Global assessment of landscape pattern changes from 1992 to 2020

Global assessment of landscape pattern changes from 1992 to 2020 | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Global assessment of landscape pattern changes from 1992 to 2020 Tamsin L. Woodman, Peter Alexander, David F.R.P. Burslem, Justin M.J. Travis, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6522006/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 16 Oct, 2025 Read the published version in Landscape Ecology → Version 1 posted 12 You are reading this latest preprint version Abstract Context Landscape patterns are driven by complex natural and anthropogenic factors and are important for a range of environmental processes, including species movement and wildfire risks. Previous assessments of global-scale landscape pattern change have focused on a single land use and land cover (LULC) type or landscape-level measurements, hence there is a lack of knowledge on landscape pattern change across multiple LULC classes. Objectives We assessed global-scale change in landscape patterns for the six LULC classes used in the HILDA + dataset (urban, cropland, pasture/rangeland, forest, unmanaged grass/shrubland, and sparse/no vegetation) from 1992 to 2020. Methods Six class-level landscape metrics which showed predictable scaling behaviour with landscape extent were calculated for each LULC class and year of the study period. Landscape metrics were quantified for five landscape extents (100, 400, 1600, 6400 and 25600 km 2 ). Trends in global landscape patterns over time were evaluated with a particular focus on area and fragmentation. Results Unmanaged grass/shrubland LULC expanded in area and showed increased fragmentation, while pasture/rangeland and forest LULC tended to decline in area and exhibit decreased fragmentation. Even though there was high spatial heterogeneity in landscape pattern change for all LULC classes, neighbouring 100 km 2 landscapes often showed the same directional change in area and fragmentation. Conclusions These findings highlight the variability in landscape pattern change at global scales and indicate that drivers of landscape pattern vary across the globe at local to regional scales, with implications for environmental processes such as biodiversity loss and carbon storage. Landscape patterns Landscape metrics Land use and land cover change Global spatial scale Landscape fragmentation Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Understanding how global landscape patterns have fluctuated through time is key for deciphering the drivers of changes and how these changes impact the Earth System. Anthropogenic land use and land cover (LULC) change is driving the loss and fragmentation of the Earth’s remaining natural ecosystems, with potentially damaging consequences for the Earth System (Haddad et al. 2015 ). Approximately 2.3 million km 2 of forest was lost between 2000 and 2012 (Hansen et al. 2013 ), for example, and over 70% of remaining forest cover is within 1 km of a forest edge (Haddad et al. 2015 ). By contrast, the amount of cropland has increased 9% from 2003 to 2019 (Potapov et al. 2021 ). LULC change alone can have adverse effects on the environment, such as through driving biodiversity loss (Maxwell et al. 2016 ; Jaureguiberry et al. 2022 ). Changes in landscape patterns can also have negative environmental impacts; for example, the increasing amount of tropical forest edge habitat has been estimated to release 0.34 Gt of carbon per year (Brinck et al. 2017 ). Moreover, landscape patterns are important for a range of environmental processes, including the movement of organisms (Fischer and Lindenmayer 2007 ), fire spread (Ryu et al. 2007 ) and ignition (Pais et al. 2021 ), and accumulation of soil carbon (Liu et al. 2022 ). Therefore, knowledge of how LULC has changed historically in terms of both area and pattern is important for estimating the impacts of LULC change on the Earth System. Landscape patterns are generated by a complicated array of natural and anthropogenic factors, such as anthropogenic LULC change, climate, soil properties, and biotic interactions (Turner and Gardner 2015 ). LULC change leads to changes in both the amount (composition) and configuration (spatial arrangement) of LULC classes, and usually also leads to smaller individual patches of a specific LULC class. The process of generating smaller LULC patches, which affects both the amount and configuration of LULC classes simultaneously, is often referred to as the process of fragmentation. However, fragmentation can also refer to changes in the spatial arrangement for a given area of a LULC class in a landscape (‘fragmentation per se’; sensu (Fahrig 2003 ). These differences in how fragmentation is defined, combined with varied correlation structures between measures of landscape structure at the patch versus landscape level, make it difficult to compare between studies and have led to considerable disagreement around the impacts of fragmentation on biodiversity (see (Fletcher Jr et al. 2018 ; Fahrig et al. 2019 ). Although fragmentation has been recognised as an important process of environmental change, the distribution and trends of fragmentation patterns for each LULC category remain unclear. The drivers of landscape pattern change and the impact of landscape patterns on the environment may differ across scales (Ewers and Laurance 2006 ; Cattarino et al. 2014 ; Jin et al. 2023 ), hence the choice of scale on which to study the effects of landscape patterns can have considerable influence on the conclusions from a study (Turner 1989 ; Miguet et al. 2016 ). For example, indices of macroinvertebrate richness were most closely associated with landscape patterns calculated at a scale of 200 m-wide riparian corridors (Sponseller et al. 2008 ), suggesting that this scale would be most appropriate for studying the effects of landscape patterns on macroinvertebrate diversity. Similarly, the relationships between landscape patterns and plant diversity are strongly dependent on the scale at which landscape patterns are quantified (Martello et al. 2023 ; Jin et al. 2023 ). The choice of scale on which landscape metrics, which are used to quantify landscape patterns (Gustafson 1998 , 2019 ), are calculated is therefore important as it may affect how landscape patterns appear to relate to their drivers and to environmental processes. Calculating landscape metrics at a range of scales may be preferable to better understand their cross-scale associations with landscape pattern drivers and environmental processes (Miguet et al. 2016 ). Although it is important to study landscape patterns at multiple scales, there are relatively few landscape metrics which behave predictably as the scale of a landscape increases in terms of both extent and resolution (Turner 1989 ; Wu et al. 2002 ; Wu 2004 ; Uuemaa et al. 2005 ; Argañaraz and Entraigas 2014 ). The lack of landscape metrics with predictable relationships across scales may impede our ability to make cross-scale comparisons of the drivers and impacts of landscape patterns. An increasing number of landscape metrics have become available in recent years compared to those tested for scaling relationships in previous studies (Wu et al. 2002 ; Wu 2004 ; Šímová and Gdulová 2012 ), which suggests there is an opportunity to assess the scaling behaviour of these new landscape metrics and update our knowledge of the behaviour of landscape metrics across scales. The choice of landscape metrics and definition of fragmentation used to assess landscape patterns at a global-scale has varied between studies, making it hard to compare studies that have focused on a single LULC class or landscape extent (for example, (Haddad et al. 2015 ; Hu et al. 2020 ; Ma et al. 2023 ). To our knowledge, previous estimations of global landscape pattern change have focused on forest or cropland patterns only (Riitters et al. 2000 ; Haddad et al. 2015 ; Hu et al. 2020 ; Ma et al. 2023 ), or quantified change at the level of entire landscapes rather than for individual LULC classes (Jacobson et al. 2019 ). Moreover, few studies have assessed global-scale landscape patterns across scales, except for (Riitters et al. 2000 ) who calculated global forest fragmentation across four spatial scales for a single time point. Although natural land cover classes such as forest are thought to have decreased and become more fragmented through time (Hansen et al. 2013 ; Haddad et al. 2015 ; Jacobson et al. 2019 ), a more recent study indicated that the majority of forested landscapes across the globe may have exhibited a trend of declining fragmentation between 2000 and 2020 (Ma et al. 2023 ). However, the metrics used to quantify fragmentation differ between studies; for instance, Haddad et al. ( 2015 ) utilised distance to edge, number of fragments and fragment area to quantify global forest fragmentation, whereas Ma et al. ( 2023 ) employed a fragmentation index constructed from edge density, patch density, and mean patch area. Consequently, there is a need for global-scale landscape pattern change assessments which encompass multiple LULC classes and landscape extents to give a better understanding of how landscape patterns are changing over time. Here, we address this research gap and quantify global-scale landscape patterns for several LULC classes, and assess how landscape patterns have changed over recent decades. This is the first study to quantify global landscape patterns for multiple LULC classes, landscape extents, and years. Given that most class-level landscape metrics show unpredictable scaling relationships across landscape extents (Wu 2004 ), we first use a single country, Colombia, which has highly heterogeneous landscape patterns, to identify metrics with predictable scaling relationships when landscape extent increases. The selected metrics are then used to calculate global landscape patterns for a range of landscape extents between 1992 and 2020, to address how global landscape patterns have changed over time. We also attempt to identify where different LULC classes have shown changes in both area and fragmentation per se over the study period. Our study intends to generate broader knowledge of the overall trends and spatial variability in global-scale landscape patterns over the past three decades, which has implications for understanding the drivers of landscape pattern change and their impacts on the Earth System. Methods Land use and land cover data We used land use and land cover (LULC) data from the HILDA + version 2b dataset in the Eckert IV projection for the calculation of landscape metrics at global scale (Winkler et al. 2020 , 2021 , 2024 ; Woodman et al. 2025 ). HILDA + provides yearly 1 km spatial resolution LULC data from 1960 to 2020. Each grid cell contains a single LULC class which is derived from an aggregation of multiple LULC maps and other related datasets, such as FAO land use statistics (Winkler et al. 2020 , 2021 , 2024 ). There are six LULC classes in HILDA + version 2b: urban, cropland, pasture/rangeland, forest, unmanaged grass/shrubland and sparse/no vegetation. HILDA + distinguishes between managed pasture/rangelands and unmanaged grass/shrublands, giving it an advantage over other global LULC datasets that treat managed and unmanaged grasslands as the same LULC class. Pasture/rangeland is defined in HILDA + as managed herbaceous plants with at least 10% cover, including areas that are used for livestock and hay production. Unmanaged grass/shrublands are natural herbaceous plants with at least 10% cover that are not managed by people, including wetland areas. Both the pasture/rangeland and unmanaged grass/shrubland classes include mosaics of herbaceous plants with trees and shrubs. A grid cell must have at least 10% cover of trees that are taller than 5 metres to be classed as forest (Winkler et al. 2020 , 2021 , 2024 ). We restricted the study period to between 1992 and 2020 because 1992 is the first year that a high resolution, yearly LULC dataset (the ESA CCI Land Cover time series) is used as input to HILDA+ (ESA 2017 ; Winkler et al. 2020 , 2024 ). For global analysis we cropped the HILDA + maps to exclude the continent of Antarctica and sub-Antarctic islands as very little LULC change occurred here during the study period. The HILDA + LULC maps were cropped in R software version 4.1.3 (R Core Team 2022 ) using the ‘terra’ R package version 1.7–23 (Hijmans 2022 ) and an outline of Antarctica from the ‘rnaturalearth’ package version 0.3.2 (Massicotte and South 2023 ). Selection of landscape metrics A key goal of our study was to look at changes in LULC classes across multiple spatial extents, for which we required class-level landscape metrics with consistent behaviour across scales. To achieve this, we first assessed the behaviour of class-level landscape metrics from the ‘landscapemetrics’ R package (Hesselbarth et al. 2019 ) with increasing landscape extent for Colombia, before moving to global level analyses. There have been large changes in LULC over time in Colombia, and the rate and drivers of LULC change have varied both spatially and temporally. For instance, in the late twentieth century the Andean region experienced the highest rates of deforestation (Etter et al. 2008 ). Current drivers of LULC change in Colombia include clearing of forests for cattle grazing, legal and illegal crop production, mining, and urbanization (Etter et al. 2008 ; Armenteras et al. 2011 ; González-González et al. 2021 ). Given its large area, diverse land covers and varied drivers of LULC change, all six LULC classes from HILDA + were represented in Colombia (Winkler et al. 2020 , 2021 , 2024 ; Woodman et al. 2023 ). Colombia was therefore considered a suitable case study to test the scaling relationships of landscape metrics across ten landscape extents (100, 400, 900, 1600, 2500, 3600, 4900, 6400, 8100 and 10000 km 2 ). First, an outline of Colombia from the ‘rnaturalearth’ version 0.3.2 package (Massicotte and South 2023 ) was used to crop the HILDA + dataset from 1992 to 2020 to the same extent as Colombia. Next, a set of ten regular grids covering the terrestrial surface of Colombia were created to represent landscapes with different extents. Each grid had landscapes (grid cells) with sides of between 10 and 100 km length at 10 km increments, giving a total of ten grids. Each grid was overlaid with LULC in Colombia from HILDA + and all grid cells that were entirely classified as ocean by HILDA + in every year from 1992 to 2020 were removed. Each individual cell in a grid was treated as a landscape for the calculation of landscape metrics. Two examples of landscapes with different extents (sides of length 20 km and 80 km, or 400 km 2 and 6400 km 2 landscape extent, respectively) overlaid on LULC in Colombia in 1992 are shown in Fig. 1 . Next, all 55 class-level landscape metrics implemented in the ‘landscapemetrics’ R package version 1.5.6 (Hesselbarth et al. 2019 ) were calculated in the first year of the study period (1992) for each landscape in the ten grids. After calculating class-level landscape metrics using the ‘landscapemetrics’ R package, we analysed the scaling relationship between the mean of each landscape metric and landscape extent. First, we removed the two HILDA + water LULC classes (ocean and water) from the dataset of landscape metrics in Colombia in 1992. The ‘landscapemetrics’ R package does not return a value for class-level landscape metrics when a LULC class is not present in a landscape (Hesselbarth et al. 2019 ), which resulted in missing values for metrics in many landscapes. It is not possible to calculate a value for the majority of landscape metrics when a LULC class is not present in a landscape, so we treated missing values for these metrics as missing when summarising the dataset (Table S 1). However, for other metrics, such as class area (CA) and total edge length (TE), the value of the metric is equivalent to zero when a LULC class is not present in a landscape. Therefore, we replaced missing values with zeros for eleven out of the 55 landscape metrics available in the ‘landscapemetrics’ package (Hesselbarth et al. 2019 ). The mean of each landscape metric was then calculated across all landscapes for every landscape extent, and the mean of landscape metrics as predicted by landscape extent in terms of the length of each landscape in kilometres was plotted (as in (Wu 2004 ). The resulting plots were examined to establish which landscape metrics showed consistent scaling relationships as landscape extent increased. Six landscape metrics were found to have predictable scaling behaviour as landscape extent increased (Fig. S 1, Fig. S 2, Table 1 ). The six landscape metrics which demonstrated predictable scaling with landscape extent across Colombia were: class area (CA), Landscape Shape Index (LSI), number of disjunct core area patches (NDCA), number of patches (NP), total core area (TCA), and total edge length (TE). Four of these six metrics were previously identified as having consistent behaviour as landscape extent was increased across a set of landscapes in the United States (Wu 2004 ). CA is the total area of a LULC class in one landscape in hectares, TE is the total edge length of a LULC class in a landscape in metres, and NP gives the number of non-contiguous patches of a LULC class. NDCA and TCA are both core area metrics, where ‘core area’ consists of grid cells which are surrounded by cells of the same class. NDCA counts the number of non-contiguous core area patches for a LULC within a landscape, meaning it is a measure of the number of ‘patches within patches’. Meanwhile, TCA is the total area of a LULC class that can be considered as the core area within a landscape, with units of hectares. LSI is calculated as a ratio of the total edge length of a LULC class to the hypothetical minimum edge length of that class if it was as aggregated as possible; hence, LSI quantifies the fragmentation of a LULC class within a landscape, with lower values indicating lower fragmentation (Hesselbarth et al. 2019 ). These six landscape metrics