Contrasting ecological mechanisms mediate the impact of land conversion on ecosystem multifunctionality

Land use/cover (LULC) changes have unequivocally affected biodiversity and ecosystem functioning, with enormous repercussions for human well-being. However, the mechanistic ecological mechanisms underlying the impact of land conversion on ecosystem multifunctionality (EMF) remain insufficiently examined from the perspective of multiple biodiversity attributes in dryland regions with increasing deforestation rates. We investigated how the conversion of natural forests and savannas to agroforestry parklands alters the relationships between multiple biodiversity attributes (taxonomic, functional, phylogenetic, and structural) and EMF, while accounting for the effects of environmental factors in the dryland landscapes in Benin. We used forest inventory data from 145 plots spanning forests, savannas, and agroforestry parklands and assessed the implications of three land conversion scenarios. We quantified EMF using eight functions that are central to primary productivity and nutrient cycling. We found that EMF was positively related solely to structural diversity in forests. The conversion of forests and savannas to agroforestry parklands decreased EMF both directly and indirectly. The indirect effects were mediated by two contrasting biodiversity effects. When forests were converted to agroforestry parklands through savannas, indirect effects were driven by shifts in functional composition toward the dominance of species with acquisitive traits. In contrast, species diversity reduction explained the indirect effects when savannas were converted to agroforestry parklands. The aridity index and soil texture influenced biodiversity attributes, but not EMF. The present study provides evidence that the biodiversity–EMF relationship is dependent on the LULC type and was evident only in the natural ecosystem through the effects of structural diversity, thereby emphasizing the importance of enhancing structural diversity for promoting EMF in forests. Our findings also demonstrate that land conversion weakened natural EMF through biotic homogenization resulting from two contrasting biodiversity-related mechanisms including loss of species diversity and dominance of species with acquisitive resource-use strategy.

Methods
Data were collected across three dominant land use (LU) types in the Sudanian and Sudano–Guinean zones in Benin, West Africa. The LU types included forests, savannas and agroforestry parklands. Vegetation and soil data were collected from 145 circular plots of 0.1 ha each. Within each plot, the floristic inventory consisted of counting and measuring the diameter at breast height (DBH, cm) and height (H, m) of all living trees with DBH > 5 cm. Leaf samples were collected from 5–16 individual trees of abundant species across the sampling plots to determine their dry matter content (mg g-1) andnitrogen content (%). Soil samples were collected at a 0–20 cm depth from the center of four subplots of 0.01 ha each that were installed within the main plot. The soil samples were analyzed for organic carbon (%), total nitrogen (%), total phosphorus (%), and available phosphorus (%) content. Litter samples were collected from four smaller plots of 1 m radius that were established within the four 0.01 ha subplots to quantify litter biomass (Mg ha-1).

Data on environmental conditions were also collected. We characterized the environmental conditions of each sampling plot based on climate, topography, and soil. The climatic variables included aridity index and precipitation seasonality, which were extracted from the WorldClim2 database at a resolution of ~1 km for 1971–2000. We used elevation data recorded by a handheld GPS to represent the topography. Additionally, we downloaded data on soil content (%) of clay, silt, sand, and soil pH from the Africa Soil Profiles Database at a resolution of 1 km (https://www.isric.org) for three standard soil depth intervals (0‒5, 5‒15 and 15–30 cm). We used the weighted mean values of the additional soil variables for the 0–20 cm soil layer in the subsequent analyses.