A remote sensing-based approach for detecting and mapping the accumulation of heavy metal in Phragmites australis (Cav.) Trin. ex Steud. and Arundo donax L.in wetlands influenced by gold mining
Date
2021
Authors
Mabhungu, Loveness
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Arundo donax L. and Phragmites australis( Cav.) Trin. ex Steud. are two morphologically similar species which play an important role in removal of heavy metals from acid mine drainage (AMD) polluted wetlands. Mapping the distribution of the two species in wetlands and monitoring the level of heavy metal contaminants accumulated in their tissues is very crucial for effective wetland management. Determining the phytoremediating efficiency between the two wetland species would help promoting the native P. australis for a wide use across the country by replacing A. donax, declared as invasive weed category ‘1b’ according to the National Environmental Management: Biodiversity (NEM:BA) Act, 2004 (Act NO. 10 of 2004) of Alien and Invasive Species Lists, 2016,that places it under strict environmental restrictions prohibiting any propagation or spreading of the species. This study investigated the efficiency of A. donax and P. australis in the uptake of copper from artificially simulated acid mine drainage by mixing copper (CuSO4.5H2O)and nitric acid(HNO3) in tubs in a glasshouse experiment. Both A. donax and P. australis were able to uptake substantial amounts of copper from the tubs. There were significant differences between the control and treatment of the leaf and root samples of A. donax with mean differences of 2.876 ± 0.364 and 25.603 ± 3.119 mg/kg, respectively, and the stems and roots of P. australis with mean differences of6.512 ± 2.01 and 16.10 ± 3.62 mg/kg, respectively. The native P. australis had a higher translocation factor (0.4), which makes it more preferable for phytoremediation where harvesting of the above ground biomass is required to remove the accumulated copper for safe disposal, than the alien A. donax (0.2).Using ASD hyperspectral data measured from the glasshouse plants, the utility of the Random Forest (RF) and Support Vector Machines (SVM) classification algorithms coupled with the Guided Regularized Random Forest (GRRF) for feature selection, was tested in the discrimination between the two species grown under control and treatment conditions (copper + nitric acid). Results showed that leaf copper concentration is inversely related to mean leaf spectral reflectance. Spectral discrimination was more efficient between treatment plants of the two species (Overall Accuracy = 85.2% and 81.4% for RF and SVM, respectively) than between the control plants (Overall Accuracy = 64.2% and 75% for RF and SVM, respectively). This was because although the two species are morphologically similar under healthy conditions, and hence similar spectral reflectance, they responded to heavy metal pollution accumulation in their tissues differently, which resulted in SVM in the discrimination of the two plants under treatment conditions. The potential of the multispectral Sentinel-2 data to detect and map A. donax and P. australis effectively was investigated in wetlands around the city of Johannesburg and the results showed that P. australis was more abundant and wide spread than A. donax. Although there was some spectral confusion in separating the two species, the RF and SVM yielded recommendable and equal accuracies in the mapping process (overall accuracy =91.2% and kappa coefficient = 0.89 for both RF and SVM, respectively). The study concludes that the Sentinel-2 data coupled with either RF or SVM classification algorithms were effective for monitoring the A. donax and P. australis species in wetlands
Description
A thesis submitted to School of Geography, Archaeology and Environmental studies, Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfillment of the academic requirements for the degree of Doctor of Philosophy in Geography and Environmental Studies, 2021