversity-rich regions, Myanmar ranked third (DF = 1.34 × 10^–04 E/MSY/km²; 95% CI: 9.40 × 10^–05 to 1.74 × 10^–04), followed by Laos (1.11 × 10^–04 E/MSY/km²; 95% CI: 7.78 × 10^–05 to 1.45 × 10^–04), Bhutan (1.08 × 10^–04 E/MSY/km²; 95% CI: 7.54 × 10^–05 to 1.40 × 10^–04), Vietnam (9.68 × 10^–05 E/MSY/km²; 95% CI: 6.78 × 10^–05 to 1.26 × 10^–04), and Brunei (1.11 × 10^–04 E/MSY/km²; 95% CI: 7.78 × 10^–05 to 1.45 × 10^–04). Despite having minimal land-use change, Brunei’s high species richness implies that even limited development can result in intense biodiversity loss. The East Himalayas and Southeast Asia exemplify pristine ecological structures and rich biodiversity. However, economic development pressures have led to extensive exploitation of these ecological resources, resulting in notable biodiversity declines.

Similarly, Equatorial Guinea exhibited a damage factor of 8.54 × 10^–05 E/MSY/km² (95% CI: 5.98 × 10^–05 to 1.11 × 10^–04). Although we disaggregated damage factor values to the country or regional level, the representative values and associated CI ranges demonstrate limited heterogeneity across these areas. Malaysia had an estimated damage factor of 7.76 × 10^–05 E/MSY/km² (95% CI: 5.43 × 10^–05 to 1.01 × 10^–04), attributable not only to ongoing urbanization and the reduction of primary natural lands but also to its role as a biogeographical bridge between the Asian continent and the Pacific region, supporting a diversity of endemic species.

3.4. Land-Use-Based Biodiversity Damage Variability

To address the need for robust LCA study implementations, a representative damage factor for land use was developed for users of secondary inventory or nonlocation-specific LCA data sets. In this study, distinct land-use change impacts within each grid were assigned the same damage factor without disaggregation, as biodiversity loss in any given grid typically resulted from cumulative local pressures rather than a single land use change. These pressures were associated with specific combinations of environmental variables on the grid scale. However, the same land-use changes exhibited different damage factors across different locations, effectively capturing spatial heterogeneity in the biodiversity impact.This grid-based approach enables the application of location-specific damage factors, even though transition-specific values are not differentiated within the individual grids. Complete grid-level damage factor maps are provided in the Supporting Information (Figure S1). Despite this grid-level aggregation, we derived land-use-type-specific statistics by analyzing damage factor distributions across all grids where each type occurs, yielding differentiated mean factors for urban, cropland, pasture, and forest conversions.

We conducted uncertainty analyses for each globally distributed, land-use-oriented damage factor, presenting 95% CIs and weight-average value. The urban-induced damage factor had the highest weight-average at 1.85 × 10^–02 (95% CI: 1.61 × 10^–02 to 2.09 × 10^–02; Figure ) E/MSY/km², followed by the cropland-induced damage factor at 8.82 × 10^–03 (95% CI: 7.65 × 10^–03 to 9.99 × 10^–03) E/MSY/km² and grazing land-induced damage factor at 5.54 × 10^–04 E/MSY/km² (95% CI: 4.16 × 10^–04 to 7.22 × 10^–04) E/MSY/km². Secondary nonforested land exhibited relatively minor global impacts compared to other land-use types, with a weighted mean damage factor for biodiversity of 1.09 × 10^–04 (95% CI: 7.93 × 10^–05 to 1.44 × 10^–04) E/MSY/km², representing the lowest among the five types. In contrast, the weighted average damage factor for secondary forested land was approximately 3.25 × 10^–04 (95% CI = 1.96 × 10^–04 to 5.11 × 10^–04) E/MSY/km².

4.

Frequency plots for each land use-induced damage factor occurrence: (a) cropland, (b) grazing land, (c) urban land, (d) secondary nonforested land, and (e) secondary forested land. (f) Comparison of five distribution types; hatched line: mean value’s position; x-axis: marginal damage factor (DF) values for land use, expressed as E/MSY/km²; y-axis: global distribution frequency of values, rescaled to a frequency of <500; hatched purple line: 95% CI range, red line: mean value’s position; blue and red are separated by 0.

