Multi-Temporal Changes Analysis of Natural Vegetation Cover Using Serial NDVI and Metric Indices: Case of Tlemcen National Park (Northwest of Algeria)

Abstract

Assessing the evolution of vegetation cover fragmentation over time is a key-issue for monitoring biodiversity and sustainable development in land ecosystems. In the present study, landscape spatial metrics were investigated to assess changes of multitemporal vegetation cover within the Tlemcen National Park (TNP) in Northwest of Algeria over 31 years. To achieve this, five landscape metrics including number of patches (NP), landscape shape index (LSI), class area (CA), largest patch index (LPI), and percentage of landscape (PLAND) were applied and analysed to assess the fragmentation at class level. While, only NP and LSI were analyzed at the landscape level. Three Landsat images acquired during September 1987, 1999 and 2018 with TM, ETM+ and OLI sensors, respectively, were used. After rigorous preprocessing steps, the images were transformed and threshold into NDVI maps representing the major land use classes in TNP, such as forest, shrub lands, open shrub, scattered vegetation and bare soil. For validation purposes, Kappa coefficient analysis was carried out based on 120 ground control points representing these land use classes. The obtained results revealed that the generated confusion matrix ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{p} &lt; 0.05$</tex> ) yield a Kappa coefficient of 0.77 and a significant concordance coefficient of 0.83 between the observed (ground truth) and predicted classes by NDVI-2018. These results indicate that the thresholding procedure is very appropriate and extendable to the other NDVI maps derived for the years 1987 and 1999. Moreover, at the landscape level, the NP and LSI metrics have increased over time by 67.6% in 1999 and 27.9% in 2018, revealing a gradual and important fragmentation of the landscape at TNP. In addition, this trend is corroborated at the class level by the five considered metrics. Accordingly, the synergy between metric indices and remote sensing data offers great potential for sustainable and efficient management policies to address the issues of spatiotemporal fragmentation and degradation across national parks.