therefore encompass measures of LULC area, edge length, core area, and fragmentation. Calculating landscape metrics at global scale The six landscape metrics which showed consistent scaling relationships with increasing landscape extent across Colombia in 1992 (Table 1 ) were calculated at global scale to create a cohesive dataset of landscape patterns for multiple LULC classes and landscape extents. We decided to calculate global landscape metrics for five landscape extents: 10 by 10 km, 20 by 20 km, 40 by 40 km, 80 by 80 km, and 160 by 160 km (100, 400, 1600, 6400 and 25600 km 2 , respectively). The 10 by 10 km (i.e. 100 km 2 ) extent was chosen because it approximately matches the resolution of available socioeconomic datasets (for example: (Center for International Earth Science Information Network - CIESIN - Columbia University 2018 ; Kummu et al. 2018 ; Fischer et al. 2021 ) that could be utilised in future to investigate drivers of landscape patterns. The further four landscape extents were selected by doubling the number of kilometres per side of a landscape. Most global land use models typically generate projections at coarse resolutions (for example, in regions or 0.5° grids; Alexander et al. 2017 ), so the 1600, 6400 and 25600 km 2 landscapes (40, 80, and 160 km per side, respectively) aim to approximate the range of outputs obtained from global land use models. One grid covering the terrestrial surface of the Earth, excluding Antarctica, was created per landscape extent using the same method as for creating landscapes across Colombia. Each of the six class-level landscape metrics were then calculated within each landscape in the five global-scale grids using the ‘landscapemetrics’ R package version 1.5.6 (Hesselbarth et al. 2019 ) in R version 4.0.0 (R Core Team 2020 ) for all years in the study period (1992 to 2020). We assessed whether the selected landscape metrics showed predictable scaling across landscape extents at global scale using the same method as for Colombia (Fig. 2 ). Table 1 Class-level landscape metrics used to quantify global-scale landscape patterns. All six metrics showed consistent scaling relationships with increasing landscape extent across Colombia from 1992. Description of metrics based on Hesselbarth et al. ( 2019 ) Landscape metric Abbreviation Units Description Class area CA Hectares Total area of a LULC class in a landscape. Landscape Shape Index LSI None Ratio between the actual edge length of a LULC class in a landscape and its theoretical minimum edge length if it were as aggregated as possible. Number of disjunct core area patches NDCA None Number of core area patches of a LULC class in a landscape, where a core area patch has no neighbouring cells of another LULC class. Number of patches NP None Number of patches of a LULC class in a landscape. Total core area TCA Hectares Total core area of a LULC class in a landscape, where core area is made up of grid cells with no neighbouring cells of a different LULC class. Total edge length TE Metres Total edge length of a LULC class in a landscape. Analysis of landscape pattern change The dataset of global landscape patterns from 1992 to 2020 was used to analyse landscape pattern change of six LULC classes through time. Given the large number of analyses carried out, we do not present all the results for all landscape extents in the main Results. However, as all six landscape metrics show consistent scaling relationships globally (Fig. 2 ) we expect our conclusions will be applicable across scales. We focused our analyses on the smallest landscape extent to increase the sample size for subregions (i.e. continents) and also to better characterize relatively rare land cover classes, such as urban land cover (Fig. 2 ). Prior to analysing landscape pattern change during the study period, we first processed landscape patterns in the same way as for assessing the scaling relationships of landscape metrics. First, we removed HILDA + classes representing water from the dataset. Next, missing values were replaced with zero for any landscape metrics where we considered a missing value to be equivalent to zero (Table S 1). Landscape metrics were summarised by calculating the mean and standard deviation across all landscapes of a given extent in each year for every LULC class. Calculating the mean and standard deviation allowed us to assess the average and variation in landscape patterns across the globe from 1992 and 2020, and whether the trends were consistent through time. Additionally, we assessed the spatiotemporal variation in landscape pattern change by calculating the net change within each landscape between 1992 and 2020 for every landscape metric and LULC class. Net change was plotted for each metric and LULC class for 100 km 2 extent landscapes only to evaluate whether the magnitude and direction of change varied between landscapes and regions (Fig. S 29–34). We also calculated the average net change in landscapes for each continent to test for differences between continental- and global-scale landscape pattern change. To disentangle the impacts of changes in the area and configuration of LULC classes, we focused on two particular indices: class area (CA), which measures the area of a LULC class in a landscape, and Landscape Shape Index (LSI), which measures the fragmentation of a LULC class within a landscape (Hesselbarth et al. 2019 ). LSI is a good measure of changes in configuration independently of CA because it calculates edge length in relation to a hypothetical minimum edge length, and hence accounts for the area of a LULC class within the landscape. A high LSI value indicates a LULC class with high fragmentation per se, as the actual edge length is much longer than the hypothetical minimum if the LULC class was as aggregated as possible. Comparatively, low LSI indicates that a LULC class has low fragmentation per se. Overall, LSI provides a better measure of configuration independently of CA than total edge (TE), the patch metric NP, or the core area metrics NCDA and TCA, as all of these are known to be correlated with class area at the landscape scale (Wang et al. 2014 ; Lockhart and Koper 2018 ). To identify landscapes where both CA and LSI were changing in the same direction for one LULC class, we first selected all landscapes of 100 km 2 extent which contained that class in both 1992 and 2020. Next, the selected landscapes were classified into nine categories: both CA and LSI increasing (CA + LSI+); CA increasing and LSI decreasing (CA + LSI-); CA increasing and no change in LSI (CA + LSI=); CA decreasing and LSI increasing (CA-LSI+); both CA and LSI decreasing (CA-LSI-); CA decreasing and no change in LSI (CA-LSI=); no change in CA and increasing LSI (CA = LSI+); no change in CA and decreasing LSI (CA = LSI-), and no change in either CA or LSI (CA = LSI=). Maps of the nine categories of CA and LSI change were created to examine whether there was a tendency for regions and LULC classes to become more fragmented over time as LULC area changed. The percentage of landscapes in each of the nine categories was calculated at global- and continental scales for each landscape extent, to test whether the prevailing direction of change differed between LULC classes and across scales. Results Scaling relationships of landscape patterns All six class-level landscape metrics identified as showing consistent scaling relationships when landscape extent was increased across Colombia from 1992 were also confirmed to exhibit predictable behaviour with increasing landscape extent at global scale (Fig. 2 ). Five out of the six metrics appeared to exhibit a power-law relationship with landscape extent, whereas LSI had an approximately linear relationship with landscape extent. The relationship between each metric and landscape extent followed the same pattern across LULC classes, although the rate of increase differed between LULC classes. For example, forest CA increased much more rapidly with landscape extent compared to urban CA, likely because urban areas will not cover more than a small fraction of a 25600 km 2 landscape. Global trends in landscape patterns The mean values for the six landscape metrics in landscapes of 100 and 25600 km 2 extent varied by LULC class at global scale from 1992 to 2020 (Fig. 3 ). For example, the mean area of unmanaged grass/shrubland increased from 1362 ± 2545 ha (mean ± standard deviation) in 1992 to 1492 ± 2658 ha in 2020 in 100 km 2 landscapes, with a corresponding increase in all other metrics. Similarly, there were increases in five out of six landscape metrics for urban LULC in 100 km 2 landscapes, with only LSI exhibiting a small decrease from 1.35 ± 0.45 in 1992 to 1.34 ± 0.45 in 2020. Therefore, unmanaged grass/shrubland and urban land cover both increased on average in 100 km 2 landscapes between 1992 and 2020, which coincided with increasing core area and number of patches, and in the case of unmanaged grass/shrubland an increase in fragmentation per se as measured by LSI. By contrast, pasture/rangeland and forest LULC exhibited a decrease in area on average in most landscape metrics in 100 km 2 landscapes across the study period. For instance, at global scale all metrics declined for pasture/rangeland from 1992 to 2020, except for TCA which showed no net change between the two years. For forest cover, there was a decrease in five out of six landscape metrics from 1992 to 2020 (CA, LSI, NP, TCA, and TE), and a very small expansion in NDCA (0.52 ± 0.77 in 1992 and 0.53 ± 0.77 in 2020). Thus, the pasture/rangeland and forest classes declined in area and became less fragmented on average in 100 km 2 landscapes over the course of the study period. The magnitude of changes in landscape patterns were larger for pasture/rangeland than forest. The direction of change in landscape metrics in 100 km 2 landscapes from 1992 to 2020 was more variable for the cropland and sparse/no vegetation LULC classes, and the changes in each metric were smaller compared to the fluctuations in other LULC classes. Three out of six landscape metrics demonstrated an increase on average in 100 km 2 landscapes for cropland from 1992 to 2020, with LSI, NP and TCA showing small reductions (LSI = 1.67 ± 0.58 in 1992 and LSI = 1.63 ± 0.53 in 2020; NP = 0.52 ± 1.12 in 1992 and NP = 0.51 ± 1.06 in 2020; TCA = 454 ± 1395 ha in 1992 and 452 ± 1364 ha in 2020). Comparatively, there were small decreases in CA, NDCA, and TCA for sparse/no vegetation, increases in NP and TE, and no change in LSI, suggesting that sparse/no vegetation cover decreased and became more fragmented on average in 100 km 2 landscapes from 1992 to 2020. The behaviour of landscape metrics over time between 1992 and 2020 was generally similar in 100 km 2 and 25600 km 2 landscapes (Fig. 3 ), which was expected as the landscape metrics were selected to show predictable scaling behaviour across landscape extents. However, there were differences in the comparative magnitude of some metrics within 100 km 2 versus 25600 km 2 landscapes. For instance, unmanaged grass/shrubland NP was 10.2% more than the pasture/rangeland NP in 100 km 2 landscapes in 1992, and 38.6% higher in landscapes of 25600 km 2 extent. Cropland LSI was 6.5% higher than forest LSI in 100 km 2 extent landscapes in 1992, whereas in 25600 km 2 landscapes cropland LSI was 2.6% less than forest LSI. These differences when comparing landscape metrics between LULC classes across scales could be due to variability in the shape of the relationship between landscape metrics and landscape extent for different LULC classes (Fig. 2 ). There was considerable variation in the direction and magnitude of landscape pattern changes across continents (Fig. 4 ). For example, while pasture/rangeland CA declined globally the largest average net decrease was for Oceania (mean and standard deviation of -1047 ± 3074 ha) whereas there was net expansion on average in Africa (117 ± 1112 ha) and Asia (22 ± 1046 ha). LSI for pasture/rangeland declined on average across all continents, with the largest decrease in Europe (-0.15 ± 0.46) and smallest in South America on average (-0.01 ± 0.37). Similarly, unmanaged grass/shrubland CA expanded on average in all continents while NP and TE increased for all continents except Africa, although the decreases in these two metrics across Africa were small (-0.10 ± 0.96, and − 758 ± 9156 m for NP and TE, respectively). Net changes in the patterns of forest and cropland were particularly variable among continents; for instance, forest CA increased in Oceania, Europe and Asia but declined in South America, Africa and North America. In general, the trends in average landscape metrics were consistent through time between 1992 and 2020 in landscapes of 100 km 2 extent (Fig. S11-16), although there were exceptions to this pattern. For example, average NP of pasture/rangeland increased in North America from 1998 to 2005 but showed a consistent decline from 2005 onwards. Overall, there was significant variability in landscape pattern change from 1992 to 2020 between continents, and the direction of change at continental-scale was not always the same as at global-scale. Directional change in LULC area and fragmentation In addition to examining the global trends in landscape patterns, we investigated how often increased area of a LULC class was associated with increased fragmentation across scales. To assess changes in area and fragmentation, landscapes of 100 km 2 extent which contained a LULC class in both 1992 and 2020 were classified as to whether they showed an increase (+), decrease (-), or no change (=) in CA and LSI between 1992 and 2020. The most common pattern for all LULC classes was no net change in CA or LSI between 1992 and 2020, so the majority of landscapes were categorised as CA = LSI= (Fig. 5 and Fig. 6 ). However, the second-most common category of CA and LSI change varied between LULC classes. For instance, the second-most prevalent category for the urban and unmanaged grass/shrubland LULC classes was CA + LSI+ (19.97% and 23.22% for urban and unmanaged grass/shrubland, respectively), whereas the second-most frequent category was CA-LSI- (16.91%) for pasture/rangeland and CA-LSI + for forest (17.70%). The predominant patterns of LULC change were therefore divergent across LULC classes over the past three decades, although in multiple cases there was little difference in the frequency of each change category at global-scale; for example, for cropland the frequencies of the categories where both CA and LSI changed were 20.28% for CA + LSI-, 19.15% for CA-LSI+, 16.27% for CA-LSI-, and 15.20% for CA + LSI+. There was more variation in the prevalence of each category at continental-scale, however (Fig. S 23). For instance, more than 50% of landscapes exhibited a decrease in cropland CA in Europe across the study period (CA-LSI + = 26.36%, CA-LSI- = 26.29%, and CA-LSI = = 2.69%), whereas in Africa the majority of landscapes containing cropland showed an increase in CA from 1992 to 2020 (CA + LSI + = 22.34%, CA + LSI- = 30.60% and CA + LSI = = 3.56%). Hence, cropland was more likely to increase in landscapes in Africa and decrease in landscapes in Europe, while changes in fragmentation were split relatively evenly between increasing and declining fragmentation. There were also differences in the most common categories of CA and LSI change across landscape extents, with the CA = LSI = category becoming less frequent as landscape extent was increased at global scale (Fig. S 24). For all six LULC classes there was considerable spatial variation in the directional net change in CA and LSI within 100 km 2 landscapes which contained that LULC class in both 1992 and 2020 (Fig. 6 ). For example, cropland decreased and became more fragmented across large areas of the eastern United States of America (US), with the dominant category being CA-LSI+. Forest loss in landscapes in the southern Amazon was also predominantly associated with more fragmentation as most landscapes where forest LULC changed were in the CA-LSI + category. Loss of pasture area in Europe was more commonly classified as CA-LSI- rather than the CA-LSI+, indicating that remaining pasture/rangeland area was less fragmented in 2020 compared to 1992. Given that CA, NP, and TE of pasture/rangeland declined and TCA increased slightly on average across Europe during the study period, the apparent decrease in pasture/rangeland fragmentation may be due to the loss of small farms, which would leave large-scale farms that have larger patch size and lower fragmentation. In general, increasing area of LULC classes was distributed between the CA + LSI + and CA + LSI- categories, with increasing forest cover in northern Russia classed as CA + LSI- (increased area and decreased fragmentation) and unmanaged grass/shrubland in eastern Australia mostly classified as CA + LSI+, for instance. There were extensive changes in area and fragmentation of pasture/rangeland and unmanaged grass/shrubland across Australia, with a trend towards increased unmanaged grass/shrubland and decreased pasture cover. Discussion Overall, our findings show considerable variation in change across LULC classes between 1992 and 2020. At a global scale, the unmanaged grass/shrubland LULC class expanded and became more fragmented on average across the study period, whereas forest and pasture/rangeland declined and became less fragmented (Fig. 3 ). Global urban LULC increased in area on average but showed a decline in fragmentation per se. Urban LULC is known to have expanded globally since 1970 (van Vliet 2019 ; Güneralp et al. 2020 ; Liu et al. 2020 ), hence our findings agree with previous studies. Urban LULC change can drive both increased and decreased fragmentation of landscapes (Irwin and Bockstael 2007 ; Schneider and Woodcock 2008 ), likely due to factors such as the price of land and geographic constraints on expansion (Angel et al. 2012 ). The continued expansion and decreased fragmentation of urban LULC at global scale has implications for food security and the environment depending on whether urban areas expand into cropland or natural LULC classes. Changes