Abstract
In many parts of Africa, spatially-explicit information on plant α-diversity, i.e., the number of species in a given area, is missing as baseline information for spatial planning. We present an approach on how to combine vegetation-plot databases and remotely-sensed land surface phenology (LSP) metrics to predict plant α-diversity on a regional scale. We gathered data on plant α-diversity, measured as species density, from 999 vegetation plots sized 20 m × 50 m covering all major vegetation units of the Okavango basin in the countries of Angola, Namibia and Botswana. As predictor variables, we used MODIS LSP metrics averaged over 12 years (250-m spatial resolution) and three topographic attributes calculated from the SRTM digital elevation model. Furthermore, we tested whether additional climatic data could improve predictions. We tested three predictor subsets: (1) remote sensing variables; (2) climatic variables; and (3) all variables combined. We used two statistical modeling approaches, random forests and boosted regression trees, to predict vascular plant α-diversity. The resulting maps showed that the Miombo woodlands of the Angolan Central Plateau featured the highest diversity, and the lowest values were predicted for the thornbush savanna in the Okavango Delta area. Models built on the entire dataset exhibited the best performance followed by climate-only models and remote sensing-only models. However, models including climate data showed artifacts. In spite of lower model performance, models based only on LSP metrics produced the most realistic maps. Furthermore, they revealed local differences in plant diversity of the landscape mosaic that were blurred by homogenous belts as predicted by climate-based models. This study pinpoints the high potential of LSP metrics used in conjunction with biodiversity data derived from vegetation-plot databases to produce spatial information on a regional scale that is urgently needed for basic natural resource management applications.
Highlights
Biodiversity is declining at a high rate [1], and international treaties, such as the Convention on Biological Diversity, pledged to halt biodiversity loss
This study pinpoints the high potential of land surface phenology (LSP) metrics used in conjunction with biodiversity data derived from vegetation-plot databases to produce spatial information on a regional scale that is urgently needed for basic natural resource management applications
Depending on the dataset, the algorithmic prediction error varies in magnitude, boosted regression trees (BRT) and random forests (RF) are both machine learning techniques based on regression trees and exhibited comparable model performance
Summary
Biodiversity is declining at a high rate [1], and international treaties, such as the Convention on Biological Diversity, pledged to halt biodiversity loss. The recent discussion on ‘essential biodiversity variables’ has shown that remote sensing applications are indispensable in the process and are needed to monitor changes in biodiversity over large areas with a consistent methodology [2,3]. In this context, field-based ecological data play a prominent role as baseline data for biodiversity models and as ground truth information for remote sensing applications. Created a meta-database containing over 200 existing vegetation-plot databases worldwide with over three million vegetation plots [4,5] These databases harbor an enormous potential as ground truth information for future remote sensing studies and spatial modeling approaches from local to global scales; yet, so far this potential remains unexploited.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
