Potionential of Spectral Indices for Halophyte Vegetation Cover Detection in Arid and Salt-Affected Landscape

In arid and marginal landscape, characterized by saline soils and significant climatic fluctuations, halophyte vegetation covers (tolerating salinity) are subject to special interest to maximize crop productivity and soil rehabilitation. In such context, monitoring vegetation dynamic in space and time is a fundamental process for biodiversity and sustainable development. In this context, the present paper evaluate the potential of the most popular vegetation indices (VIs), such as NDVI, SAVI, EVI and TDVI, for providing a better relationships with halophyte vegetation covers; as well as for eliminating or at least minimising the soil background artefacts in arid and saline landscape. This evaluation is based on a Landsat-OLI image, corrected from all radiometric anomalies, and a field survey. The latter allowed the identification and localization (using accurate GPS) of different halophytic vegetation cover rates (5%, 10%, 50%, 70%, and 100%). Moreover, 100 bare-soil samples with different contents of salinity and optical properties were selected. Statistical test has been undertaken using regression analysis (p < 0.05) between observed and predicted variables. The results obtained show that the considered VIs fit significantly with vegetation cover rates yielding an R2 of 0.91, 0.92, 0.96 and 0.98 for EVI, SAVI, NDVI and TDVI, respectively. However, different behaviors and sensitivities are observed. The EVI exhibit some sensitivity to the artefacts of underlying saline soils with an RMSE of 35%, as well it overestimates the scattered and medium vegetation cover (0 to 70%), and it saturate for the dense cover (75% to 100%). Although the RMSE caused by the soil background sensitivity are very low for NDVI and SAVI, these two indices underestimate medium and dense covers with serious linearity-saturation problems. Indeed, for a completely dense cover (100%), their prediction values do not exceed 40% for NDVI and 55% for SAVI. Despite its relatively limited sensitivity to soil artifacts with an RMSE around 9%, the TDVI perform relatively better illustrating reasonably a linear behavior. Their predictive values do not exceed 80% for a completely dense cover (100%). Accordingly, this index can allow relatively better detection and quantification of halophytic vegetation cover rates in arid and saline landscapes.