in the landscape patterns of forest LULC were spatially variable, with forests increasing in area and becoming less fragmented in Europe, Asia, and Oceania but declining and fragmenting in Africa, South America and North America. Hence, forest areas were particularly prone to loss and fragmentation in tropical regions, which agrees with previous research (Hansen et al. 2013 ; Ma et al. 2023 ) and has implications for biodiversity loss (Alroy 2017 ; Giam 2017 ), carbon storage (Chaplin-Kramer et al. 2015 ; Shapiro et al. 2016 ) and carbon emissions (Brinck et al. 2017 ; Broggio et al. 2024 ). Ma et al. ( 2023 ) found that approximately 75% of global forest landscapes with exactly 25 km 2 extent became less fragmented between 2000 and 2020, whereas only 26.8% of forest landscapes with 100 km 2 extent showed decreased fragmentation between 1992 and 2020 here. There were multiple differences in methodology between the two studies that may explain this discrepancy. For example, net change in forest LSI was calculated from 1992 to 2020 in this study, whereas net change in forest fragmentation was assessed between 2000 and 2020 in (Ma et al. 2023 ). Increases in forest fragmentation from 1992 to 2000 may have counteracted decreases between 2000 and 2020, leading to a lower proportion of landscapes identified as having decreased fragmentation in this study. Similarly, input LULC maps of 1 km resolution (this study) versus 30 m resolution (Ma et al. 2023 ) were used to calculate forest landscape metrics. Coarse resolution maps are more homogeneous than fine resolution ones (Wiens 1989 ), so Ma et al. ( 2023 ) may have detected fine-scale changes in forest patterns that were apparent at 30 m but not 1 km resolution. Furthermore, different metrics were used to quantify fragmentation (LSI versus a normalised index of edge density, patch density, and mean patch area; Ma et al. 2023 ), which could have influenced the classification of landscapes into increased versus decreased fragmentation in the two studies. Indeed, Wang et al. ( 2014 ) show these indices vary in complex ways in how well they capture fragmentation independently of habitat amount. A similar definition of forest was used in both studies, with forest was defined as trees with height of ≥ 5 metres in Ma et al. ( 2023 ) and as trees with height of > 5 metres and cover ≥ 10% in the HILDA + dataset (Winkler et al., 2024 , 2021 ), but even these small differences may have contributed to the differing conclusions between the studies. The pasture/rangeland and unmanaged grass/shrubland LULC classes showed the largest changes in landscape pattern at global scale from 1992 to 2020. Pasture/rangeland is acknowledged to have declined globally since about the year 2000 (Blaustein-Rejto et al. 2019 ; Winkler et al. 2021 ), but the outcomes of pasture/rangeland declines will depend on what they are replaced by. For example, replacement of pasture/rangeland with unmanaged grass/shrublands may have positive implications for the environment such as through providing additional land for biodiversity (Poore 2016 ) and carbon sequestration (Silver et al. 2000 ), whereas replacing pasture/rangeland with urban LULC would likely not have environmental benefits. However, the replacement of pasture/rangeland with unmanaged grass/shrubland does not necessarily lead to positive outcomes. For instance, plant diversity does not always recover after land abandonment (Cava et al. 2018 ; Isbell et al. 2019 ) and land abandonment in cultural landscapes with low intensity land use can lead to biodiversity loss (Daskalova and Kamp 2023 ). Similarly, the occurrence of new unmanaged grass/shrubland patches at a distance from existing ones may increase the time taken for species to colonise the new patches, leading to much longer time lags in biodiversity recovery compared to creating new patches near to current ones (Synes et al. 2020 ). Overall, the global decrease in pasture/rangelands and increase in unmanaged grass/shrublands suggests that there are emerging opportunities for ecosystem recovery and restoration, although management actions may be needed to ensure that environmental outcomes are positive. Although unmanaged grass/shrubland appears to have become more fragmented at a global scale, difficulties in classifying LULC as pasture or rangeland versus unmanaged land make this uncertain (Phelps and Kaplan 2017 ). Moreover, the changes in the patterns of pasture/rangeland and unmanaged grass/shrubland LULC in our study are largely driven by shifts in LULC across Oceania, and particularly Australia (Fig. 4 and Fig. 6 ). Australia has extensive areas of both managed and unmanaged grasslands, which drives high uncertainty in these regions in HILDA + because the underlying input datasets use different classifications for pasture and grasslands and it is difficult to delineate unmanaged versus low intensity grasslands within heterogeneous rangeland landscapes (Winkler et al. 2021 ). Therefore, it is unclear whether the spatial patterns of pasture/rangeland and unmanaged grass/shrubland across Australia, and the rest of the globe, are real or an artifact of the HILDA + data. Further research is needed to: a) advance our capacity to identify land use intensity, and therefore distinguish between managed pastures and unmanaged grass/shrublands, in LULC datasets, and b) develop a unified approach for defining pastures and grasslands given that LULC datasets currently use different definitions to represent a range of animal production systems with differing intensities (Phelps and Kaplan 2017 ). Improved mapping and consolidated definitions of pasture/rangelands and unmanaged grass/shrublands would help to establish whether unmanaged grass/shrublands are actually becoming more fragmented at global scale, or whether the spatial pattern of unmanaged grass/shrubland change is less fragmented than suggested in HILDA+. Overall, changes in landscape patterns were highly spatially variable for all LULC classes, with differences apparent at local- to continental-scales. The spatial pattern of CA and LSI categories did not appear to be random (Fig. 5 and Fig. 6 ), as clusters of neighbouring landscapes often belonged to the same category. Therefore, the drivers of landscape pattern change likely act on local- to regional-scales, which has implications for representing landscape patterns in land use models. Moreover, when modelling LULC change, we cannot always assume that natural LULC classes become more and anthropogenic LULC classes less fragmented over time, or vice versa; instead, landscape pattern changes are spatially and temporally heterogeneous, as has been suggested by previous studies (Turner and Gardner 2015 ). We only assessed the temporal change in landscape patterns at global- and continental-scales, so future work could expand this temporal analysis by identifying regions or landscapes which showed varied temporal dynamics from 1992 to 2020, such as those where the temporal trend of fragmentation changed direction. Additional temporal analyses could support further research into the drivers of landscape pattern change across scales and how these relate to LULC change drivers, in order to better predict landscape pattern change and its impacts on the Earth System. Future work could also examine the patterns of LULC transitions in addition to changes in the patterns of individual LULC classes, to understand whether fragmenting LULC classes are being converted to natural or anthropogenic LULC classes and how this in turn impacts the environment. Disentangling the effects of habitat amount and configuration using more landscape metrics at a single landscape extent could be another focus of future research, as unfortunately none of the landscape indices showing predictable behaviour across landscape extents are also entirely uncorrelated with habitat amount, though LSI is known to be a reasonable measure of fragmentation per se between 20 and 60% class area (Wang et al. 2014 ). In conclusion, global-scale landscape pattern change demonstrated considerable heterogeneity between 1992 and 2020, which has implications for the impacts of change on environmental processes such as biodiversity and carbon emissions. Although trends in landscape metrics were generally consistent across landscape extents, the differing relationships between landscape metrics and landscape extent across LULC classes highlights the importance of considering multiple spatial scales and LULC classes when assessing landscape pattern change. The spatial heterogeneity in landscape patterns detected here suggests that landscape pattern change should be accounted for when quantifying and predicting LULC change, especially as landscape patterns are key for a range of environmental processes. Declarations Acknowledgements: TLW was funded by an EASTBIO Doctoral Training Partnership Biotechnology and Biological Sciences Research Council grant number BB/T00875X/1 and a ERC Starting Grant 'SCALEFORES' (grant no. 680176), awarded to FE. TLW and PA were supported by the UKRI projects (ForestPaths, 10039590) and (MOSAIC, 10075849). Funding TLW was funded by an EASTBIO Doctoral Training Partnership Biotechnology and Biological Sciences Research Council grant number BB/T00875X/1 and a ERC Starting Grant 'SCALEFORES' (grant no. 680176), awarded to FE. TLW and PA were supported by the UKRI projects (ForestPaths, 10039590) and (MOSAIC, 10075849). Competing interests The authors have no relevant financial or non-financial interests to disclose. Author contributions The study was conceptualised by TLW, FE, PA, and JMJT. TLW carried out all analysis with data generated and supplied by KW. The first draft of the manuscript was written by TLW and all authors contributed to revising and editing the manuscript. Data availability The HILDA+ version 2b dataset will be made publicly available on Zenodo at a later date (https://doi.org/10.5281/zenodo.15017066). The global dataset of landscape patterns generated in this study is publicly available on Zenodo (https://doi.org/10.5281/zenodo.15120267), as is the code used to generate and analyse the dataset (https://doi.org/10.5281/zenodo.15124527). References Alexander P, Prestele R, Verburg PH, et al (2017) Assessing uncertainties in land cover projections. Glob Change Biol 23:767–781. https://doi.org/10.1111/gcb.13447 Alroy J (2017) Effects of habitat disturbance on tropical forest biodiversity. Proc Natl Acad Sci USA 114:6056–6061. https://doi.org/10.1073/pnas.1611855114 Angel S, Parent J, Civco DL (2012) The fragmentation of urban landscapes: global evidence of a key attribute of the spatial structure of cities, 1990–2000. Environ Urban 24:249–283. https://doi.org/10.1177/0956247811433536 Argañaraz JP, Entraigas I (2014) Scaling functions evaluation for estimation of landscape metrics at higher resolutions. Ecol Inform 22:1–12. https://doi.org/10.1016/j.ecoinf.2014.02.004 Armenteras D, Rodríguez N, Retana J, Morales M (2011) Understanding deforestation in montane and lowland forests of the Colombian Andes. Reg Environ Change 11:693–705. https://doi.org/10.1007/s10113-010-0200-y Blaustein-Rejto D, Blomqvist L, McNamara J, de Kirby K (2019) Achieving Peak Pasture: shrinking Pasture’s footprint by spreading the livestock revolution. The Breakthrough Institute Brinck K, Fischer R, Groeneveld J, et al (2017) High resolution analysis of tropical forest fragmentation and its impact on the global carbon cycle. Nat Commun 8:14855. https://doi.org/10.1038/ncomms14855 Broggio IS, Silva-Junior CHL, Nascimento MT, et al (2024) Quantifying landscape fragmentation and forest carbon dynamics over 35 years in the Brazilian Atlantic Forest. Environ Res Lett 19:034047. https://doi.org/10.1088/1748-9326/ad281c Cattarino L, McAlpine CA, Rhodes JR (2014) Land‐use drivers of forest fragmentation vary with spatial scale. Glob Ecol Biogeogr 23:1215–1224. https://doi.org/10.1111/geb.12187 Cava MGB, Pilon NAL, Ribeiro MC, Durigan G (2018) Abandoned pastures cannot spontaneously recover the attributes of old‐growth savannas. J Appl Ecol 55:1164–1172. https://doi.org/10.1111/1365-2664.13046 Center for International Earth Science Information Network - CIESIN - Columbia University (2018) Gridded Population of the World, Version 4 (GPWv4): Population Density, Revision 11. Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). Chaplin-Kramer R, Ramler I, Sharp R, et al (2015) Degradation in carbon stocks near tropical forest edges. Nat Commun 6:10158. https://doi.org/10.1038/ncomms10158 Daskalova GN, Kamp J (2023) Abandoning land transforms biodiversity. Science 380:581–583. https://doi.org/10.1126/science.adf1099 ESA (2017) Land Cover CCI Product User Guide Version 2. Tech. Rep. Etter A, McAlpine C, Possingham H (2008) Historical patterns and drivers of landscape change in Colombia since 1500: A regionalized spatial approach. Ann Am Assoc Geogr 98:2–23. https://doi.org/10.1080/00045600701733911 Ewers RM, Laurance WF (2006) Scale-dependent patterns of deforestation in the Brazilian Amazon. Environ Conserv 33:203–211. https://doi.org/10.1017/S0376892906003250 Fahrig L (2003) Effects of Habitat Fragmentation on Biodiversity. Annu Rev Ecol Evol Syst 34:487–515. https://doi.org/10.1146/annurev.ecolsys.34.011802.132419 Fahrig L, Arroyo-Rodríguez V, Bennett JR, et al (2019) Is habitat fragmentation bad for biodiversity? Biological Conservation 230:179–186. https://doi.org/10.1016/j.biocon.2018.12.026 Fischer G, Nachtergaele FO, van Velthuizen HT, et al (2021) Global Agro-Ecological Zones (GAEZ v4) Model Documentation. FAO & IIASA Fischer J, Lindenmayer DB (2007) Landscape modification and habitat fragmentation: a synthesis. Glob Ecol Biogeogr 16:265–280. https://doi.org/10.1111/j.1466-8238.2007.00287.x Fletcher Jr RJ, Didham RK, Banks-Leite C, et al (2018) Is habitat fragmentation good for biodiversity? Biol Conserv 226:9–15. https://doi.org/10.1016/j.biocon.2018.07.022 Giam X (2017) Global biodiversity loss from tropical deforestation. Proc Natl Acad Sci USA 114:5775–5777. https://doi.org/10.1073/pnas.1706264114 González-González A, Clerici N, Quesada B (2021) Growing mining contribution to Colombian deforestation. Environ Res Lett 16:064046. https://doi.org/10.1088/1748-9326/abfcf8 Güneralp B, Reba M, Hales BU, et al (2020) Trends in urban land expansion, density, and land transitions from 1970 to 2010: a global synthesis. Environ Res Lett 15:044015. https://doi.org/10.1088/1748-9326/ab6669 Gustafson EJ (1998) Quantifying Landscape Spatial Pattern: What Is the State of the Art? Ecosyst 1:143–156. https://doi.org/10.1007/s100219900011 Gustafson EJ (2019) How has the state-of-the-art for quantification of landscape pattern advanced in the twenty-first century? Landsc Ecol 34:2065–2072. https://doi.org/10.1007/s10980-018-0709-x Haddad NM, Brudvig LA, Clobert J, et al (2015) Habitat fragmentation and its lasting impact on Earth’s ecosystems. Sci Adv 1:e1500052. https://doi.org/10.1126/sciadv.1500052 Hansen MC, Potapov PV, Moore R, et al (2013) High-Resolution Global Maps of 21st-Century Forest Cover Change. Science 342:850–853. https://doi.org/10.1126/science.1244693 Hesselbarth MHK, Sciaini M, With KA, et al (2019) landscapemetrics: an open-source R tool to calculate landscape metrics. Ecography 42:1648–1657. https://doi.org/10.1111/ecog.04617 Hijmans RJ (2022) terra: Spatial Data Analysis. R package version 1.7-23. https://rspatial.org/terra/ Hu Q, Xiang M, Chen D, et al (2020) Global cropland intensification surpassed expansion between 2000 and 2010: A spatio-temporal analysis based on GlobeLand30. Sci Total Environ 746:141035. https://doi.org/10.1016/j.scitotenv.2020.141035 Irwin EG, Bockstael NE (2007) The evolution of urban sprawl: Evidence of spatial heterogeneity and increasing land fragmentation. Proc Natl Acad Sci USA 104:20672–20677. https://doi.org/10.1073/pnas.0705527105 Isbell F, Tilman D, Reich PB, Clark AT (2019) Deficits of biodiversity and productivity linger a century after agricultural abandonment. Nat Ecol Evol 3:1533–1538. https://doi.org/10.1038/s41559-019-1012-1 Jacobson AP, Riggio J, Tait AM, Baillie JEM (2019) Global areas of low human impact (‘Low Impact Areas’) and fragmentation of the natural world. Sci Rep 9:14179. https://doi.org/10.1038/s41598-019-50558-6 Jaureguiberry P, Titeux N, Wiemers M, et al (2022) The direct drivers of recent global anthropogenic biodiversity loss. Sci Adv 2022:eabm9982. https://doi.org/10.1126/sciadv.abm9982 Jin H, Xu J, Peng Y, et al (2023) Impacts of landscape patterns on plant species diversity at a global scale. Sci Total Environ 896:165193. https://doi.org/10.1016/j.scitotenv.2023.165193 Kummu M, Taka M, Guillaume JHA (2018) Gridded global datasets for Gross Domestic Product and Human Development Index over 1990–2015. Sci Data 5:180004. https://doi.org/10.1038/sdata.2018.4 Liu X, Huang Y, Xu X, et al (2020) High-spatiotemporal-resolution mapping of global urban change from 1985 to 2015. Nat Sustain 3:564–570. https://doi.org/10.1038/s41893-020-0521-x Liu X, Li S, Wang S, et al (2022) Effects of farmland landscape pattern on spatial distribution of soil organic carbon in Lower Liaohe Plain of northeastern China. Ecol Indic 145:109652. https://doi.org/10.1016/j.ecolind.2022.109652 Lockhart J, Koper N (2018) Northern prairie songbirds are more strongly influenced by grassland configuration than grassland amount. Landscape Ecol 33:1543–1558. https://doi.org/10.1007/s10980-018-0681-5 Ma J, Li J, Wu W, Liu J (2023) Global forest fragmentation change from 2000 to 2020. Nat Commun 14:3752. https://doi.org/10.1038/s41467-023-39221-x Martello F, dos Santos JS, Silva-Neto CM, et al (2023) Landscape structure shapes the diversity of plant reproductive traits in agricultural landscapes in the Brazilian Cerrado. Agric Ecosyst Environ 341:108216. https://doi.org/10.1016/j.agee.2022.108216 Massicotte P, South A (2023) rnaturalearth: World Map Data from Natural Earth. R package version 0.3.2. https://CRAN.R-project.org/package=rnaturalearth Maxwell SL, Fuller RA, Brooks TM, Watson JEM (2016) Biodiversity: The ravages of guns, nets and bulldozers. Nature 536:143–145. https://doi.org/10.1038/536143a