Marginal extinction rates associated with the expansion of various land types suggest that secondary natural lands generally support ecosystem restoration and improved species habitats (blue areas). However, land management strategies must be refined to account for location-specific biodiversity challenges, as significant species extinctions persist in certain nonforested areas. Additionally, most other land-use types that intensified through anthropogenic activities tended to increase species extinction rates.

4. Discussion

4.1. Characterization of Location-Based DF

As shown in the results, some regions as exhibiting negative (indicating habitat gain) and others with positive (indicating habitat loss) damage factors. There are three primary reasons for the negative damage factor areas. First, their inland location, characterized by land continuity and expansiveness, facilitated the rapid relocation of species in response to habitat degradation resulting from environmental change pressures. , Second, these inland areas, characterized by relatively stable climatic conditions, predominantly featured pasture and primary/secondary vegetation as the ecotone, allowing species to access newly suitable habitats within their species-specific dispersal limits during ecological migration, rather than through ecological adaptation. , Third, according to land use projections, these regions experienced minimal mean net land-use transitions, indicating less intensive land use and reduced biodiversity pressurenot necessarily active habitat restoration.

Certain regions like northern Amazon also exhibited negative values; however, these areas reflect projected increases in habitat suitability under the SSP1-RCP2.6 sustainability-oriented scenario, rather than realized biodiversity recovery or policy-driven restoration outcomes. , These negative values should not be interpreted as direct biodiversity recovery or beneficial effects of land-use change. Instead, they reflect potential shifts in habitat suitability and a possible reduction in extinction rates under specific scenarios. These contrasting patterns highlight the spatial heterogeneity of biodiversity responses to land-use change.

High positive damage factor valuessuggesting greater habitat losswere frequently observed in regions outside the continents, particularly in ecotones where extinction rates can be substantially higher than in stable ecosystems. Warm, humid ecoregions helped maintain biodiversity richness, notably from East to South Asia, the Indonesian archipelago, outer Australian regions, New Zealand, southern North America, Meso-America, southern Africa, and western Madagascar. High positive damage factor values in ecotonal and warm, humid regions suggest greater habitat sensitivity to projected land-use transitions. These regions further illustrate ecosystem-specific responses to land-use change. Because damage factors are defined relative to land-use change area, regions with limited transitions may still show high DF values, influenced by regional species vulnerability.

Land-use change represents a relatively complex alteration of ecological structures, where pre- and postchange combinations can synergistically affect the ecosystem. Focusing on individual variables may weaken or disrupt the synergistic effects of the combinations. Ecosystem imbalance and biodiversity loss at finer scales are often driven by cumulative stressors rather than the direct influence of a single variable. −

Despite technological advancements improving existing large land-use data sets, significant temporal and spatial discrepancies persist across data sets. Standardizing damage factors at small scales can partially address the heterogeneity encountered by users when employing various data sets, simplifying model complexity, and enhancing the feasibility of user assessments. Moreover, unifying the damage factor across grid cells can mitigate land competition that may emerge in future policy formulations, offering a more scientifically verifiable theoretical basis for policy development and facilitating decoupling from traditional empirical approaches. For instance, this approach may rectify overestimating the benefits of increased forest vegetation for biodiversity conservation or underestimating the absolute disadvantages of urban land expansion for species. An accurate assessment of land-use structures and species within specific locations is essential to conserving local biodiversity.

4.2. Comparison with Literature

Numerous research tools have been developed to assess biodiversity loss and address its spatial significance. However, most studies lack thorough spatial comparisons to existing indicators. In analyzing the regional representative damage factors, we considered the homogeneous distribution across global hotspot regions, where most high-value regions were categorized as global hotspots. Furthermore, we compared our findings with the STAR metric. The STAR_r score emphasizes regions that historically supported numerous endangered species but currently do not, while the STAR_t score reflects the potential reduction in the extinction risk through threat abatement actions.

