Biodiversity Impact Assessment Considering Land Use Intensities and Fragmentation

Land use is a major threat to terrestrial biodiversity. Life cycle assessment is a tool that can assess such threats and thereby support environmental decision-making. Within the Global Guidance for Life Cycle Impact Assessment (GLAM) project, the Life Cycle Initiative hosted by UN Environment aims to create a life cycle impact assessment method across multiple impact categories, including land use impacts on ecosystem quality represented by regional and global species richness. A working group of the GLAM project focused on such land use impacts and developed new characterization factors to combine the strengths of two separate recent advancements in the field: the consideration of land use intensities and land fragmentation. The data sets to parametrize the underlying model are also updated from previous models. The new characterization factors cover five species groups (plants, amphibians, birds, mammals, and reptiles) and five broad land use types (cropland, pasture, plantations, managed forests, and urban land) at three intensity levels (minimal, light, and intense). They are available at the level of terrestrial ecoregions and countries. This paper documents the development of the characterization factors, provides practical guidance for their use, and critically assesses the strengths and remaining shortcomings.


Characterization factor framework
The probability of dispersal is based on the least-cost distance (w, m) between patches x and y and the median dispersal distance (α, m) of species group g in region j (eq.S1). 1  ,,,,, =  − ,,,,,  ,, ⁄ (Eq.S1) The least-cost distance is determined by the interpatch distance (d, m) and the resistance (r, dimensionless) of the surrounding land type k separating the patches (eq.S2).
,,,,, = (∑  ,,,, •  ,,,,  ) (Eq.S2) The allocation factor (a, dimensionless) is based on the area (A, m 2 ) and the habitat affinity (h, dimensionless) of the land type i (here excluding primary vegetation) and use intensity m (eq.S3). 2  ,,, = ( (Eq.S3) Model parametrization for plants In the raw data of Gallego-Zamorano et al. 3 , we changed the classification of urban forests from intense to light use.Moreover, we removed six data records (one urban, two cropland, and three plantation forests), as they were identified as outliers based on the modified z-score, leaving 319 data records for further analysis.While Gallego-Zamorano et al. 3 only considered three intensity levels of cropland and two of pasture and otherwise only considered broad land use types, we distinguished more often three intensity levels, including pasture and urban areas.We considered primary and secondary vegetation with the vegetation type forest as managed forest.Primary vegetation with a minimal use intensity served as the reference land use.If the change in species richness relative to the reference land use exceeded 1, which happened for minimally used urban land, it was cut off at 1 to avoid giving benefits under the high uncertainty of the underlying data.
A z-value for the biome "mangroves" was missing and assumed to equal flooded grasslands and savannas, as both are water-inundated and have equal z-values for mammals. 4 link the species with traits to those with spatial distributions, the names in both databases were standardized based on the World Flora Online taxonomic backbone 5 and exact and fuzzy matching. 6If only the genus could be matched, we replaced this part in the species name and tried to match it again.If the fuzzy distance was larger than 2 or equal to 2 but both changes were substitutions as opposed to insertions or deletions, we considered the matching too imprecise and the species unmatched.Following this name standardization, some names turned out to be synonyms, and duplicates were removed from the plant species distributions, leaving 26,573 vascular plant species for further analysis.
The trait categories of the "dispersal syndrome" and "growth form" needed to be harmonized and aligned with the categories of the linear regression model.The dispersal syndrome could include the categories animal (vertebrate), ant, ballistic, wind (none, i.e. without special adaptation), and wind (special, i.e., with the help of seed appendages).Since TRY did not allow for a distinction between different forms of wind dispersal, we assumed all to be wind (special), as this category was more common in the dataset of Tamme et al. 7 .The growth form included the categories tree, shrub, and herb.If TRY provided multiple values for a certain species and trait, we would select the most frequent value.

Model parametrization for vertebrates
We received the dataset used for Newbold et al. 8 and the demo script to obtain the relative species richness per land use class via private communication with Tim Newbold (July 2022).
The demo script used the functions CorrectSamplingEffort, MergeSites and SiteMetrics from the library predictsFunctions of Newbold's GitHub repository. 9We adapted the script so that we could include some confounding variables when modelling the species richness response to land use (as done in the original publication): the distance to road, 10 the population density 11 and the estimated travel time to the nearest city of 50'000 or more people. 12ubsequent to using the GLMER function, we used the function PlotGLMERFactor, 9 which we adapted to retrieve the species richness difference expressed as percentage between natural habitat and the land use intensity levels resulting from the model.This way, we could obtain the relative species richness per land use intensity level by dividing the species richness difference expressed as a percentage by 100 and adding 1.
The data from the meta-analysis 13 are specific to the forest management category and species group at continental resolution (the raw data can be found in the supporting information of the reference, Excel sheet Raw data, columns Xc and Xe).Xe represents the "mean species richness in disturbed (managed) forest sites", and Xc represents the "mean species richness in reference (unmanaged) forest sites".For intense management, we used the data on clear-cut (all trees in the harvest areas are removed at once, resulting in even-aged silviculture); for light management, we used the data on selective logging (largest, highest and oldest quality trees are removed, while the remaining vegetation is left standing); for minimum management, we used the data on reduced impact logging (similar to selective logging, but adopted with more sustainable practices, e.g., to reduce the soil damage, biodiversity impact, etc.).
As for plants, a cut-off at 1 was adopted when the response ratios were above 1.
Since the habitat affinities of vertebrates depend on the species numbers (eq.9), the characterization factors for ecoregions with only one species recorded were considered unreliable and set to a missing value.