Miguet P, Jackson HB, Jackson ND, et al (2016) What determines the spatial extent of landscape effects on species? Landsc Ecol 31:1177–1194. https://doi.org/10.1007/s10980-015-0314-1 Pais C, Miranda A, Carrasco J, Shen Z-JM (2021) Deep fire topology: Understanding the role of landscape spatial patterns in wildfire occurrence using artificial intelligence. Environ Model Softw 143:105122. https://doi.org/10.1016/j.envsoft.2021.105122 Phelps LN, Kaplan JO (2017) Land use for animal production in global change studies: Defining and characterizing a framework. Glob Change Biol 23:4457–4471. https://doi.org/10.1111/gcb.13732 Poore JAC (2016) Call for conservation: Abandoned pasture. Science 351:132. https://doi.org/10.1126/science.351.6269.132-a Potapov P, Turubanova S, Hansen MC, et al (2021) Global maps of cropland extent and change show accelerated cropland expansion in the twenty-first century. Nat Food 3:19–28. https://doi.org/10.1038/s43016-021-00429-z R Core Team (2022) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/. R Core Team (2020) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/. Riitters K, Wickham J, O’Neill R, et al (2000) Global-scale patterns of forest fragmentation. Conserv Ecol 4:3 Ryu S-R, Chen J, Zheng D, Lacroix JJ (2007) Relating surface fire spread to landscape structure: An application of FARSITE in a managed forest landscape. Landsc Urban Plan 83:275–283. https://doi.org/10.1016/j.landurbplan.2007.05.002 Schneider A, Woodcock CE (2008) Compact, Dispersed, Fragmented, Extensive? A Comparison of Urban Growth in Twenty-five Global Cities using Remotely Sensed Data, Pattern Metrics and Census Information. Urban Stud 45:659–692. https://doi.org/10.1177/0042098007087340 Shapiro AC, Aguilar-Amuchastegui N, Hostert P, Bastin J-F (2016) Using fragmentation to assess degradation of forest edges in Democratic Republic of Congo. Carbon Balance Manage 11:11. https://doi.org/10.1186/s13021-016-0054-9 Silver WL, Ostertag R, Lugo AE (2000) The Potential for Carbon Sequestration Through Reforestation of Abandoned Tropical Agricultural and Pasture Lands. Restor Ecol 8:394–407. https://doi.org/10.1046/j.1526-100x.2000.80054.x Šímová P, Gdulová K (2012) Landscape indices behavior: A review of scale effects. Appl Geogr 34:385–394. https://doi.org/10.1016/j.apgeog.2012.01.003 Sponseller RA, Benfield EF, Valett HM (2008) Relationships between land use, spatial scale and stream macroinvertebrate communities. Freshw Biol 46:1409–1424. https://doi.org/10.1046/j.1365-2427.2001.00758.x Synes NW, Ponchon A, Palmer SCF, et al (2020) Prioritising conservation actions for biodiversity: Lessening the impact from habitat fragmentation and climate change. Biol Conserv 252:108819. https://doi.org/10.1016/j.biocon.2020.108819 Turner MG (1989) Landscape Ecology: The Effect of Pattern on Process. Annu Rev Ecol Evol Syst 20:171–97. https://doi.org/10.1146/annurev.es.20.110189.001131 Turner MG, Gardner RH (2015) Causes of Landscape Pattern. In: Landscape Ecology in Theory and Practice. Springer, New York, NY, pp 33–62 Uuemaa E, Roosaare J, Mander Ü (2005) Scale dependence of landscape metrics and their indicatory value for nutrient and organic matter losses from catchments. Ecol Indic 5:350–369. https://doi.org/10.1016/j.ecolind.2005.03.009 van Vliet J (2019) Direct and indirect loss of natural area from urban expansion. Nat Sustain 2:755–763. https://doi.org/10.1038/s41893-019-0340-0 Wang X, Blanchet FG, Koper N (2014) Measuring habitat fragmentation: An evaluation of landscape pattern metrics. Methods Ecol Evol 5:634–646. https://doi.org/10.1111/2041-210X.12198 Wiens JA (1989) Spatial Scaling in Ecology. Funct Ecol 3:385–397. https://doi.org/10.2307/2389612 Winkler K, Fuchs R, Rounsevell M, Herold M (2021) Global land use changes are four times greater than previously estimated. Nat Commun 12:1–10. https://doi.org/10.1038/s41467-021-22702-2 Winkler K, Fuchs R, Rounsevell M, Herold M (2024) HILDA+ version 2.0: Global Land Use Change between 1960 and 2020 [dataset]. PANGAEA (in review) Winkler K, Fuchs R, Rounsevell MDA, Herold M (2020) HILDA+ Global Land Use Change between 1960 and 2019 [dataset] Woodman TL, Arendarczyk B, Winkler K, et al (2025) Harmonised global land use and land cover maps between 1960 and 2100 [dataset] Woodman TL, Rueda-Uribe C, Henry RC, et al (2023) Introducing LandScaleR: A novel method for spatial downscaling of land use projections. Environ Model Softw 169:105826. https://doi.org/10.1016/j.envsoft.2023.105826 Wu J (2004) Effects of changing scale on landscape pattern analysis: scaling relations. Landsc Ecol 19:125–138. https://doi.org/10.1023/B:LAND.0000021711.40074.ae Wu J, Shen W, Sun W, Tueller PT (2002) Empirical patterns of the effects of changing scale on landscape metrics. Landsc Ecol 17:761–782. https://doi.org/10.1023/A:1022995922992 Additional Declarations No competing interests reported. Supplementary Files SupplementaryInformationGloballandscapepatterns.pdf Cite Share Download PDF Status: Published Journal Publication published 16 Oct, 2025 Read the published version in Landscape Ecology → Version 1 posted Editorial decision: Revision requested 22 Jun, 2025 Reviews received at journal 20 Jun, 2025 Reviews received at journal 18 Jun, 2025 Reviewers agreed at journal 07 Jun, 2025 Reviewers agreed at journal 06 Jun, 2025 Reviews received at journal 18 May, 2025 Reviewers agreed at journal 06 May, 2025 Reviewers agreed at journal 28 Apr, 2025 Reviewers invited by journal 27 Apr, 2025 Editor assigned by journal 24 Apr, 2025 Submission checks completed at journal 24 Apr, 2025 First submitted to journal 24 Apr, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6522006","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":449110025,"identity":"848b4d04-6f9d-42a3-9b6e-64406b508e13","order_by":0,"name":"Tamsin L. Woodman","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7UlEQVRIiWNgGAWjYLCCBAhlAMQ2DAwHkMWwAB40LWlEamFAaDlMWIs9++nkDw93MCT2szdv3fBzx3k5vgPMDz8wtqXhtoUnd5tE4hmGxJk9x8pu9p65bSx5gM1YgrEtB4/DcrcxJLYx5G64kWN2g7ftduKGAwxmDIxtFbi18L/d/AGm5ebftnP1Gw6wf8OvRSJ3gwRMy23etgMJBgd4QLbgcdiNt0C/tEnUg/xyW7Yt2XDmYZ5iiYRzuL3P3p+7+ePPNhtjfvbmbTffttnJ8x1v3/jhQ1kyTi1QIIHEZmbAFyujYBSMglEwCogBAIUoWHzfU9HkAAAAAElFTkSuQmCC","orcid":"","institution":"School of Biological Sciences, University of Aberdeen","correspondingAuthor":true,"prefix":"","firstName":"Tamsin","middleName":"L.","lastName":"Woodman","suffix":""},{"id":449110026,"identity":"8e86b12e-c5d6-4d0e-84a0-8c8c0e3b455b","order_by":1,"name":"Peter Alexander","email":"","orcid":"","institution":"School of GeoSciences, University of Edinburgh","correspondingAuthor":false,"prefix":"","firstName":"Peter","middleName":"","lastName":"Alexander","suffix":""},{"id":449110027,"identity":"811ca145-66b6-474a-a8fe-04b582973474","order_by":2,"name":"David F.R.P. Burslem","email":"","orcid":"","institution":"School of Biological Sciences, University of Aberdeen","correspondingAuthor":false,"prefix":"","firstName":"David","middleName":"F.R.P.","lastName":"Burslem","suffix":""},{"id":449110028,"identity":"8b13c76a-8480-4016-a13d-96affa11ac61","order_by":3,"name":"Justin M.J. Travis","email":"","orcid":"","institution":"School of Biological Sciences, University of Aberdeen","correspondingAuthor":false,"prefix":"","firstName":"Justin","middleName":"M.J.","lastName":"Travis","suffix":""},{"id":449110029,"identity":"d1680329-72a9-461f-b891-a5fca793f67f","order_by":4,"name":"Karina Winkler","email":"","orcid":"","institution":"Land Use Change and Climate, IMKIFU, Karlsruhe Institute of Technology (KIT)","correspondingAuthor":false,"prefix":"","firstName":"Karina","middleName":"","lastName":"Winkler","suffix":""},{"id":449110030,"identity":"c5137f5d-f6d2-494a-b804-1b866c344ab8","order_by":5,"name":"Felix Eigenbrod","email":"","orcid":"","institution":"School of Geography and Environmental Science, University of Southampton","correspondingAuthor":false,"prefix":"","firstName":"Felix","middleName":"","lastName":"Eigenbrod","suffix":""}],"badges":[],"createdAt":"2025-04-24 15:08:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6522006/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6522006/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10980-025-02210-0","type":"published","date":"2025-10-16T15:56:59+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":82091967,"identity":"3945eecf-a707-4210-a8e2-708589b060de","added_by":"auto","created_at":"2025-05-06 16:24:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":473435,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eExample landscapes for Colombia with extents of a) 400 km\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003cem\u003e and b) 6400 km\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003cem\u003e. Landscapes are depicted using black lines and are overlaid on land use and land cover maps from HILDA+ in 1992\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/ce9d2c438041e336c26b9f77.png"},{"id":82092178,"identity":"650dfa16-d6b5-45a9-823d-7ae83e02536c","added_by":"auto","created_at":"2025-05-06 16:32:20","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":270642,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eScaling relationships for six landscape metrics at global scale. Lines give the mean value of a metric across landscapes of different extents for one LULC class. Landscape extent is the length of each side of a landscape in kilometres. CA = class area (hectares), LSI = Landscape Shape Index, NDCA = number of disjunct core area patches, NP = number of patches, TCA = total core area (hectares), TE = total edge length (m)\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/7d3b58290176725d5bafb629.png"},{"id":82091969,"identity":"8b109712-e80c-4e10-885f-ce6843455717","added_by":"auto","created_at":"2025-05-06 16:24:20","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":345070,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eAverage of global-scale landscape patterns in 100 km\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003cem\u003e (a) and 25600 km\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003cem\u003e (b) extent landscapes from 1992 to 2020. Lines give the mean of one landscape metric for one LULC class across \u003c/em\u003e\u003csup\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/sup\u003e\u003cem\u003elandscapes of that extent in every year from 1992 to 2020. See Fig. 2 for landscape pattern definitions and units. Standard deviations were large (minimum 0.15 for NDCA and maximum 18185 m for TE in 100 km\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003cem\u003e landscapes, for example) and are plotted separately in Fig. S 6 and Fig. S 10\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/2ff1b0871bae250fdcdaca8d.png"},{"id":82091971,"identity":"93e9481e-e0a9-43d3-a25d-a9d396cbbbfc","added_by":"auto","created_at":"2025-05-06 16:24:20","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":265970,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eAverage net change in landscape metrics across continents from 1992 to 2020. Average net change is shown for six LULC classes for landscapes of 100 km\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003cem\u003e extent. See Fig. 2 for descriptions of landscape metrics and units\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/1caea686a3f81c04895573e3.png"},{"id":82092181,"identity":"a93a2b38-2a1e-4d47-9b3c-192b556e8017","added_by":"auto","created_at":"2025-05-06 16:32:20","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":388876,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eDirection of net change in area and fragmentation in 100 km\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003cem\u003e landscapes from 1992 to 2020. Direction of net change is shown for landscapes which contained the land use and land cover (LULC) class of interest in both 1992 and 2020. CA = class area, LSI = Landscape Shape Index. The square inset in each panel shows the relative proportion of landscapes assigned to each of the nine categories of CA and LSI change for that LULC class. Grey shading indicates the absence of a LULC class in a landscape in both 1992 and 2020. Note that LSI+ indicates increased fragmentation of a LULC class and LSI- represents decreased fragmentation\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/ed0b873c3e372c95ddcac6ac.png"},{"id":82091976,"identity":"a70a3d13-0898-41f5-acf5-96b198c2c396","added_by":"auto","created_at":"2025-05-06 16:24:20","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":641824,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eDirection of net change in area and fragmentation from 1992 to 2020 for four example regions. Directional change is shown for a) forest in the southern Amazon, b) cropland in the United States of America, c) pasture/rangeland across central and eastern Europe, and d) pasture/rangeland plus unmanaged grass/shrubland in Australia. CA = class area, LSI = Landscape Shape Index. See Fig. 5 for further information on CA and LSI categories\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/e35d4d10ca0d29b4fbc03f4e.png"},{"id":93955960,"identity":"b5c6c6f6-69ae-46eb-9d26-a8c4dc96dd41","added_by":"auto","created_at":"2025-10-20 16:08:14","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3117981,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/dd949804-34b1-4b1f-a014-74c2cf23668f.pdf"},{"id":82092185,"identity":"5baf462d-3bfd-42b5-ae49-0cd652a63d6d","added_by":"auto","created_at":"2025-05-06 16:32:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":6492272,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformationGloballandscapepatterns.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6522006/v1/6e50754a39ca20906deba0e8.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Global assessment of landscape pattern changes from 1992 to 2020","fulltext":[{"header":"Introduction","content":"\u003cp\u003eUnderstanding how global landscape patterns have fluctuated through time is key for deciphering the drivers of changes and how these changes impact the Earth System. Anthropogenic land use and land cover (LULC) change is driving the loss and fragmentation of the Earth\u0026rsquo;s remaining natural ecosystems, with potentially damaging consequences for the Earth System (Haddad et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Approximately 2.3\u0026nbsp;million km\u003csup\u003e2\u003c/sup\u003e of forest was lost between 2000 and 2012 (Hansen et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), for example, and over 70% of remaining forest cover is within 1 km of a forest edge (Haddad et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). By contrast, the amount of cropland has increased 9% from 2003 to 2019 (Potapov et al. \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). LULC change alone can have adverse effects on the environment, such as through driving biodiversity loss (Maxwell et al. \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Jaureguiberry et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Changes in landscape patterns can also have negative environmental impacts; for example, the increasing amount of tropical forest edge habitat has been estimated to release 0.34 Gt of carbon per year (Brinck et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Moreover, landscape patterns are important for a range of environmental processes, including the movement of organisms (Fischer and Lindenmayer \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2007\u003c/span\u003e), fire spread (Ryu et al. \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) and ignition (Pais et al. \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), and accumulation of soil carbon (Liu et al. \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Therefore, knowledge of how LULC has changed historically in terms of both area and pattern is important for estimating the impacts of LULC change on the Earth System.\u003c/p\u003e \u003cp\u003eLandscape patterns are generated by a complicated array of natural and anthropogenic factors, such as anthropogenic LULC change, climate, soil properties, and biotic interactions (Turner and Gardner \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). LULC change leads to changes in both the amount (composition) and configuration (spatial arrangement) of LULC classes, and usually also leads to smaller individual patches of a specific LULC class. The process of generating smaller LULC patches, which affects both the amount and configuration of LULC classes simultaneously, is often referred to as the process of fragmentation. However, fragmentation can also refer to changes in the spatial arrangement for a given area of a LULC class in a landscape (\u0026lsquo;fragmentation per se\u0026rsquo;; sensu (Fahrig \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). These differences in how fragmentation is defined, combined with varied correlation structures between measures of landscape structure at the patch versus landscape level, make it difficult to compare between studies and have led to considerable disagreement around the impacts of fragmentation on biodiversity (see (Fletcher Jr et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Fahrig et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Although fragmentation has been recognised as an important process of environmental change, the distribution and trends of fragmentation patterns for each LULC category remain unclear.