Mapping an alien invasive shrub within conservation corridors using super-resolution satellite imagery

Abstract

Alien invasive plant species are one of the main drivers of global biodiversity loss. Methods for monitoring the spread of alien invasive plants are needed to improve management and mitigate impact on local biodiversity. Recent advances in deep learning and image fusion holds great potential for mapping and managing alien invasive plants. One such method is super-resolution image reconstruction, where a neural network learns to downscale images from coarse to fine resolution. Within the commercial timber production landscape of KwaZulu-Natal, endangered grassland corridors are threatened by American bramble invasion, impacting plants, birds, arthropods, and soil restoration. Here we aim to improve our understanding of bramble invasion dynamics through using super-resolved satellite mosaics. Bramble was classified with very high accuracies (85%) from the super-resolved satellite mosaic, compared to other conventional satellite imagery with different spectral and spatial resolutions. Using landscape analyses, we identified plantation tree harvesting and prescribed burning to be major drivers increasing bramble cover within the landscape. Bramble cover was highest one year following plantation tree harvesting. Continuous prescribed burning positively influenced bramble. Bramble cover was also high close to streams, and under future invasion projections, bramble will severely impact Ensifera species alongside low priority grasshopper species habitat. Results also indicate that bramble has a significant negative impact on intermediate priority grasshoppers and plant species richness. For controlling bramble invasion within commercial timber production landscapes, we recommend the adoption rotational harvesting, as harvesting entire plantation blocks throughout the landscape will dramatically increase invasion potential of bramble. Current bramble removal programmes should prioritize riparian areas. Special attention is needed to control bramble one year after timber harvesting, as this is when bramble cover is highest. We show the benefits of using super-resolved mosaics to gain new insights into alien invasive species dynamics, while further development of this technique will aid in managing invasive alien plant species at local scales.

Keywords: Conservation corridors; Deep learning; Invasion dynamics; Management; Plantation forestry; Remote sensing; Rubus cuneifolius.