Structural diversity and tree density drives variation in the biodiversity-ecosystem function relationship of woodlands and savannas.

Positive biodiversity-ecosystem function relationships (BEFRs) have been widely documented, but it is unclear if BEFRs should be expected in disturbance-driven systems. Disturbance may limit competition and niche differentiation, which are frequently posited to underlie BEFRs. We provide the first exploration of the relationship between tree species diversity and biomass, one measure of ecosystem function, across southern African woodlands and savannas, an ecological system rife with disturbance from fire, herbivores and humans. We used > 1000 vegetation plots distributed across 10 southern African countries and structural equation modelling to determine the relationship between tree species diversity and above-ground woody biomass, accounting for interacting effects of resource availability, disturbance by fire, tree stem density and vegetation type. We found positive effects of tree species diversity on above-ground biomass, operating via increased structural diversity. The observed BEFR was highly dependent on organismal density, with a minimum threshold of c. 180 mature stems ha-1 . We found that water availability mainly affects biomass indirectly, via increasing species diversity. The study underlines the close association between tree diversity, ecosystem structure, environment and function in highly disturbed savannas and woodlands. We suggest that tree diversity is an under-appreciated determinant of wooded ecosystem structure and function.

Human and natural controls of the variation in aboveground tree biomass in African dry tropical forests.

Understanding the anthropogenic and natural controls that affect the patterns, distribution, and dynamics of terrestrial carbon is crucial to meeting climate change mitigation objectives. We assessed the human and natural controls over aboveground tree biomass density in African dry tropical forests, using Zambia's first nationwide forest inventory. We identified predictors that best explain the variation in biomass density, contrasted anthropogenic and natural sites at different spatial scales, and compared sites with different stand structure characteristics and species composition. In addition, we evaluated the effects of different management and conservation practices on biomass density. Variation in biomass density was mostly determined by biotic processes, linked with both species richness and dominance (evenness), and to a lesser extent, by land use, environmental controls, and spatial structure. Biomass density was negatively associated with tree species evenness and positively associated with species richness for both natural and human-modified sites. Human influence variables (including distance to roads, distance to town, fire occurrence, and the population on site) did not explain substantial variation in biomass density in comparison to biodiversity variables. The relationship of human activities to biomass density in managed sites appears to be mediated by effects on species diversity and stand structure characteristics, with lower values in human-modified sites for all metrics tested. Small contrasts in carbon density between human-modified and natural forest sites signal the potential to maintain carbon in the landscape inside but also outside forestlands in this region. Biodiversity is positively related to biomass density in both human and natural sites, demonstrating potential synergies between biodiversity conservation and climate change mitigation. This is the first evidence of positive outcomes of protected areas and participatory forest management on carbon storage at national scale in Zambia. This research shows that understanding controls over biomass density can provide policy relevant inputs for carbon management and on ecological processes affecting carbon storage.

A global land cover training dataset from 1984 to 2020.

State-of-the-art cloud computing platforms such as Google Earth Engine (GEE) enable regional-to-global land cover and land cover change mapping with machine learning algorithms. However, collection of high-quality training data, which is necessary for accurate land cover mapping, remains costly and labor-intensive. To address this need, we created a global database of nearly 2 million training units spanning the period from 1984 to 2020 for seven primary and nine secondary land cover classes. Our training data collection approach leveraged GEE and machine learning algorithms to ensure data quality and biogeographic representation. We sampled the spectral-temporal feature space from Landsat imagery to efficiently allocate training data across global ecoregions and incorporated publicly available and collaborator-provided datasets to our database. To reflect the underlying regional class distribution and post-disturbance landscapes, we strategically augmented the database. We used a machine learning-based cross-validation procedure to remove potentially mis-labeled training units. Our training database is relevant for a wide array of studies such as land cover change, agriculture, forestry, hydrology, urban development, among many others.