\u003c/p\u003e \u003cp\u003eThe drivers of landscape pattern change and the impact of landscape patterns on the environment may differ across scales (Ewers and Laurance \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Cattarino et al. \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Jin et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), hence the choice of scale on which to study the effects of landscape patterns can have considerable influence on the conclusions from a study (Turner \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e1989\u003c/span\u003e; Miguet et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). For example, indices of macroinvertebrate richness were most closely associated with landscape patterns calculated at a scale of 200 m-wide riparian corridors (Sponseller et al. \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), suggesting that this scale would be most appropriate for studying the effects of landscape patterns on macroinvertebrate diversity. Similarly, the relationships between landscape patterns and plant diversity are strongly dependent on the scale at which landscape patterns are quantified (Martello et al. \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Jin et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The choice of scale on which landscape metrics, which are used to quantify landscape patterns (Gustafson \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e1998\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), are calculated is therefore important as it may affect how landscape patterns appear to relate to their drivers and to environmental processes. Calculating landscape metrics at a range of scales may be preferable to better understand their cross-scale associations with landscape pattern drivers and environmental processes (Miguet et al. \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAlthough it is important to study landscape patterns at multiple scales, there are relatively few landscape metrics which behave predictably as the scale of a landscape increases in terms of both extent and resolution (Turner \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e1989\u003c/span\u003e; Wu et al. \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Wu \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Uuemaa et al. \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Arga\u0026ntilde;araz and Entraigas \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The lack of landscape metrics with predictable relationships across scales may impede our ability to make cross-scale comparisons of the drivers and impacts of landscape patterns. An increasing number of landscape metrics have become available in recent years compared to those tested for scaling relationships in previous studies (Wu et al. \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Wu \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Š\u0026iacute;mov\u0026aacute; and Gdulov\u0026aacute; \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), which suggests there is an opportunity to assess the scaling behaviour of these new landscape metrics and update our knowledge of the behaviour of landscape metrics across scales.\u003c/p\u003e \u003cp\u003eThe choice of landscape metrics and definition of fragmentation used to assess landscape patterns at a global-scale has varied between studies, making it hard to compare studies that have focused on a single LULC class or landscape extent (for example, (Haddad et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Hu et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Ma et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). To our knowledge, previous estimations of global landscape pattern change have focused on forest or cropland patterns only (Riitters et al. \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; Haddad et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Hu et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Ma et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), or quantified change at the level of entire landscapes rather than for individual LULC classes (Jacobson et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Moreover, few studies have assessed global-scale landscape patterns across scales, except for (Riitters et al. \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2000\u003c/span\u003e) who calculated global forest fragmentation across four spatial scales for a single time point. Although natural land cover classes such as forest are thought to have decreased and become more fragmented through time (Hansen et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Haddad et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Jacobson et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), a more recent study indicated that the majority of forested landscapes across the globe may have exhibited a trend of declining fragmentation between 2000 and 2020 (Ma et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). However, the metrics used to quantify fragmentation differ between studies; for instance, Haddad et al. (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) utilised distance to edge, number of fragments and fragment area to quantify global forest fragmentation, whereas Ma et al. (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) employed a fragmentation index constructed from edge density, patch density, and mean patch area. Consequently, there is a need for global-scale landscape pattern change assessments which encompass multiple LULC classes and landscape extents to give a better understanding of how landscape patterns are changing over time.\u003c/p\u003e \u003cp\u003eHere, we address this research gap and quantify global-scale landscape patterns for several LULC classes, and assess how landscape patterns have changed over recent decades. This is the first study to quantify global landscape patterns for multiple LULC classes, landscape extents, and years. Given that most class-level landscape metrics show unpredictable scaling relationships across landscape extents (Wu \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2004\u003c/span\u003e), we first use a single country, Colombia, which has highly heterogeneous landscape patterns, to identify metrics with predictable scaling relationships when landscape extent increases. The selected metrics are then used to calculate global landscape patterns for a range of landscape extents between 1992 and 2020, to address how global landscape patterns have changed over time. We also attempt to identify where different LULC classes have shown changes in both area and fragmentation per se over the study period. Our study intends to generate broader knowledge of the overall trends and spatial variability in global-scale landscape patterns over the past three decades, which has implications for understanding the drivers of landscape pattern change and their impacts on the Earth System.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eLand use and land cover data\u003c/h2\u003e \u003cp\u003eWe used land use and land cover (LULC) data from the HILDA\u0026thinsp;+\u0026thinsp;version 2b dataset in the Eckert IV projection for the calculation of landscape metrics at global scale (Winkler et al. \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Woodman et al. \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). HILDA\u0026thinsp;+\u0026thinsp;provides yearly 1 km spatial resolution LULC data from 1960 to 2020. Each grid cell contains a single LULC class which is derived from an aggregation of multiple LULC maps and other related datasets, such as FAO land use statistics (Winkler et al. \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). There are six LULC classes in HILDA\u0026thinsp;+\u0026thinsp;version 2b: urban, cropland, pasture/rangeland, forest, unmanaged grass/shrubland and sparse/no vegetation. HILDA\u0026thinsp;+\u0026thinsp;distinguishes between managed pasture/rangelands and unmanaged grass/shrublands, giving it an advantage over other global LULC datasets that treat managed and unmanaged grasslands as the same LULC class. Pasture/rangeland is defined in HILDA\u0026thinsp;+\u0026thinsp;as managed herbaceous plants with at least 10% cover, including areas that are used for livestock and hay production. Unmanaged grass/shrublands are natural herbaceous plants with at least 10% cover that are not managed by people, including wetland areas. Both the pasture/rangeland and unmanaged grass/shrubland classes include mosaics of herbaceous plants with trees and shrubs. A grid cell must have at least 10% cover of trees that are taller than 5 metres to be classed as forest (Winkler et al. \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). We restricted the study period to between 1992 and 2020 because 1992 is the first year that a high resolution, yearly LULC dataset (the ESA CCI Land Cover time series) is used as input to HILDA+ (ESA \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Winkler et al. \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). For global analysis we cropped the HILDA\u0026thinsp;+\u0026thinsp;maps to exclude the continent of Antarctica and sub-Antarctic islands as very little LULC change occurred here during the study period. The HILDA\u0026thinsp;+\u0026thinsp;LULC maps were cropped in R software version 4.1.3 (R Core Team \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) using the \u0026lsquo;terra\u0026rsquo; R package version 1.7\u0026ndash;23 (Hijmans \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) and an outline of Antarctica from the \u0026lsquo;rnaturalearth\u0026rsquo; package version 0.3.2 (Massicotte and South \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eSelection of landscape metrics\u003c/h3\u003e\n\u003cp\u003eA key goal of our study was to look at changes in LULC classes across multiple spatial extents, for which we required class-level landscape metrics with consistent behaviour across scales. To achieve this, we first assessed the behaviour of class-level landscape metrics from the \u0026lsquo;landscapemetrics\u0026rsquo; R package (Hesselbarth et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) with increasing landscape extent for Colombia, before moving to global level analyses. There have been large changes in LULC over time in Colombia, and the rate and drivers of LULC change have varied both spatially and temporally. For instance, in the late twentieth century the Andean region experienced the highest rates of deforestation (Etter et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Current drivers of LULC change in Colombia include clearing of forests for cattle grazing, legal and illegal crop production, mining, and urbanization (Etter et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Armenteras et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Gonz\u0026aacute;lez-Gonz\u0026aacute;lez et al. \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Given its large area, diverse land covers and varied drivers of LULC change, all six LULC classes from HILDA\u0026thinsp;+\u0026thinsp;were represented in Colombia (Winkler et al. \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2020\u003c/span\u003e, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Woodman et al. \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Colombia was therefore considered a suitable case study to test the scaling relationships of landscape metrics across ten landscape extents (100, 400, 900, 1600, 2500, 3600, 4900, 6400, 8100 and 10000 km\u003csup\u003e2\u003c/sup\u003e).\u003c/p\u003e \u003cp\u003eFirst, an outline of Colombia from the \u0026lsquo;rnaturalearth\u0026rsquo; version 0.3.2 package (Massicotte and South \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) was used to crop the HILDA\u0026thinsp;+\u0026thinsp;dataset from 1992 to 2020 to the same extent as Colombia. Next, a set of ten regular grids covering the terrestrial surface of Colombia were created to represent landscapes with different extents. Each grid had landscapes (grid cells) with sides of between 10 and 100 km length at 10 km increments, giving a total of ten grids. Each grid was overlaid with LULC in Colombia from HILDA\u0026thinsp;+\u0026thinsp;and all grid cells that were entirely classified as ocean by HILDA\u0026thinsp;+\u0026thinsp;in every year from 1992 to 2020 were removed. Each individual cell in a grid was treated as a landscape for the calculation of landscape metrics. Two examples of landscapes with different extents (sides of length 20 km and 80 km, or 400 km\u003csup\u003e2\u003c/sup\u003e and 6400 km\u003csup\u003e2\u003c/sup\u003e landscape extent, respectively) overlaid on LULC in Colombia in 1992 are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Next, all 55 class-level landscape metrics implemented in the \u0026lsquo;landscapemetrics\u0026rsquo; R package version 1.5.6 (Hesselbarth et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) were calculated in the first year of the study period (1992) for each landscape in the ten grids.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAfter calculating class-level landscape metrics using the \u0026lsquo;landscapemetrics\u0026rsquo; R package, we analysed the scaling relationship between the mean of each landscape metric and landscape extent. First, we removed the two HILDA\u0026thinsp;+\u0026thinsp;water LULC classes (ocean and water) from the dataset of landscape metrics in Colombia in 1992. The \u0026lsquo;landscapemetrics\u0026rsquo; R package does not return a value for class-level landscape metrics when a LULC class is not present in a landscape (Hesselbarth et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), which resulted in missing values for metrics in many landscapes. It is not possible to calculate a value for the majority of landscape metrics when a LULC class is not present in a landscape, so we treated missing values for these metrics as missing when summarising the dataset (Table S 1). However, for other metrics, such as class area (CA) and total edge length (TE), the value of the metric is equivalent to zero when a LULC class is not present in a landscape. Therefore, we replaced missing values with zeros for eleven out of the 55 landscape metrics available in the \u0026lsquo;landscapemetrics\u0026rsquo; package (Hesselbarth et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The mean of each landscape metric was then calculated across all landscapes for every landscape extent, and the mean of landscape metrics as predicted by landscape extent in terms of the length of each landscape in kilometres was plotted (as in (Wu \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2004\u003c/span\u003e). The resulting plots were examined to establish which landscape metrics showed consistent scaling relationships as landscape extent increased. Six landscape metrics were found to have predictable scaling behaviour as landscape extent increased (Fig. S 1, Fig. S 2, Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe six landscape metrics which demonstrated predictable scaling with landscape extent across Colombia were: class area (CA), Landscape Shape Index (LSI), number of disjunct core area patches (NDCA), number of patches (NP), total core area (TCA), and total edge length (TE). Four of these six metrics were previously identified as having consistent behaviour as landscape extent was increased across a set of landscapes in the United States (Wu \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2004\u003c/span\u003e). CA is the total area of a LULC class in one landscape in hectares, TE is the total edge length of a LULC class in a landscape in metres, and NP gives the number of non-contiguous patches of a LULC class. NDCA and TCA are both core area metrics, where \u0026lsquo;core area\u0026rsquo; consists of grid cells which are surrounded by cells of the same class. NDCA counts the number of non-contiguous core area patches for a LULC within a landscape, meaning it is a measure of the number of \u0026lsquo;patches within patches\u0026rsquo;. Meanwhile, TCA is the total area of a LULC class that can be considered as the core area within a landscape, with units of hectares. LSI is calculated as a ratio of the total edge length of a LULC class to the hypothetical minimum edge length of that class if it was as aggregated as possible; hence, LSI quantifies the fragmentation of a LULC class within a landscape, with lower values indicating lower fragmentation (Hesselbarth et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). These six landscape metrics therefore encompass measures of LULC area, edge length, core area, and fragmentation.\u003c/p\u003e\n\u003ch3\u003eCalculating landscape metrics at global scale\u003c/h3\u003e\n\u003cp\u003eThe six landscape metrics which showed consistent scaling relationships with increasing landscape extent across Colombia in 1992 (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) were calculated at global scale to create a cohesive dataset of landscape patterns for multiple LULC classes and landscape extents. We decided to calculate global landscape metrics for five landscape extents: 10 by 10 km, 20 by 20 km, 40 by 40 km, 80 by 80 km, and 160 by 160 km (100, 400, 1600, 6400 and 25600 km\u003csup\u003e2\u003c/sup\u003e, respectively). The 10 by 10 km (i.e. 100 km\u003csup\u003e2\u003c/sup\u003e) extent was chosen because it approximately matches the resolution of available socioeconomic datasets (for example: (Center for International Earth Science Information Network - CIESIN - Columbia University \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Kummu et al. \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Fischer et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) that could be utilised in future to investigate drivers of landscape patterns. The further four landscape extents were selected by doubling the number of kilometres per side of a landscape. Most global land use models typically generate projections at coarse resolutions (for example, in regions or 0.5\u0026deg; grids; Alexander et al. \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), so the 1600, 6400 and 25600 km\u003csup\u003e2\u003c/sup\u003e landscapes (40, 80, and 160 km per side, respectively) aim to approximate the range of outputs obtained from global land use models. One grid covering the terrestrial surface of the Earth, excluding Antarctica, was created per landscape extent using the same method as for creating landscapes across Colombia. Each of the six class-level landscape metrics were then calculated within each landscape in the five global-scale grids using the \u0026lsquo;landscapemetrics\u0026rsquo; R package version 1.5.6 (Hesselbarth et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) in R version 4.0.0 (R Core Team \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) for all years in the study period (1992 to 2020). We assessed whether the selected landscape metrics showed predictable scaling across landscape extents at global scale using the same method as for Colombia (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClass-level landscape metrics used to quantify global-scale landscape patterns. All six metrics showed consistent scaling relationships with increasing landscape extent across Colombia from 1992. Description of metrics based on Hesselbarth et al. (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLandscape metric\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAbbreviation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eUnits\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClass area\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHectares\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTotal area of a LULC class in a landscape.