Habitat Suitability Modeling for the Invasive Opuntia stricta Using Remote Sensing and Maxent in Tsavo East National Park, Kenya
Lilian Adionyi *
Department of Informatics and Computing, Taita Taveta University, Kenya.
Nashon Adero
School of Mines & Engineering, Taita Taveta University, PO Box 635, 80300, Voi, Kenya.
Samuel Mutua
School of Science and Informatics, Taita Taveta University, PO Box 635, 80300, Voi, Kenya.
David Korir
Wildlife Research and Training Institute, PO Box 842 20117, Naivasha, Kenya.
Mika Siljander
Department of Geosciences and Geography, Helsinki Laboratory of Interdisciplinary Conservation Science (HELICS), University of Helsinki, PO Box 64, 00014, Helsinki, Finland.
*Author to whom correspondence should be addressed.
Abstract
Biological invasions represent a major threat to ecosystem services and products, with the potential to disrupt ecosystems across a broad spectrum of bioclimatic regions. Consequently, it is essential to monitor the spread of invasive species systematically and over extensive areas. Remote sensing and geographic information systems have long been recognized as valuable tools for achieving this goal. This paper examines the efficacy of optical satellite data from different seasons in detecting the invasive species Opuntia stricta (Australian pest pear) within the southern portion of the Tsavo East National Park. Maximum Entropy (MaxEnt) modelling was employed to determine the most relevant environmental variables, with a total of ten predictors tested. The results demonstrated that ndvi2017dry, rvi2018wet, rvi2017wet, msavi2018dry and rvi2018dry were the most effective Vegetation Indices (Vis) for detection of Opuntia stricta. The study also found that seasonal variations played a significant role in enhancing detection accuracy. The fine-scale MaxEnt modelling predicted core areas of invasion, yielding a mean AUC of 0.718. Suitable habitat within the study area was classified into high, medium, low, and very low categories, with 51 km² identified as highly suitable for Opuntia stricta growth. Further classifications included 37 km², 83 km², and 3,589 km² for medium, low, and very low suitability, respectively. Opuntia stricta was detected in 4.15% of the study area. From this study we concluded that the invasive species remain a big risk to protected areas especially in the Tsavo East National Park. Continued monitoring is recommended especially in areas predicted to have high invasion in future. This study provides baseline data for prioritizing invasive species monitoring and management strategies.
Keywords: AUC, biological invasion, MaxEnt, opuntia stricta, remote sensing, sentinel-2, vegetation indices (Vis)