Model parametrization regarding land use and intensity
Regarding cropland, we used the cropland from HILDA+ 14 as the base map and the area equipped for irrigation 15 and phosphorus and nitrogen fertilizer use 16 as supplementary maps.
The fertilizer use maps were disaggregated from 0.5° to 5 arcminutes by assigning the value of the original larger cell to all smaller cells within it.If the area equipped for irrigation and both fertilizer uses did not exceed the first quartile of non-zero values, it would be considered minimal use.If the area equipped for irrigation and fertilizer uses were below or equal to the third quartile, the cropland would be considered light use.If the area equipped for irrigation or any of the fertilizer uses exceeded the third quartile, the cropland would be considered intense use.
Regarding pasture, we used the pasture/rangeland from HILDA+ as the base map and land use from GLOBIO 4 17 as a supplementary map.We considered "grassland" from GLOBIO 4 as minimal use, "rangeland" as light use, and "pasture" as intense use.If there was a mismatch between the broad land use types of HILDA+ and GLOBIO 4, we also assumed light use.The GLOBIO 4 data was first reclassified to assign numeric values to the three intensity levels and then averaged to align the resolution with the base map.
Regarding plantations, we used the forest from HILDA+ as the base map and forest management 18 and oil palm plantation data 19 as supplementary maps.We considered "agroforestry" among the forest management classes as minimal use, "planted forests" among the forest management classes and smallholder oil palm plantations as light use, and "plantation forests" among the forest management classes and industrial oil palm plantations as intense use.The oil palm plantation data were first aggregated based on the most frequent category from 10 to 100 m to match the resolution of the forest management data, both were reclassified to the three intensity levels, and then averaged to align the resolution again with the base map.
Regarding managed forests, we used the forest from HILDA+ again as the base map but removed natural forests and plantations.This mainly left the type "naturally regenerating forests with signs of forest management, e.g., logging, clear cuts etc.", 18 as well as those cells for which no forest management class was indicated.Within the remaining forest areas, we estimated the share of 30-m cells with forest in 2000 that experienced forest extent loss in 2020 within 5-arcmin cells based on data from Potapov et al. 20 .We considered forest extent loss below or equal to 2% as minimal use, between 2% and 20% as light use, and above 20% as intense use.The resulting share in grid cells with minimal use is similar to the one for the classes from the Global Land Systems 21 that Newbold et al. 8 linked to primary and secondary vegetation with minimal use and that are relevant to forests.We included natural forests in this comparison, as Newbold et al. 8 do not distinguish between natural and managed forests.
Our share for light use is lower and for intense use higher.This seems justified because 1) the Global Land Systems only allowed considering livestock for the definition of intensity levels, while also other factors can contribute to light or intense use of managed forests, 2) their map represents the year 2000, while our CFs represent the year 2015, and there is a general trend towards more intensification, and 3) following a precautionary principle, overestimating the intensity level would be preferred over underestimating it.
Regarding urban areas, we used the urban area from HILDA+ as the base map and the Global Human Settlement 22 as a supplementary map.We considered the categories "Very low density rural", "Low density rural", and "rural cluster" as minimal use, "suburban or periurban" and "semi-dense urban cluster" as light use, and "dense urban cluster" and "urban cluster" as intense use., and 18 to Chaudhary and Brooks (2018) 24 .The correlation between the CFs from this study and those from GLAM1 is 0.79 for global, taxa-aggregated CFs, and the percent bias is -50.See also Figure S29 and Figure S30.

Figure S1 .
Figure S1.Conceptual overview of the methodological steps and main inputs necessary to produce the characterization factors (CFs).

Figure S2 .
Figure S2.Land occupation characterization factors at the ecoregion level for cropland with light use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S3 .
Figure S3.Land occupation characterization factors at the ecoregion level for cropland with minimal use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use

Figure S5 .
Figure S5.Land occupation characterization factors at the ecoregion level for pasture with light use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S6 .
Figure S6.Land occupation characterization factors at the ecoregion level for pasture with minimal use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal

Figure S8 .
Figure S8.Land occupation characterization factors at the ecoregion level for plantations with light or minimal use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.Note that the impacts of plantations with light or minimal use on plant species richness are not differentiated.More characterization factors are available through the use of proxies.