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLandscape Shape Index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLSI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRatio between the actual edge length of a LULC class in a landscape and its theoretical minimum edge length if it were as aggregated as possible.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of disjunct core area patches\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNDCA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNumber of core area patches of a LULC class in a landscape, where a core area patch has no neighbouring cells of another LULC class.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of patches\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNumber of patches of a LULC class in a landscape.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal core area\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTCA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHectares\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTotal core area of a LULC class in a landscape, where core area is made up of grid cells with no neighbouring cells of a different LULC class.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal edge length\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMetres\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTotal edge length of a LULC class in a landscape.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003eAnalysis of landscape pattern change\u003c/h3\u003e\n\u003cp\u003eThe dataset of global landscape patterns from 1992 to 2020 was used to analyse landscape pattern change of six LULC classes through time. Given the large number of analyses carried out, we do not present all the results for all landscape extents in the main Results. However, as all six landscape metrics show consistent scaling relationships globally (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) we expect our conclusions will be applicable across scales. We focused our analyses on the smallest landscape extent to increase the sample size for subregions (i.e. continents) and also to better characterize relatively rare land cover classes, such as urban land cover (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Prior to analysing landscape pattern change during the study period, we first processed landscape patterns in the same way as for assessing the scaling relationships of landscape metrics. First, we removed HILDA\u0026thinsp;+\u0026thinsp;classes representing water from the dataset. Next, missing values were replaced with zero for any landscape metrics where we considered a missing value to be equivalent to zero (Table S 1).\u003c/p\u003e \u003cp\u003eLandscape metrics were summarised by calculating the mean and standard deviation across all landscapes of a given extent in each year for every LULC class. Calculating the mean and standard deviation allowed us to assess the average and variation in landscape patterns across the globe from 1992 and 2020, and whether the trends were consistent through time. Additionally, we assessed the spatiotemporal variation in landscape pattern change by calculating the net change within each landscape between 1992 and 2020 for every landscape metric and LULC class. Net change was plotted for each metric and LULC class for 100 km\u003csup\u003e2\u003c/sup\u003e extent landscapes only to evaluate whether the magnitude and direction of change varied between landscapes and regions (Fig. S 29\u0026ndash;34). We also calculated the average net change in landscapes for each continent to test for differences between continental- and global-scale landscape pattern change.\u003c/p\u003e \u003cp\u003eTo disentangle the impacts of changes in the area and configuration of LULC classes, we focused on two particular indices: class area (CA), which measures the area of a LULC class in a landscape, and Landscape Shape Index (LSI), which measures the fragmentation of a LULC class within a landscape (Hesselbarth et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). LSI is a good measure of changes in configuration independently of CA because it calculates edge length in relation to a hypothetical minimum edge length, and hence accounts for the area of a LULC class within the landscape. A high LSI value indicates a LULC class with high fragmentation per se, as the actual edge length is much longer than the hypothetical minimum if the LULC class was as aggregated as possible. Comparatively, low LSI indicates that a LULC class has low fragmentation per se. Overall, LSI provides a better measure of configuration independently of CA than total edge (TE), the patch metric NP, or the core area metrics NCDA and TCA, as all of these are known to be correlated with class area at the landscape scale (Wang et al. \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Lockhart and Koper \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo identify landscapes where both CA and LSI were changing in the same direction for one LULC class, we first selected all landscapes of 100 km\u003csup\u003e2\u003c/sup\u003e extent which contained that class in both 1992 and 2020. Next, the selected landscapes were classified into nine categories: both CA and LSI increasing (CA\u0026thinsp;+\u0026thinsp;LSI+); CA increasing and LSI decreasing (CA\u0026thinsp;+\u0026thinsp;LSI-); CA increasing and no change in LSI (CA\u0026thinsp;+\u0026thinsp;LSI=); CA decreasing and LSI increasing (CA-LSI+); both CA and LSI decreasing (CA-LSI-); CA decreasing and no change in LSI (CA-LSI=); no change in CA and increasing LSI (CA\u0026thinsp;=\u0026thinsp;LSI+); no change in CA and decreasing LSI (CA\u0026thinsp;=\u0026thinsp;LSI-), and no change in either CA or LSI (CA\u0026thinsp;=\u0026thinsp;LSI=). Maps of the nine categories of CA and LSI change were created to examine whether there was a tendency for regions and LULC classes to become more fragmented over time as LULC area changed. The percentage of landscapes in each of the nine categories was calculated at global- and continental scales for each landscape extent, to test whether the prevailing direction of change differed between LULC classes and across scales.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eScaling relationships of landscape patterns\u003c/h2\u003e \u003cp\u003eAll six class-level landscape metrics identified as showing consistent scaling relationships when landscape extent was increased across Colombia from 1992 were also confirmed to exhibit predictable behaviour with increasing landscape extent at global scale (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Five out of the six metrics appeared to exhibit a power-law relationship with landscape extent, whereas LSI had an approximately linear relationship with landscape extent. The relationship between each metric and landscape extent followed the same pattern across LULC classes, although the rate of increase differed between LULC classes. For example, forest CA increased much more rapidly with landscape extent compared to urban CA, likely because urban areas will not cover more than a small fraction of a 25600 km\u003csup\u003e2\u003c/sup\u003e landscape.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eGlobal trends in landscape patterns\u003c/h3\u003e\n\u003cp\u003eThe mean values for the six landscape metrics in landscapes of 100 and 25600 km\u003csup\u003e2\u003c/sup\u003e extent varied by LULC class at global scale from 1992 to 2020 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For example, the mean area of unmanaged grass/shrubland increased from 1362\u0026thinsp;\u0026plusmn;\u0026thinsp;2545 ha (mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation) in 1992 to 1492\u0026thinsp;\u0026plusmn;\u0026thinsp;2658 ha in 2020 in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes, with a corresponding increase in all other metrics. Similarly, there were increases in five out of six landscape metrics for urban LULC in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes, with only LSI exhibiting a small decrease from 1.35\u0026thinsp;\u0026plusmn;\u0026thinsp;0.45 in 1992 to 1.34\u0026thinsp;\u0026plusmn;\u0026thinsp;0.45 in 2020. Therefore, unmanaged grass/shrubland and urban land cover both increased on average in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes between 1992 and 2020, which coincided with increasing core area and number of patches, and in the case of unmanaged grass/shrubland an increase in fragmentation per se as measured by LSI.\u003c/p\u003e \u003cp\u003eBy contrast, pasture/rangeland and forest LULC exhibited a decrease in area on average in most landscape metrics in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes across the study period. For instance, at global scale all metrics declined for pasture/rangeland from 1992 to 2020, except for TCA which showed no net change between the two years. For forest cover, there was a decrease in five out of six landscape metrics from 1992 to 2020 (CA, LSI, NP, TCA, and TE), and a very small expansion in NDCA (0.52\u0026thinsp;\u0026plusmn;\u0026thinsp;0.77 in 1992 and 0.53\u0026thinsp;\u0026plusmn;\u0026thinsp;0.77 in 2020). Thus, the pasture/rangeland and forest classes declined in area and became less fragmented on average in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes over the course of the study period. The magnitude of changes in landscape patterns were larger for pasture/rangeland than forest.\u003c/p\u003e \u003cp\u003eThe direction of change in landscape metrics in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes from 1992 to 2020 was more variable for the cropland and sparse/no vegetation LULC classes, and the changes in each metric were smaller compared to the fluctuations in other LULC classes. Three out of six landscape metrics demonstrated an increase on average in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes for cropland from 1992 to 2020, with LSI, NP and TCA showing small reductions (LSI\u0026thinsp;=\u0026thinsp;1.67\u0026thinsp;\u0026plusmn;\u0026thinsp;0.58 in 1992 and LSI\u0026thinsp;=\u0026thinsp;1.63\u0026thinsp;\u0026plusmn;\u0026thinsp;0.53 in 2020; NP\u0026thinsp;=\u0026thinsp;0.52\u0026thinsp;\u0026plusmn;\u0026thinsp;1.12 in 1992 and NP\u0026thinsp;=\u0026thinsp;0.51\u0026thinsp;\u0026plusmn;\u0026thinsp;1.06 in 2020; TCA\u0026thinsp;=\u0026thinsp;454\u0026thinsp;\u0026plusmn;\u0026thinsp;1395 ha in 1992 and 452\u0026thinsp;\u0026plusmn;\u0026thinsp;1364 ha in 2020). Comparatively, there were small decreases in CA, NDCA, and TCA for sparse/no vegetation, increases in NP and TE, and no change in LSI, suggesting that sparse/no vegetation cover decreased and became more fragmented on average in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes from 1992 to 2020.\u003c/p\u003e \u003cp\u003eThe behaviour of landscape metrics over time between 1992 and 2020 was generally similar in 100 km\u003csup\u003e2\u003c/sup\u003e and 25600 km\u003csup\u003e2\u003c/sup\u003e landscapes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), which was expected as the landscape metrics were selected to show predictable scaling behaviour across landscape extents. However, there were differences in the comparative magnitude of some metrics within 100 km\u003csup\u003e2\u003c/sup\u003e versus 25600 km\u003csup\u003e2\u003c/sup\u003e landscapes. For instance, unmanaged grass/shrubland NP was 10.2% more than the pasture/rangeland NP in 100 km\u003csup\u003e2\u003c/sup\u003e landscapes in 1992, and 38.6% higher in landscapes of 25600 km\u003csup\u003e2\u003c/sup\u003e extent. Cropland LSI was 6.5% higher than forest LSI in 100 km\u003csup\u003e2\u003c/sup\u003e extent landscapes in 1992, whereas in 25600 km\u003csup\u003e2\u003c/sup\u003e landscapes cropland LSI was 2.6% less than forest LSI. These differences when comparing landscape metrics between LULC classes across scales could be due to variability in the shape of the relationship between landscape metrics and landscape extent for different LULC classes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThere was considerable variation in the direction and magnitude of landscape pattern changes across continents (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). For example, while pasture/rangeland CA declined globally the largest average net decrease was for Oceania (mean and standard deviation of -1047\u0026thinsp;\u0026plusmn;\u0026thinsp;3074 ha) whereas there was net expansion on average in Africa (117\u0026thinsp;\u0026plusmn;\u0026thinsp;1112 ha) and Asia (22\u0026thinsp;\u0026plusmn;\u0026thinsp;1046 ha). LSI for pasture/rangeland declined on average across all continents, with the largest decrease in Europe (-0.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.46) and smallest in South America on average (-0.01\u0026thinsp;\u0026plusmn;\u0026thinsp;0.37). Similarly, unmanaged grass/shrubland CA expanded on average in all continents while NP and TE increased for all continents except Africa, although the decreases in these two metrics across Africa were small (-0.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.96, and \u0026minus;\u0026thinsp;758\u0026thinsp;\u0026plusmn;\u0026thinsp;9156 m for NP and TE, respectively). Net changes in the patterns of forest and cropland were particularly variable among continents; for instance, forest CA increased in Oceania, Europe and Asia but declined in South America, Africa and North America. In general, the trends in average landscape metrics were consistent through time between 1992 and 2020 in landscapes of 100 km\u003csup\u003e2\u003c/sup\u003e extent (Fig. S11-16), although there were exceptions to this pattern. For example, average NP of pasture/rangeland increased in North America from 1998 to 2005 but showed a consistent decline from 2005 onwards. Overall, there was significant variability in landscape pattern change from 1992 to 2020 between continents, and the direction of change at continental-scale was not always the same as at global-scale.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eDirectional change in LULC area and fragmentation\u003c/h3\u003e\n\u003cp\u003eIn addition to examining the global trends in landscape patterns, we investigated how often increased area of a LULC class was associated with increased fragmentation across scales. To assess changes in area and fragmentation, landscapes of 100 km\u003csup\u003e2\u003c/sup\u003e extent which contained a LULC class in both 1992 and 2020 were classified as to whether they showed an increase (+), decrease (-), or no change (=) in CA and LSI between 1992 and 2020. The most common pattern for all LULC classes was no net change in CA or LSI between 1992 and 2020, so the majority of landscapes were categorised as CA\u0026thinsp;=\u0026thinsp;LSI= (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). However, the second-most common category of CA and LSI change varied between LULC classes. For instance, the second-most prevalent category for the urban and unmanaged grass/shrubland LULC classes was CA\u0026thinsp;+\u0026thinsp;LSI+ (19.97% and 23.22% for urban and unmanaged grass/shrubland, respectively), whereas the second-most frequent category was CA-LSI- (16.91%) for pasture/rangeland and CA-LSI\u0026thinsp;+\u0026thinsp;for forest (17.70%). The predominant patterns of LULC change were therefore divergent across LULC classes over the past three decades, although in multiple cases there was little difference in the frequency of each change category at global-scale; for example, for cropland the frequencies of the categories where both CA and LSI changed were 20.28% for CA\u0026thinsp;+\u0026thinsp;LSI-, 19.15% for CA-LSI+, 16.27% for CA-LSI-, and 15.20% for CA\u0026thinsp;+\u0026thinsp;LSI+.\u003c/p\u003e \u003cp\u003eThere was more variation in the prevalence of each category at continental-scale, however (Fig. S 23). For instance, more than 50% of landscapes exhibited a decrease in cropland CA in Europe across the study period (CA-LSI\u0026thinsp;+\u0026thinsp;=\u0026thinsp;26.36%, CA-LSI- = 26.29%, and CA-LSI\u0026thinsp;=\u0026thinsp;=\u0026thinsp;2.69%), whereas in Africa the majority of landscapes containing cropland showed an increase in CA from 1992 to 2020 (CA\u0026thinsp;+\u0026thinsp;LSI\u0026thinsp;+\u0026thinsp;=\u0026thinsp;22.34%, CA\u0026thinsp;+\u0026thinsp;LSI- = 30.60% and CA\u0026thinsp;+\u0026thinsp;LSI\u0026thinsp;=\u0026thinsp;=\u0026thinsp;3.56%). Hence, cropland was more likely to increase in landscapes in Africa and decrease in landscapes in Europe, while changes in fragmentation were split relatively evenly between increasing and declining fragmentation. There were also differences in the most common categories of CA and LSI change across landscape extents, with the CA\u0026thinsp;=\u0026thinsp;LSI\u0026thinsp;=\u0026thinsp;category becoming less frequent as landscape extent was increased at global scale (Fig. S 24).