Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Titre : Estimating the leaf area index of Eucalyptus dunnii in the Midlands area, KwaZulu-Natal province over two seasons using vegetation indices and image texture measures derived from Worldview-3 imagery

Auteur : Mthembu, Nokukhanya Fundiswa.

Université de soutenance : University of KwaZulu-Natal

Grade : Master of Science in the Discipline of Geography2022

Résumé
Leaf Area Index (LAI) remains one of the important forest structural attributes, accurate estimations of LAI are crucial as LAI is a major input variable for 3-PGS to predict growth of different commercial forest species and their water use. While remote sensing offers a faster and effective means of estimating LAI, LAI is seldom available at spatio-temporal scales that can be used to guide and inform management decisions for localised applications. Furthermore, the knowledge relating to spatial and temporal variation of LAI is still limited. This study sought to estimate LAI of Eucalyptus dunnii in the Midlands area using vegetation indices and texture measures derived from WorldView-3 imagery. The first objective was to review previous work on remote sensing methods of estimating LAI across different forest ecosystems, crops and grasslands. The results revealed that during the last decade, the use of remote sensing to estimate and map LAI has increased for crops and natural forests. However, with regards to commercial forests and grasslands, there is still a need for more research as the number of studies is still small. The second objective was to use a combination of vegetation indices and texture measures to estimate LAI. The relationships between LAI and vegetation indices (VI), and LAI and texture were modelled using Partial Least Squares Regression (PLS-R). In terms of LAI estimation using texture, the results showed that combining two or more texture bands leads to improved LAI estimation accuracy. Although texture measures can improve LAI estimation accuracy, very few studies focusing on estimating LAI using texture measures have been published. Vegetation indices alone achieved poor LAI estimation accuracy. The best performing model incorporated texture ratios and it achieved an estimation accuracy of R2=70, RMSE 1.21 in 2019 and R2=0.72, M=RMSE=1.26. Overall, this study demonstrated that texture band ratios can estimate LAI of Eucalyptus dunnii in the Midlands area with acceptable accuracy

Présentation

Version intégrale (1,9 Mb)