Spectral Discrimination of Arundo donax L. and Phragmites australis (Cav.) Grown Under Control and Accumulation of Heavy Metal Conditions

Abstract

An approximate 50% decline in wetland loss worldwide has been estimated since the end of the 19th century (Kumari et al., 2020). The escalating contamination of wetlands with heavy metals further poses significant threats to wetlands. As much as wetland plants may uptake heavy metals to aid in heavy metal reduction in the environment, contamination may have an impact on their physiological and biochemical properties. Impact in these properties may influence the spectral signature of the species. This study aims to spectrally discriminate between Arundo donax L. and Phragmites australis (Cav.) species under control and heavy metal conditions. The specific objectives are: (i) To discriminate between A. donax and P. australis species, both under control conditions and subjected to heavy metal treatment using field spectral measurements. (ii) Evaluate the discriminatory potential of resampled HyMap, WorldView-2 and WorldView-3, and Sentinel 2 data for distinguishing A. donax and P. australis under control and treatment conditions. (iii) To analyse the shift in the REP of A. donax and P. australis under both control and copper + acid treatment conditions. To do this, a glasshouse experiment with the two species where half of the tubs were exposed to copper and acid treatment (which mimic an acid mine drainage) and others left as controls was setup. Field spectroradiometer was used to take spectral reflectance measurements and features obtained from GGRF were used to discriminate the species using RF technique. Hyperspectral data was resampled to the different multispectral and hyperspectral sensors. Moreover, the concentration of copper in all the tubs were measured. There was no significant difference in the spectral reflectance of A. donax under control and treatment conditions. There is also no significant difference in the spectral reflectance when discriminating P. australis under control and treatment conditions. However, there is a slight significance when discriminating the two species when they are both under control. A significant difference in the spectral reflectance was obtained when discriminating the two species when exposed to the treatment. Random Forest classification showed an improved discrimination accuracy of 83.5% post-treatment, compared to 76.1% under control conditions. This shows discrimination was more effective after treatment, emphasizing the impact of environmental stressors on spectral responses. Resampling of hyperspectral data, especially using HyMap and WorldView-3, showed promise for discriminating between the species. The red edge position analysis indicated a slight stress-induced shifts of 2nm in P. australis (719 to 717 nm), while A. donax remained stable (719nm), suggesting resilience. It can be concluded that the species were better discriminated after the treatment of copper and acid, compared to when they are under control conditions.