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFor all six LULC classes there was considerable spatial variation in the directional net change in CA and LSI within 100 km\u003csup\u003e2\u003c/sup\u003e landscapes which contained that LULC class in both 1992 and 2020 (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). For example, cropland decreased and became more fragmented across large areas of the eastern United States of America (US), with the dominant category being CA-LSI+. Forest loss in landscapes in the southern Amazon was also predominantly associated with more fragmentation as most landscapes where forest LULC changed were in the CA-LSI\u0026thinsp;+\u0026thinsp;category. Loss of pasture area in Europe was more commonly classified as CA-LSI- rather than the CA-LSI+, indicating that remaining pasture/rangeland area was less fragmented in 2020 compared to 1992. Given that CA, NP, and TE of pasture/rangeland declined and TCA increased slightly on average across Europe during the study period, the apparent decrease in pasture/rangeland fragmentation may be due to the loss of small farms, which would leave large-scale farms that have larger patch size and lower fragmentation. In general, increasing area of LULC classes was distributed between the CA\u0026thinsp;+\u0026thinsp;LSI\u0026thinsp;+\u0026thinsp;and CA\u0026thinsp;+\u0026thinsp;LSI- categories, with increasing forest cover in northern Russia classed as CA\u0026thinsp;+\u0026thinsp;LSI- (increased area and decreased fragmentation) and unmanaged grass/shrubland in eastern Australia mostly classified as CA\u0026thinsp;+\u0026thinsp;LSI+, for instance. There were extensive changes in area and fragmentation of pasture/rangeland and unmanaged grass/shrubland across Australia, with a trend towards increased unmanaged grass/shrubland and decreased pasture cover.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOverall, our findings show considerable variation in change across LULC classes between 1992 and 2020. At a global scale, the unmanaged grass/shrubland LULC class expanded and became more fragmented on average across the study period, whereas forest and pasture/rangeland declined and became less fragmented (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Global urban LULC increased in area on average but showed a decline in fragmentation per se. Urban LULC is known to have expanded globally since 1970 (van Vliet \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; G\u0026uuml;neralp et al. \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Liu et al. \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), hence our findings agree with previous studies. Urban LULC change can drive both increased and decreased fragmentation of landscapes (Irwin and Bockstael \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Schneider and Woodcock \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), likely due to factors such as the price of land and geographic constraints on expansion (Angel et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The continued expansion and decreased fragmentation of urban LULC at global scale has implications for food security and the environment depending on whether urban areas expand into cropland or natural LULC classes.\u003c/p\u003e \u003cp\u003eChanges in the landscape patterns of forest LULC were spatially variable, with forests increasing in area and becoming less fragmented in Europe, Asia, and Oceania but declining and fragmenting in Africa, South America and North America. Hence, forest areas were particularly prone to loss and fragmentation in tropical regions, which agrees with previous research (Hansen et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Ma et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) and has implications for biodiversity loss (Alroy \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Giam \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), carbon storage (Chaplin-Kramer et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Shapiro et al. \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) and carbon emissions (Brinck et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Broggio et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eMa et al. (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) found that approximately 75% of global forest landscapes with exactly 25 km\u003csup\u003e2\u003c/sup\u003e extent became less fragmented between 2000 and 2020, whereas only 26.8% of forest landscapes with 100 km\u003csup\u003e2\u003c/sup\u003e extent showed decreased fragmentation between 1992 and 2020 here. There were multiple differences in methodology between the two studies that may explain this discrepancy. For example, net change in forest LSI was calculated from 1992 to 2020 in this study, whereas net change in forest fragmentation was assessed between 2000 and 2020 in (Ma et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Increases in forest fragmentation from 1992 to 2000 may have counteracted decreases between 2000 and 2020, leading to a lower proportion of landscapes identified as having decreased fragmentation in this study. Similarly, input LULC maps of 1 km resolution (this study) versus 30 m resolution (Ma et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) were used to calculate forest landscape metrics. Coarse resolution maps are more homogeneous than fine resolution ones (Wiens \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e1989\u003c/span\u003e), so Ma et al. (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) may have detected fine-scale changes in forest patterns that were apparent at 30 m but not 1 km resolution. Furthermore, different metrics were used to quantify fragmentation (LSI versus a normalised index of edge density, patch density, and mean patch area; Ma et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), which could have influenced the classification of landscapes into increased versus decreased fragmentation in the two studies. Indeed, Wang et al. (\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) show these indices vary in complex ways in how well they capture fragmentation independently of habitat amount. A similar definition of forest was used in both studies, with forest was defined as trees with height of \u0026ge;\u0026thinsp;5 metres in Ma et al. (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) and as trees with height of \u0026gt;\u0026thinsp;5 metres and cover\u0026thinsp;\u0026ge;\u0026thinsp;10% in the HILDA\u0026thinsp;+\u0026thinsp;dataset (Winkler et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2024\u003c/span\u003e, \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), but even these small differences may have contributed to the differing conclusions between the studies.\u003c/p\u003e \u003cp\u003eThe pasture/rangeland and unmanaged grass/shrubland LULC classes showed the largest changes in landscape pattern at global scale from 1992 to 2020. Pasture/rangeland is acknowledged to have declined globally since about the year 2000 (Blaustein-Rejto et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Winkler et al. \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), but the outcomes of pasture/rangeland declines will depend on what they are replaced by. For example, replacement of pasture/rangeland with unmanaged grass/shrublands may have positive implications for the environment such as through providing additional land for biodiversity (Poore \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) and carbon sequestration (Silver et al. \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), whereas replacing pasture/rangeland with urban LULC would likely not have environmental benefits. However, the replacement of pasture/rangeland with unmanaged grass/shrubland does not necessarily lead to positive outcomes. For instance, plant diversity does not always recover after land abandonment (Cava et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Isbell et al. \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) and land abandonment in cultural landscapes with low intensity land use can lead to biodiversity loss (Daskalova and Kamp \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Similarly, the occurrence of new unmanaged grass/shrubland patches at a distance from existing ones may increase the time taken for species to colonise the new patches, leading to much longer time lags in biodiversity recovery compared to creating new patches near to current ones (Synes et al. \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Overall, the global decrease in pasture/rangelands and increase in unmanaged grass/shrublands suggests that there are emerging opportunities for ecosystem recovery and restoration, although management actions may be needed to ensure that environmental outcomes are positive.\u003c/p\u003e \u003cp\u003eAlthough unmanaged grass/shrubland appears to have become more fragmented at a global scale, difficulties in classifying LULC as pasture or rangeland versus unmanaged land make this uncertain (Phelps and Kaplan \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Moreover, the changes in the patterns of pasture/rangeland and unmanaged grass/shrubland LULC in our study are largely driven by shifts in LULC across Oceania, and particularly Australia (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). Australia has extensive areas of both managed and unmanaged grasslands, which drives high uncertainty in these regions in HILDA\u0026thinsp;+\u0026thinsp;because the underlying input datasets use different classifications for pasture and grasslands and it is difficult to delineate unmanaged versus low intensity grasslands within heterogeneous rangeland landscapes (Winkler et al. \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Therefore, it is unclear whether the spatial patterns of pasture/rangeland and unmanaged grass/shrubland across Australia, and the rest of the globe, are real or an artifact of the HILDA\u0026thinsp;+\u0026thinsp;data. Further research is needed to: a) advance our capacity to identify land use intensity, and therefore distinguish between managed pastures and unmanaged grass/shrublands, in LULC datasets, and b) develop a unified approach for defining pastures and grasslands given that LULC datasets currently use different definitions to represent a range of animal production systems with differing intensities (Phelps and Kaplan \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Improved mapping and consolidated definitions of pasture/rangelands and unmanaged grass/shrublands would help to establish whether unmanaged grass/shrublands are actually becoming more fragmented at global scale, or whether the spatial pattern of unmanaged grass/shrubland change is less fragmented than suggested in HILDA+.\u003c/p\u003e \u003cp\u003eOverall, changes in landscape patterns were highly spatially variable for all LULC classes, with differences apparent at local- to continental-scales. The spatial pattern of CA and LSI categories did not appear to be random (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e), as clusters of neighbouring landscapes often belonged to the same category. Therefore, the drivers of landscape pattern change likely act on local- to regional-scales, which has implications for representing landscape patterns in land use models. Moreover, when modelling LULC change, we cannot always assume that natural LULC classes become more and anthropogenic LULC classes less fragmented over time, or vice versa; instead, landscape pattern changes are spatially and temporally heterogeneous, as has been suggested by previous studies (Turner and Gardner \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). We only assessed the temporal change in landscape patterns at global- and continental-scales, so future work could expand this temporal analysis by identifying regions or landscapes which showed varied temporal dynamics from 1992 to 2020, such as those where the temporal trend of fragmentation changed direction. Additional temporal analyses could support further research into the drivers of landscape pattern change across scales and how these relate to LULC change drivers, in order to better predict landscape pattern change and its impacts on the Earth System. Future work could also examine the patterns of LULC transitions in addition to changes in the patterns of individual LULC classes, to understand whether fragmenting LULC classes are being converted to natural or anthropogenic LULC classes and how this in turn impacts the environment. Disentangling the effects of habitat amount and configuration using more landscape metrics at a single landscape extent could be another focus of future research, as unfortunately none of the landscape indices showing predictable behaviour across landscape extents are also entirely uncorrelated with habitat amount, though LSI is known to be a reasonable measure of fragmentation per se between 20 and 60% class area (Wang et al. \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn conclusion, global-scale landscape pattern change demonstrated considerable heterogeneity between 1992 and 2020, which has implications for the impacts of change on environmental processes such as biodiversity and carbon emissions. Although trends in landscape metrics were generally consistent across landscape extents, the differing relationships between landscape metrics and landscape extent across LULC classes highlights the importance of considering multiple spatial scales and LULC classes when assessing landscape pattern change. The spatial heterogeneity in landscape patterns detected here suggests that landscape pattern change should be accounted for when quantifying and predicting LULC change, especially as landscape patterns are key for a range of environmental processes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTLW was funded by an EASTBIO Doctoral Training Partnership Biotechnology and Biological Sciences Research Council grant number BB/T00875X/1 and a ERC Starting Grant \u0026apos;SCALEFORES\u0026apos; (grant no. 680176), awarded to FE. TLW and PA were supported by the UKRI projects (ForestPaths, 10039590) and (MOSAIC, 10075849).\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eTLW was funded by an EASTBIO Doctoral Training Partnership Biotechnology and Biological Sciences Research Council grant number BB/T00875X/1 and a ERC Starting Grant \u0026apos;SCALEFORES\u0026apos; (grant no. 680176), awarded to FE. TLW and PA were supported by the UKRI projects (ForestPaths, 10039590) and (MOSAIC, 10075849).\u003c/p\u003e\n\u003ch2\u003eCompeting interests\u003c/h2\u003e\n\u003cp\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003ch2\u003eAuthor contributions\u003c/h2\u003e\n\u003cp\u003eThe study was conceptualised by TLW, FE, PA, and JMJT. TLW carried out all analysis with data generated and supplied by KW. The first draft of the manuscript was written by TLW and all authors contributed to revising and editing the manuscript.\u003c/p\u003e\n\u003ch2\u003eData availability\u003c/h2\u003e\n\u003cp\u003eThe HILDA+ version 2b dataset will be made publicly available on Zenodo at a later date (https://doi.org/10.5281/zenodo.15017066). The global dataset of landscape patterns generated in this study is publicly available on Zenodo (https://doi.org/10.5281/zenodo.15120267), as is the code used to generate and analyse the dataset (https://doi.org/10.5281/zenodo.15124527).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAlexander P, Prestele R, Verburg PH, et al (2017) Assessing uncertainties in land cover projections. Glob Change Biol 23:767\u0026ndash;781. https://doi.org/10.1111/gcb.13447\u003c/li\u003e\n\u003cli\u003eAlroy J (2017) Effects of habitat disturbance on tropical forest biodiversity. Proc Natl Acad Sci USA 114:6056\u0026ndash;6061. https://doi.org/10.1073/pnas.1611855114\u003c/li\u003e\n\u003cli\u003eAngel S, Parent J, Civco DL (2012) The fragmentation of urban landscapes: global evidence of a key attribute of the spatial structure of cities, 1990\u0026ndash;2000. Environ Urban 24:249\u0026ndash;283. https://doi.org/10.1177/0956247811433536\u003c/li\u003e\n\u003cli\u003eArga\u0026ntilde;araz JP, Entraigas I (2014) Scaling functions evaluation for estimation of landscape metrics at higher resolutions. Ecol Inform 22:1\u0026ndash;12. https://doi.org/10.1016/j.ecoinf.2014.02.004\u003c/li\u003e\n\u003cli\u003eArmenteras D, Rodr\u0026iacute;guez N, Retana J, Morales M (2011) Understanding deforestation in montane and lowland forests of the Colombian Andes. Reg Environ Change 11:693\u0026ndash;705. https://doi.org/10.1007/s10113-010-0200-y\u003c/li\u003e\n\u003cli\u003eBlaustein-Rejto D, Blomqvist L, McNamara J, de Kirby K (2019) Achieving Peak Pasture: shrinking Pasture\u0026rsquo;s footprint by spreading the livestock revolution. The Breakthrough Institute\u003c/li\u003e\n\u003cli\u003eBrinck K, Fischer R, Groeneveld J, et al (2017) High resolution analysis of tropical forest fragmentation and its impact on the global carbon cycle. Nat Commun 8:14855. https://doi.org/10.1038/ncomms14855\u003c/li\u003e\n\u003cli\u003eBroggio IS, Silva-Junior CHL, Nascimento MT, et al (2024) Quantifying landscape fragmentation and forest carbon dynamics over 35 years in the Brazilian Atlantic Forest. Environ Res Lett 19:034047. https://doi.org/10.1088/1748-9326/ad281c\u003c/li\u003e\n\u003cli\u003eCattarino L, McAlpine CA, Rhodes JR (2014) Land‐use drivers of forest fragmentation vary with spatial scale. Glob Ecol Biogeogr 23:1215\u0026ndash;1224. https://doi.org/10.1111/geb.12187\u003c/li\u003e\n\u003cli\u003eCava MGB, Pilon NAL, Ribeiro MC, Durigan G (2018) Abandoned pastures cannot spontaneously recover the attributes of old‐growth savannas. J Appl Ecol 55:1164\u0026ndash;1172. https://doi.org/10.1111/1365-2664.13046\u003c/li\u003e\n\u003cli\u003eCenter for International Earth Science Information Network - CIESIN - Columbia University (2018) Gridded Population of the World, Version 4 (GPWv4): Population Density, Revision 11. Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC).