Comparison of spatial and spectral properties of Landsat-8 and Sentinel-2 data for mapping plant chlorophyll-a

Abstract

Background Chlorophyll-a (Chl-a) is a vital parameter to assess vegetation quality in plants as an indicator of photosynthetic capacity to ensure proper flow of ecosystem services. Nowadays, with a rapid increase in human population and deforestation chl-a in higher plants remain at risk from degradation. Aim The study sought to compare the spatial and spectral properties Landsat-8 and Sentinel-2 in estimating and mapping chlorophyll-a (chl-a) concentrations in the Vhembe District Municipality (VDM), South Africa. Methods Landsat-8 and Sentinel-2 multispectral data were used in conjunction with field data collected in August 2017, Firstly, this study assessed the correlation between chl-a and satellite data. Secondly, explored the optimal spatial resolution for mapping chlorophyll-a with stepwise multiple linear regression, and lastly, this study mapped the concentration of plant chl-a across a heterogeneous landscape. Results When assessing the correlation between chl-a and satellite data there was an obvious correlation between chlorophyll-a and Band 5 entropy with the highest R² of 0.39 at 30 m spatial scale of Landsat-8. However, there was no statistical significant difference amongst the various spatial resolution. The ability of Gray Level Co-occurrence Matrix (GLCM) texture features with Landsat-8 at medium resolution 30 m with R² = 0.55, p = 0.000006, and RMSE = 0.17 μg/m² in estimating plant chl-a yielded higher performance accuracy than Sentinel-2 at 10 m resolution with R² = 0.24, p = 0, and RMSE = 0.46 μg/m², and 20 m resolution with R² = 0.52, p = 0.00001, and RMSE = 6.90 μg/m². In exploring the optimal spatial resolution, Landsat-8 at 30 m spatial resolution was optimal for mapping plant chlorophyll-a. Lastly, plant chl-a were successfully mapped with Landsat-8 multispectral data at 30 m spatial resolution using multiple linear regression. The distribution of plant chlorophyll-a varies across the study area and is unevenly distributed due to different species and height. Discussions Chlorophyll-a as a crucial parameter in plants and requires continuous monitoring to ensure and improve ecosystem services provided by plants. This study estimated plant chlorophyll-a across the Vhembe District Municipality. The correlation coefficients derived by GLCM’s features demonstrated the ability of GLCM’s features in predicting and mapping plant chlorophyll-a with Landsat-8 at 30 m spatial resolution. Several studies have successfully mapped chlorophyll-a from a homogeneous xii landscape, for instance in agricultural crops and limited studies mapped chlorophyll-a calibrated from a heterogeneous landscape which motivated this study. Recent advancement in optical remote sensing data opens new avenues for mapping plant chlorophyll-a at various spatial resolution. Conclusion The use of remote sensing data at 30 m spatial resolution with GLCM features effectively predicted plant chlorophyll-a and enable the data processing and performance accuracy assessment. Therefore, this study highlighted the importance of Landsat-8 imagery in vegetation monitoring across a heterogeneous landscape.