Given that damage factor calculations are heavily influenced by the amount of land-use change projected in the target scenario rather than by historical habitat data, we focused on spatial relationships between the projected extinction rate and the STAR_t metric’s distribution. Considering the varying dimensions and interpretations between the two data sets, and our focus on analyzing the coupling of high- and low-value distributions, we divided the data into quartiles. Data points (pixels) within the same quartile range were retained, while the matching ratio for each quartile was subsequently calculated. Detailed spatial overlap results are provided in the Supporting Information (Figure S3).

4.3. Overview of Extinction Rate

Substantial progress has been made globally in the study of extinction rates. Current estimates for the global background extinction rate range from approximately 0.1 to 2 E/MSY, with an average fossil rate approximating 1.8 E/MSY. Existing research, encompassing over 100 fossil species, indicates a widely referenced background rate of approximately 1 to 0.1 E/MSY. Increasing this background rate to 2 E/MSY is substantiated by extensive analyses of mammalian ranges, which conservatively indicate a background rate of approximately 1.8 E/MSY. One notable advantage of studying extinction rates is the wealth of data derived from historical fossil lifespans, which provide a comprehensive temporal reference.

A previous study reported that extinction rates in more recent history ranged from approximately 150 to 260 E/MSY. Across different temporal ranges and taxonomic groups, the historical extinction rate spans from approximately 1.7 to 73 E/MSY. , In stark contrast, recent extinction rates, primarily driven by anthropogenic activities, are estimated to be markedly higher, from 20 to 200 E/MSY.

Current research predominantly focuses on historical to near-recent periods due to the availability of surveys and historical records. Thus, substantial potential exists for further investigation to estimate future extinction rates. Present projections suggest that the extinction rate may be 100–1,000 times higher than the background rate, with existing studies adjusting future extinction rate estimates to fall within 200–1,500 E/MSY. Although factors such as anthropogenic drivers, species-specific characteristics, and differing scenarios affect these predictions, many converge within the range of 100–1,000 E/MSY for future estimates.

Compared to our previous findings, under the SSP1-RCP2.6 scenario, research examining land-use changes and their implications for biodiversity extinction rates, based on over 6 000 species, estimated an extinction rate of approximately 746 E/MSY. However, this reflects a worst-case scenario regarding habitat changes for birds and underscores the severity of potential losses; it serves to raise the upper limit of the overall extinction rate estimates. Conversely, the current analysis of bird habitat changes offers a relatively accurate and moderate assessment of habitat alterations, particularly considering their long migratory adaptability and the native boundary conditions.

When considering adjusted global extinction rates that integrate distributional characteristics across taxonomic groups, the calculated total extinction rate was 161.9E/MSY for all five groups standardized to a million species-year basis. This suggests a plausible rate that aligns with the lower limit of the future projections. However, these results were derived through weighted aggregation rather than independent calculations, rendering direct comparisons with conventional rates by separate taxonomic groups less appropriate (Table S1).

4.4. Limitations of DF in LCIA

Significant advances in the effectiveness and applicability of damage factors, achieved by introducing a more precise computational model, will enable the comprehensive integration of extinction impacts into the LCIA framework. Nevertheless, several challenges persist in standardizing ecological metrics within LCIA.

• (1)For geographically large countries or regions, damage factors associated with land use changes and their impacts on biodiversity demonstrate considerable advantages by capturing both temporal and spatial heterogeneity. However, in nations or island states with modest spatial extents, damage factor development is hindered by coarse resolution, which fails to capture local land-use structural changes and ecosystem responses. Future research should utilize higher-resolution land-use data sets to train big data ecological models for more precise analytical computations.
• (2)When calculating the damage factors on a grid cell basis, we regarded the total land change within each grid as the cumulative pressure accelerating local biodiversity extinction rates. However, traditional LCIA applications differentiate damage factors across various land-use categories with greater granularity. To enhance alignment with LCIA applications, conducting sensitivity analyses for each land-use categoryby quantifying marginal land-use changes corresponding to biodiversity extinction rate increases within each gridis warranted.
• (3)Developing biodiversity-related damage factors within LCIA continues to exhibit a disconnect from the ecological theory. MaxEnt analysis revealed that species exhibited equivalent preferences across all land use types, as observed through response curves incorporating linear and nonlinear dynamics. , However, current LCIA approaches insufficiently address nonlinearity in damage factors, which may lead to policy decisions that inadvertently exacerbate land-use competition. Additionally, we estimated species’ remaining lifespan using two time points under a linear habitat change assumption, assuming a constant rate of habitat change between the baseline (2015) and future (2100) scenarios and extrapolating this rate to estimate the time to potential extinction. This simplification may introduce uncertainty when habitat changes are nonlinear over time and do not account for potential conservation or land-use changes. It may also overlook extinction time lags, where species persistence can continue after habitat loss. Incorporating additional time points could improve the lifespan calibration and model reliability.
• (4)The future land-use change projections were based on an established harmonized data set under the SSP1-RCP2.6 scenario. Consequently, the spatial allocation and distribution of damage factors within grids were scenario-dependent. The emergence of negative values can be attributed to habitat expansion due to sustainable policy interventions. For optimality analyses of local damage factors at the single-grid-cell level, multiple scenario-based damage factors must be integrated to ensure precise calculations.