\u003c/li\u003e\n\u003cli\u003eChaplin-Kramer R, Ramler I, Sharp R, et al (2015) Degradation in carbon stocks near tropical forest edges. Nat Commun 6:10158. https://doi.org/10.1038/ncomms10158\u003c/li\u003e\n\u003cli\u003eDaskalova GN, Kamp J (2023) Abandoning land transforms biodiversity. Science 380:581\u0026ndash;583. https://doi.org/10.1126/science.adf1099\u003c/li\u003e\n\u003cli\u003eESA (2017) Land Cover CCI Product User Guide Version 2. Tech. Rep.\u003c/li\u003e\n\u003cli\u003eEtter A, McAlpine C, Possingham H (2008) Historical patterns and drivers of landscape change in Colombia since 1500: A regionalized spatial approach. Ann Am Assoc Geogr 98:2\u0026ndash;23. https://doi.org/10.1080/00045600701733911\u003c/li\u003e\n\u003cli\u003eEwers RM, Laurance WF (2006) Scale-dependent patterns of deforestation in the Brazilian Amazon. Environ Conserv 33:203\u0026ndash;211. https://doi.org/10.1017/S0376892906003250\u003c/li\u003e\n\u003cli\u003eFahrig L (2003) Effects of Habitat Fragmentation on Biodiversity. Annu Rev Ecol Evol Syst 34:487\u0026ndash;515. https://doi.org/10.1146/annurev.ecolsys.34.011802.132419\u003c/li\u003e\n\u003cli\u003eFahrig L, Arroyo-Rodr\u0026iacute;guez V, Bennett JR, et al (2019) Is habitat fragmentation bad for biodiversity? Biological Conservation 230:179\u0026ndash;186. https://doi.org/10.1016/j.biocon.2018.12.026\u003c/li\u003e\n\u003cli\u003eFischer G, Nachtergaele FO, van Velthuizen HT, et al (2021) Global Agro-Ecological Zones (GAEZ v4) Model Documentation. FAO \u0026amp; IIASA\u003c/li\u003e\n\u003cli\u003eFischer J, Lindenmayer DB (2007) Landscape modification and habitat fragmentation: a synthesis. Glob Ecol Biogeogr 16:265\u0026ndash;280. https://doi.org/10.1111/j.1466-8238.2007.00287.x\u003c/li\u003e\n\u003cli\u003eFletcher Jr RJ, Didham RK, Banks-Leite C, et al (2018) Is habitat fragmentation good for biodiversity? Biol Conserv 226:9\u0026ndash;15. https://doi.org/10.1016/j.biocon.2018.07.022\u003c/li\u003e\n\u003cli\u003eGiam X (2017) Global biodiversity loss from tropical deforestation. Proc Natl Acad Sci USA 114:5775\u0026ndash;5777. https://doi.org/10.1073/pnas.1706264114\u003c/li\u003e\n\u003cli\u003eGonz\u0026aacute;lez-Gonz\u0026aacute;lez A, Clerici N, Quesada B (2021) Growing mining contribution to Colombian deforestation. Environ Res Lett 16:064046. https://doi.org/10.1088/1748-9326/abfcf8\u003c/li\u003e\n\u003cli\u003eG\u0026uuml;neralp B, Reba M, Hales BU, et al (2020) Trends in urban land expansion, density, and land transitions from 1970 to 2010: a global synthesis. Environ Res Lett 15:044015. https://doi.org/10.1088/1748-9326/ab6669\u003c/li\u003e\n\u003cli\u003eGustafson EJ (1998) Quantifying Landscape Spatial Pattern: What Is the State of the Art? Ecosyst 1:143\u0026ndash;156. https://doi.org/10.1007/s100219900011\u003c/li\u003e\n\u003cli\u003eGustafson EJ (2019) How has the state-of-the-art for quantification of landscape pattern advanced in the twenty-first century? Landsc Ecol 34:2065\u0026ndash;2072. https://doi.org/10.1007/s10980-018-0709-x\u003c/li\u003e\n\u003cli\u003eHaddad NM, Brudvig LA, Clobert J, et al (2015) Habitat fragmentation and its lasting impact on Earth\u0026rsquo;s ecosystems. Sci Adv 1:e1500052. https://doi.org/10.1126/sciadv.1500052\u003c/li\u003e\n\u003cli\u003eHansen MC, Potapov PV, Moore R, et al (2013) High-Resolution Global Maps of 21st-Century Forest Cover Change. Science 342:850\u0026ndash;853. https://doi.org/10.1126/science.1244693\u003c/li\u003e\n\u003cli\u003eHesselbarth MHK, Sciaini M, With KA, et al (2019) landscapemetrics: an open-source R tool to calculate landscape metrics. Ecography 42:1648\u0026ndash;1657. https://doi.org/10.1111/ecog.04617\u003c/li\u003e\n\u003cli\u003eHijmans RJ (2022) terra: Spatial Data Analysis. R package version 1.7-23. https://rspatial.org/terra/\u003c/li\u003e\n\u003cli\u003eHu Q, Xiang M, Chen D, et al (2020) Global cropland intensification surpassed expansion between 2000 and 2010: A spatio-temporal analysis based on GlobeLand30. Sci Total Environ 746:141035. https://doi.org/10.1016/j.scitotenv.2020.141035\u003c/li\u003e\n\u003cli\u003eIrwin EG, Bockstael NE (2007) The evolution of urban sprawl: Evidence of spatial heterogeneity and increasing land fragmentation. Proc Natl Acad Sci USA 104:20672\u0026ndash;20677. https://doi.org/10.1073/pnas.0705527105\u003c/li\u003e\n\u003cli\u003eIsbell F, Tilman D, Reich PB, Clark AT (2019) Deficits of biodiversity and productivity linger a century after agricultural abandonment. Nat Ecol Evol 3:1533\u0026ndash;1538. https://doi.org/10.1038/s41559-019-1012-1\u003c/li\u003e\n\u003cli\u003eJacobson AP, Riggio J, Tait AM, Baillie JEM (2019) Global areas of low human impact (\u0026lsquo;Low Impact Areas\u0026rsquo;) and fragmentation of the natural world. Sci Rep 9:14179. https://doi.org/10.1038/s41598-019-50558-6\u003c/li\u003e\n\u003cli\u003eJaureguiberry P, Titeux N, Wiemers M, et al (2022) The direct drivers of recent global anthropogenic biodiversity loss. Sci Adv 2022:eabm9982. https://doi.org/10.1126/sciadv.abm9982\u003c/li\u003e\n\u003cli\u003eJin H, Xu J, Peng Y, et al (2023) Impacts of landscape patterns on plant species diversity at a global scale. Sci Total Environ 896:165193. https://doi.org/10.1016/j.scitotenv.2023.165193\u003c/li\u003e\n\u003cli\u003eKummu M, Taka M, Guillaume JHA (2018) Gridded global datasets for Gross Domestic Product and Human Development Index over 1990\u0026ndash;2015. Sci Data 5:180004. https://doi.org/10.1038/sdata.2018.4\u003c/li\u003e\n\u003cli\u003eLiu X, Huang Y, Xu X, et al (2020) High-spatiotemporal-resolution mapping of global urban change from 1985 to 2015. Nat Sustain 3:564\u0026ndash;570. https://doi.org/10.1038/s41893-020-0521-x\u003c/li\u003e\n\u003cli\u003eLiu X, Li S, Wang S, et al (2022) Effects of farmland landscape pattern on spatial distribution of soil organic carbon in Lower Liaohe Plain of northeastern China. Ecol Indic 145:109652. https://doi.org/10.1016/j.ecolind.2022.109652\u003c/li\u003e\n\u003cli\u003eLockhart J, Koper N (2018) Northern prairie songbirds are more strongly influenced by grassland configuration than grassland amount. Landscape Ecol 33:1543\u0026ndash;1558. https://doi.org/10.1007/s10980-018-0681-5\u003c/li\u003e\n\u003cli\u003eMa J, Li J, Wu W, Liu J (2023) Global forest fragmentation change from 2000 to 2020. Nat Commun 14:3752. https://doi.org/10.1038/s41467-023-39221-x\u003c/li\u003e\n\u003cli\u003eMartello F, dos Santos JS, Silva-Neto CM, et al (2023) Landscape structure shapes the diversity of plant reproductive traits in agricultural landscapes in the Brazilian Cerrado. Agric Ecosyst Environ 341:108216. https://doi.org/10.1016/j.agee.2022.108216\u003c/li\u003e\n\u003cli\u003eMassicotte P, South A (2023) rnaturalearth: World Map Data from Natural Earth. R package version 0.3.2. https://CRAN.R-project.org/package=rnaturalearth\u003c/li\u003e\n\u003cli\u003eMaxwell SL, Fuller RA, Brooks TM, Watson JEM (2016) Biodiversity: The ravages of guns, nets and bulldozers. Nature 536:143\u0026ndash;145. https://doi.org/10.1038/536143a\u003c/li\u003e\n\u003cli\u003eMiguet P, Jackson HB, Jackson ND, et al (2016) What determines the spatial extent of landscape effects on species? Landsc Ecol 31:1177\u0026ndash;1194. https://doi.org/10.1007/s10980-015-0314-1\u003c/li\u003e\n\u003cli\u003ePais C, Miranda A, Carrasco J, Shen Z-JM (2021) Deep fire topology: Understanding the role of landscape spatial patterns in wildfire occurrence using artificial intelligence. Environ Model Softw 143:105122. https://doi.org/10.1016/j.envsoft.2021.105122\u003c/li\u003e\n\u003cli\u003ePhelps LN, Kaplan JO (2017) Land use for animal production in global change studies: Defining and characterizing a framework. Glob Change Biol 23:4457\u0026ndash;4471. https://doi.org/10.1111/gcb.13732\u003c/li\u003e\n\u003cli\u003ePoore JAC (2016) Call for conservation: Abandoned pasture. Science 351:132. https://doi.org/10.1126/science.351.6269.132-a\u003c/li\u003e\n\u003cli\u003ePotapov P, Turubanova S, Hansen MC, et al (2021) Global maps of cropland extent and change show accelerated cropland expansion in the twenty-first century. Nat Food 3:19\u0026ndash;28. https://doi.org/10.1038/s43016-021-00429-z\u003c/li\u003e\n\u003cli\u003eR Core Team (2022) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.\u003c/li\u003e\n\u003cli\u003eR Core Team (2020) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.\u003c/li\u003e\n\u003cli\u003eRiitters K, Wickham J, O\u0026rsquo;Neill R, et al (2000) Global-scale patterns of forest fragmentation. Conserv Ecol 4:3\u003c/li\u003e\n\u003cli\u003eRyu S-R, Chen J, Zheng D, Lacroix JJ (2007) Relating surface fire spread to landscape structure: An application of FARSITE in a managed forest landscape. Landsc Urban Plan 83:275\u0026ndash;283. https://doi.org/10.1016/j.landurbplan.2007.05.002\u003c/li\u003e\n\u003cli\u003eSchneider A, Woodcock CE (2008) Compact, Dispersed, Fragmented, Extensive? A Comparison of Urban Growth in Twenty-five Global Cities using Remotely Sensed Data, Pattern Metrics and Census Information. Urban Stud 45:659\u0026ndash;692. https://doi.org/10.1177/0042098007087340\u003c/li\u003e\n\u003cli\u003eShapiro AC, Aguilar-Amuchastegui N, Hostert P, Bastin J-F (2016) Using fragmentation to assess degradation of forest edges in Democratic Republic of Congo. Carbon Balance Manage 11:11. https://doi.org/10.1186/s13021-016-0054-9\u003c/li\u003e\n\u003cli\u003eSilver WL, Ostertag R, Lugo AE (2000) The Potential for Carbon Sequestration Through Reforestation of Abandoned Tropical Agricultural and Pasture Lands. Restor Ecol 8:394\u0026ndash;407. https://doi.org/10.1046/j.1526-100x.2000.80054.x\u003c/li\u003e\n\u003cli\u003e\u0026Scaron;\u0026iacute;mov\u0026aacute; P, Gdulov\u0026aacute; K (2012) Landscape indices behavior: A review of scale effects. Appl Geogr 34:385\u0026ndash;394. https://doi.org/10.1016/j.apgeog.2012.01.003\u003c/li\u003e\n\u003cli\u003eSponseller RA, Benfield EF, Valett HM (2008) Relationships between land use, spatial scale and stream macroinvertebrate communities. Freshw Biol 46:1409\u0026ndash;1424. https://doi.org/10.1046/j.1365-2427.2001.00758.x\u003c/li\u003e\n\u003cli\u003eSynes NW, Ponchon A, Palmer SCF, et al (2020) Prioritising conservation actions for biodiversity: Lessening the impact from habitat fragmentation and climate change. Biol Conserv 252:108819. https://doi.org/10.1016/j.biocon.2020.108819\u003c/li\u003e\n\u003cli\u003eTurner MG (1989) Landscape Ecology: The Effect of Pattern on Process. Annu Rev Ecol Evol Syst 20:171\u0026ndash;97. https://doi.org/10.1146/annurev.es.20.110189.001131\u003c/li\u003e\n\u003cli\u003eTurner MG, Gardner RH (2015) Causes of Landscape Pattern. In: Landscape Ecology in Theory and Practice. Springer, New York, NY, pp 33\u0026ndash;62\u003c/li\u003e\n\u003cli\u003eUuemaa E, Roosaare J, Mander \u0026Uuml; (2005) Scale dependence of landscape metrics and their indicatory value for nutrient and organic matter losses from catchments. Ecol Indic 5:350\u0026ndash;369. https://doi.org/10.1016/j.ecolind.2005.03.009\u003c/li\u003e\n\u003cli\u003evan Vliet J (2019) Direct and indirect loss of natural area from urban expansion. Nat Sustain 2:755\u0026ndash;763. https://doi.org/10.1038/s41893-019-0340-0\u003c/li\u003e\n\u003cli\u003eWang X, Blanchet FG, Koper N (2014) Measuring habitat fragmentation: An evaluation of landscape pattern metrics. Methods Ecol Evol 5:634\u0026ndash;646. https://doi.org/10.1111/2041-210X.12198\u003c/li\u003e\n\u003cli\u003eWiens JA (1989) Spatial Scaling in Ecology. Funct Ecol 3:385\u0026ndash;397. https://doi.org/10.2307/2389612\u003c/li\u003e\n\u003cli\u003eWinkler K, Fuchs R, Rounsevell M, Herold M (2021) Global land use changes are four times greater than previously estimated. Nat Commun 12:1\u0026ndash;10. https://doi.org/10.1038/s41467-021-22702-2\u003c/li\u003e\n\u003cli\u003eWinkler K, Fuchs R, Rounsevell M, Herold M (2024) HILDA+ version 2.0: Global Land Use Change between 1960 and 2020 [dataset]. PANGAEA (in review)\u003c/li\u003e\n\u003cli\u003eWinkler K, Fuchs R, Rounsevell MDA, Herold M (2020) HILDA+ Global Land Use Change between 1960 and 2019 [dataset]\u003c/li\u003e\n\u003cli\u003eWoodman TL, Arendarczyk B, Winkler K, et al (2025) Harmonised global land use and land cover maps between 1960 and 2100 [dataset]\u003c/li\u003e\n\u003cli\u003eWoodman TL, Rueda-Uribe C, Henry RC, et al (2023) Introducing LandScaleR: A novel method for spatial downscaling of land use projections. Environ Model Softw 169:105826. https://doi.org/10.1016/j.envsoft.2023.105826\u003c/li\u003e\n\u003cli\u003eWu J (2004) Effects of changing scale on landscape pattern analysis: scaling relations. Landsc Ecol 19:125\u0026ndash;138. https://doi.org/10.1023/B:LAND.0000021711.40074.ae\u003c/li\u003e\n\u003cli\u003eWu J, Shen W, Sun W, Tueller PT (2002) Empirical patterns of the effects of changing scale on landscape metrics. Landsc Ecol 17:761\u0026ndash;782. https://doi.org/10.1023/A:1022995922992\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"landscape-ecology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"land","sideBox":"Learn more about [Landscape Ecology](https://www.springer.com/journal/10980)","snPcode":"10980","submissionUrl":"https://submission.nature.com/new-submission/10980/3","title":"Landscape Ecology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Landscape patterns, Landscape metrics, Land use and land cover change, Global spatial scale, Landscape fragmentation","lastPublishedDoi":"10.21203/rs.3.rs-6522006/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6522006/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eContext\u003c/b\u003e\u003c/p\u003e \u003cp\u003eLandscape patterns are driven by complex natural and anthropogenic factors and are important for a range of environmental processes, including species movement and wildfire risks. Previous assessments of global-scale landscape pattern change have focused on a single land use and land cover (LULC) type or landscape-level measurements, hence there is a lack of knowledge on landscape pattern change across multiple LULC classes.\u003c/p\u003e\u003cp\u003e\u003cb\u003eObjectives\u003c/b\u003e\u003c/p\u003e \u003cp\u003eWe assessed global-scale change in landscape patterns for the six LULC classes used in the HILDA\u0026thinsp;+\u0026thinsp;dataset (urban, cropland, pasture/rangeland, forest, unmanaged grass/shrubland, and sparse/no vegetation) from 1992 to 2020.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e\u003c/p\u003e \u003cp\u003eSix class-level landscape metrics which showed predictable scaling behaviour with landscape extent were calculated for each LULC class and year of the study period. Landscape metrics were quantified for five landscape extents (100, 400, 1600, 6400 and 25600 km\u003csup\u003e2\u003c/sup\u003e). Trends in global landscape patterns over time were evaluated with a particular focus on area and fragmentation.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e\u003c/p\u003e \u003cp\u003eUnmanaged grass/shrubland LULC expanded in area and showed increased fragmentation, while pasture/rangeland and forest LULC tended to decline in area and exhibit decreased fragmentation. Even though there was high spatial heterogeneity in landscape pattern change for all LULC classes, neighbouring 100 km\u003csup\u003e2\u003c/sup\u003e landscapes often showed the same directional change in area and fragmentation.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusions\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThese findings highlight the variability in landscape pattern change at global scales and indicate that drivers of landscape pattern vary across the globe at local to regional scales, with implications for environmental processes such as biodiversity loss and carbon storage.\u003c/p\u003e","manuscriptTitle":"Global assessment of landscape pattern changes from 1992 to 2020","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-06 16:24:15","doi":"10.21203/rs.3.rs-6522006/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-06-22T06:15:34+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-20T14:31:40+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-19T01:53:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"21935197225668975821672241894902922773","date":"2025-06-07T11:05:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"321310589713145686008222426618678022992","date":"2025-06-07T03:41:20+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-18T12:24:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"82871077468443889982743134545166709129","date":"2025-05-06T18:01:00+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"285454350310452514362550299349610977074","date":"2025-04-28T12:38:14+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-28T02:36:41+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-04-25T01:15:09+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-25T01:12:11+00:00","index":"","fulltext":""},{"type":"submitted","content":"Landscape Ecology","date":"2025-04-24T14:58:23+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"landscape-ecology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"land","sideBox":"Learn more about [Landscape Ecology](https://www.springer.com/journal/10980)","snPcode":"10980","submissionUrl":"https://submission.nature.com/new-submission/10980/3","title":"Landscape Ecology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"0d79fa38-07f8-4b2f-8a86-2d38c2676cae","owner":[],"postedDate":"May 6th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-10-20T16:00:31+00:00","versionOfRecord":{"articleIdentity":"rs-6522006","link":"https://doi.org/10.1007/s10980-025-02210-0","journal":{"identity":"landscape-ecology","isVorOnly":false,"title":"Landscape Ecology"},"publishedOn":"2025-10-16 15:56:59","publishedOnDateReadable":"October 16th, 2025"},"versionCreatedAt":"2025-05-06 16:24:15","video":"","vorDoi":"10.1007/s10980-025-02210-0","vorDoiUrl":"https://doi.org/10.1007/s10980-025-02210-0","workflowStages":[]},"version":"v1","identity":"rs-6522006","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6522006","identity":"rs-6522006","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}