Figure S9 .
Figure S9.Land occupation characterization factors at the ecoregion level for managed forest with intense use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S10 .
Figure S10.Land occupation characterization factors at the ecoregion level for managed forest with light use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal

Figure S12 .
Figure S12.Land occupation characterization factors at the ecoregion level for urban areas with intense use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S13 .
Figure S13.Land occupation characterization factors at the ecoregion level for urban areas with light use and potential impacts on plant species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal

Figure S15 .
Figure S15.Land occupation characterization factors at the ecoregion level for cropland with intense or light use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.Note that the impacts of cropland with intense or light use on vertebrate species richness are not differentiated.More characterization factors are available through the use of proxies.

Figure S16 .
Figure S16.Land occupation characterization factors at the ecoregion level for cropland with minimal use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species

Figure S17 .
Figure S17.Land occupation characterization factors at the ecoregion level for pasture with intense or light use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.Note that the impacts of pasture with intense or light use on vertebrate species richness are not differentiated.More characterization factors are available through the use of proxies.

Figure S18 .
Figure S18.Land occupation characterization factors at the ecoregion level for pasture with minimal use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S19 .
Figure S19.Land occupation characterization factors at the ecoregion level for plantations with intense or light use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness,

Figure S20 .
Figure S20.Land occupation characterization factors at the ecoregion level for plantations with minimal use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S21 .
Figure S21.Land occupation characterization factors at the ecoregion level for managed forest with intense use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S22 .
Figure S22.Land occupation characterization factors at the ecoregion level for managed forest with light use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d)

Figure S23 .
Figure S23.Land occupation characterization factors at the ecoregion level for managed forest with minimal use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S24 .
Figure S24.Land occupation characterization factors at the ecoregion level for urban areas with intense use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S25 .
Figure S25.Land occupation characterization factors at the ecoregion level for urban areas with light use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d)

Figure S26 .
Figure S26.Land occupation characterization factors at the ecoregion level for urban areas with minimal use and potential impacts on vertebrate species richness.The unit is PDF/m 2 .a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.Grey denotes no data, indicating either the absence of the specific land use class in these regions or missing species data.More characterization factors are available through the use of proxies.

Figure S27 .
Figure S27.Land occupation characterisation factors for different species groups.a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.

Figure S28 .
Figure S28.Land occupation characterisation factors for different species groups.a) average impacts on regional species richness, b) marginal impacts on regional species richness, c) average impacts on global species richness, d) marginal impacts on global species richness.

Figure S29 .
Figure S29.Comparison of characterization factor (CF) ranks from this study and two previous studies across species groups.The black dashed line is the line of equality, while the coloured solid lines are linear trend lines.

Figure S30 .
Figure S30.Comparison of characterization factors (CFs) from this study and two previous studies across species groups.The axes are square-root-transformed.The black dashed line is the line of equality, while the coloured solid lines are robust lines using the Theil-Sen estimator.

Table S2 .
8and use classification by broad type and intensity, following Extended Data Table1from Newbold et al.8.IntenseOne or more disturbances that are severe enough to markedly change the nature of the ecosystem; this includes clear-felling of part of the site too recently for much recovery to have occurred.Primary sites in fully urban settings should be classed as Intense use.
LightOne or more disturbances of moderate intensity (e.g., selective logging) or breadth of impact (e.g., bushmeat extraction), which are not severe enough to markedly change the nature of the ecosystem.Primary sites in suburban settings are at least Light use.Intense Pasture with significant input of fertilizer or pesticide, and with high stock density (high enough to cause significant disturbance or to stop regeneration of vegetation).CroplandMinimal Low-intensity farms, typically with small fields, mixed crops, crop rotation, little or no inorganic fertilizer use, little or no pesticide use, little or no ploughing, little or no irrigation, little or no mechanization.

Table S3 .
Application of average characterization factors to global land occupation in 2015.All characterization factors used in this application are aggregated across the species groups and represent relative global species losses.

Table S4 .
Comparison with other CFs and IUCN data.r is the Spearman correlation coefficient, pbias is the percent bias, and nthreat is the number of species threatened by land use.Subscript 21 refers toKuipers et al.  (2021)

Table S5 .
Sensitivity analysis.r is the Spearman correlation coefficient, and pbias is the percent bias.Subscripts lb and ub refer to lower and upper bounds for the local relative species richness, broad to broad land use types without distinguishing land use intensities, and sar to the species-area relationship without considering fragmentation.

Table S6 .
Contribution to variance (%).Analysis performed for global, average characterization factors for land occupation.

Table S7 .
Correlation matrix of global, average, land occupation characterization factors across species groups.