Availability and proximity of natural habitat influence cropland biodiversity in forest biomes globally

Abstract

Aim: Agriculture is one of the greatest pressures on biodiversity. Regional studies have shown that the presence of natural habitat and landscape heterogeneity are beneficial for biodiversity in agriculture, but it remains unclear whether their importance varies geographically. Here, we use local biodiversity data to determine which local and landscape variables are most associated with biodiversity patterns and whether their association varies between tropical and non-tropical regions.Location: Global terrestrial area in forest biomes.Major taxa studied: More than 21,000 species of vertebrates, invertebrates, plants and other taxa. Methods:We used generalized linear mixed-effects models to analyse the relationships between either community total abundance or species richness (derived from the PREDICTS database) and a number of site-level (predominant land use and landuse intensity) and landscape-level variables (distance to forest, the percentage of natural habitat in the surrounding landscape, landscape homogeneity, the number of land-cover types in the landscape, and total fertilizer application). We compared the associations of these variables with biodiversity in tropical and non-tropical regions. Results:In most cases, changes in biodiversity associated with landscape-level variables were greater than those associated with local land use and land-use intensity.Increased natural habitat availability was associated with the most consistent increases in biodiversity. Landscape homogeneity was also important but showed different directions of biodiversity change between regions. Associations with fertilizer application or the number of land-cover types were generally weaker, although still of greater magnitude than for the local land-use measures.Main conclusions: Our results highlight similarities and differences in the association of local-and landscape-scale variables with local biodiversity in tropical and non-tropical regions. Landscape natural habitat availability had a consistent positive association with biodiversity, highlighting the key role of landscape management in the maintenance of biodiversity in croplands. Landscape-scale variables were almost always associated with greater changes in biodiversity than the local-scale measures.

Biome type and availability of natural vegetation dictate whether farmland intensification or expansion is worse for biodiversity

“…Previous research has shown the importance of natural habitat in buffering the effects of human activity on biodiversity (Macchi et al 2020;Outhwaite et al 2022) but our results suggest nuanced effects depending on crop. In landscapes with wheat cultivation, areas with higher percentage of natural habitat experience steeper decreases in species richness and abundance with yield, likely from higher initial biodiversity as natural habitat is positively related to species richness and abundance in wheat landscapes (Table S20).…”

“…However, primary vegetation in modi ed landscapes likely represents an important source of biodiversity and ecosystem services for the surrounding farmland (Karp et al 2013;Outhwaite et al 2022).…”

“…We included percentage of natural habitat in our models because of previous studies indicating the importance of this landscape characteristic on its own (Outhwaite et al 2022) and in interaction with yield (Macchi et al 2020) for determining biodiversity patterns. We extracted for each data point the percentage of natural habitat at that location by using a modi ed version of the land-use estimates from Hoskins et al (2016).…”

“…For the yield-biodiversity models, we used percentage of natural habitat as a continuous variable and interaction term with yield, land use, biome type and subsistence yield, to account for potentially different biodiversity responses to these variables at different levels of natural habitat (Macchi et al 2020;Outhwaite et al 2022). For the land-conversion models, we used percentage of natural habitat to select sites that were located in pixels with less than 30% natural vegetation ('human-modi ed landscapes'), and pixels with more than 70% natural vegetation ('natural landscapes') (Table S4).…”

“…To capture potential differences in the impact of yield increases on cropland and primary vegetation within the same landscape, we included the PREDICTS land-use classi cation as a variable in the models. For the yield-biodiversity models, we used the original land-use classi cation from the PREDICTS database and interaction terms between land use, and yield and percentage of natural vegetation as we expect that these effects will differ depending on local land use (Outhwaite et al 2022). For the land-conversion models, we used the land-use classi cation in combination with percentage of natural habitat, as detailed above.…”