5. Implications and Future Perspectives

This study investigated future extinction rates by integrating the LCIA findings to establish new damage factors. This approach establishes a robust framework for evaluating the closed loop from upstream policy decisions to downstream applications and considers the relevance of various types of LCA data for end users. Targeted solutions are proposed to address the disconnect between previous research and policy, specifically regarding biodiversity loss compensation and restoration. Furthermore, indicators are upscaled across spatial and temporal scales, enhancing visualization and interpretation to support location-based decision-making.

The results can inform assessment tools such as the LEAP approach, which offers more precise evaluation of location-based biodiversity extinctions and facilities alignment with policy benchmarksan area previously underrepresented in LCIA models. Higher-resolution ecological models and the integration of expert knowledge allow for consideration of species-level ecological traits and variability within existing LCIA frameworks. Future research should prioritize expanding the foundational species database to reduce taxonomic bias and improve the efficiency of damage factors across spatial and temporal dimensions. Additionally, integrating outcomes from multiple impact categories within the LIME model and converting these results into monetary indicators is anticipated.

Glossary

Abbreviations

DF

damage factor

ER, extinction risk; GEP

Global Extinction Probability

GIS

geographic information systems

LCIA

life-cycle impact assessment

LPI

Living Planet Index

MSA

mean species abundance

PBs

planetary boundaries

PDF

potentially disappeared fraction

STAR

species threat abatement and restoration

TNFD

Taskforce on Nature-related Financial Disclosures

WWF

World Wide Fund for Nature.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.6c02140.

• AUC summary for 6569 species (XLSX)

• Raster file of damage factors for 50 km (PDF)

• Supporting information (PDF)

• Extinction time summary (XLSX)

• Country Specific DF_ver0 (XLSX)

R.L. conceptualized and conducted the research; H.O. and A.H. guided the modeling and provided species data input; L.T., N.I., and R.L. guided the methodologies and indicators; T.M. and K.T. provided advice on the paper. All authors have read and agreed to the publication of the manuscript.

This study was partly funded by the “Development and Demonstration Project of Nature Footprint for Promoting Nature-related Financial Disclosures by Investors and Financial Institutions and International Standardization” (JPJ013028) administered by the Ministry of the Environment as part of the “Programs for Bridging the Gap between R&D and the Ideal Society (Society 5.0) and Generating Economic and Social Value” (BRIDGE) implemented by the Cabinet Office of Japan; and the Environment Research and Technology Development Fund (JPMEERF20202002, JPMEERF20241001, JPMEERF15S11400, JPMEERF23S12108, JPMEERF25S12423) of the Environmental Restoration and Conservation Agency of Japan, JSPS KAKENHI (Grant Number JP21H04944).

Other contents of this article are licensed under a Creative Commons Attribution 4.0 International License. This license allows you to use, share, adapt, distribute, and reproduce the material in any medium or format, provided that you give proper credit to the original author(s) and source, include a link to the Creative Commons license, and indicate if any changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you must obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

The authors declare the following competing financial interest(s): K.T. is affiliated to MS&AD InterRisk Research & Consulting, Inc. The rest of the authors declare no conflicts of interest.