Investigating the Relationship between Tree Species Diversity and Landsat-8 Spectral Heterogeneity across Multiple Phenological Stages

Abstract

The emergence of the spectral variation hypothesis (SVH) has gained widespread attention in the remote sensing community as a method for deriving biodiversity information from remotely sensed data. SVH states that spectral heterogeneity on remotely sensed imagery reflects environmental heterogeneity, which in turn is associated with high species diversity and, therefore, could be useful for characterizing landscape biodiversity. However, the effect of phenology has received relatively less attention despite being an important variable influencing plant species spectral responses. The study investigated (i) the effect of phenology on the relationship between spectral heterogeneity and plant species diversity and (ii) explored spectral angle mapper (SAM), the coefficient of variation (CV) and their interaction effect in estimating species diversity. Stratified random sampling was adopted to survey all tree species with a diameter at breast height of > 10 cm in 90 × 90 m plots distributed throughout the study site. Tree species diversity was quantified by the Shannon diversity index (H′), Simpson index of diversity (D2) and species richness (S). SAM and CV were employed on Landsat-8 data to compute spectral heterogeneity. The study applied linear regression models to investigate the relationship between spectral heterogeneity metrics and species diversity indices across four phenological stages. The results showed that the end of the growing season was the most ideal phenological stage for estimating species diversity, following the SVH concept. During this period, SAM and species diversity indices (S, H′, D2) had an r2 of 0.14, 0.24, and 0.20, respectively, while CV had an r2 of 0.22, 0.22, and 0.25, respectively. The interaction of SAM and CV improved the relationship between the spectral data and H′ and D2 (from r2 of 0.24 and 0.25 to r2 of 0.32 and 0.28, respectively) at the end of the growing season. The two spectral heterogeneity metrics showed differential sensitivity to components of plant diversity. SAM had a high relationship with H′ followed by D2 and then a lower relationship with S throughout the different phenological stages. Meanwhile, CV had a higher relationship with D2 than other plant diversity indices and its relationship with S and H′ remained similar. Although the coefficient of determination was comparatively low, the relationship between spectral heterogeneity metrics and species diversity indices was statistically significant (p < 0.05) and this supports the assertion that SVH could be implemented to characterize plant species diversity. Importantly, the application of SVH should consider (i) the choice of spectral heterogeneity metric in line with the purpose of the SVH application since these metrics relate to components of species diversity differently and (ii) vegetation phenology, which affects the relationship that spectral heterogeneity has with plant species diversity.

Spatial Characterisation of Vegetation Diversity in Groundwater-Dependent Ecosystems Using In-Situ and Sentinel-2 MSI Satellite Data

“…Findings from this study are supported by empirical data from field-measured diversity, which explained at least 61% of the remote sensingderived vegetation diversity (Rao's Q). When compared to other studies-for example, Madonsela, Cho, Ramoelo and Mutanga [83]; Madonsela, Cho, Ramoelo and Mutanga [65]; Rocchini, Chiarucci and Loiselle [25]; and Lopes, Fauvel, Ouin and Girard [101]-our observations are more robust. For example, Madonsela, Cho, Ramoelo and Mutanga [83] and Rocchini, Chiarucci and Loiselle [25] reported a low R 2 (0-0.48) when explaining vegetation diversity in South Africa and Italy, respectively.…”

“…We tested 13 measures of spectral variation. These measures include coefficient of variation (CV) [65], Enhanced Vegetation Index (EVI) [66], Enhanced Vegetation Index (EVI) 2 [67], Modified Soil Adjusted Vegetation Index (MSAVI) 2 [68], Normalized Difference Vegetation Index (NDVI) [69], Normalized Difference Phenology Index (NDPI) [70], Optimized Soil Adjusted Vegetation Index (OS-AVI) [71], Simple Ratio Index [72], Soil Adjusted Vegetation Index (SAVI) [73], Renormalized Difference Vegetation Index (RDVI) [74], and Tasseled-cap Greenness Index (TCI) [75]. In addition, we also used all the spectral bands from Sentinel-2 MSI and the first principal component of the spectral bands after principal component analysis (PCA).…”

“…The low performance of the Normalized Difference Phenology Index and the first principal component in estimating the Rao's Q can be explained by how these metrics capture spectral heterogeneity more in vegetated landscapes than in arid environments with intense soil reflectance [70,82]. In addition, these measures might be more sensitive to a specific component of diversity (i.e., evenness, abundance, or dominance) [65,83]. Henceforth, these metrics might perform better in areas with high environmental heterogeneity (e.g., temperate forests) or with a different measure of diversity [76,84].…”

“…In this study, CV was used to measure the variability in spectral reflectance amongst pixels from all the spectral bands [27,84]. CV is highly sensitive to the reflectance of rare and abundant species and is a good measure of spectral variation [65,84]. The CV has been used in different studies [27,37,65], focusing on monitoring different aspects of the environment [85,86].…”

“…CV is highly sensitive to the reflectance of rare and abundant species and is a good measure of spectral variation [65,84]. The CV has been used in different studies [27,37,65], focusing on monitoring different aspects of the environment [85,86]. Owing to its ability for detecting spectral heterogeneity, the high performance of the CV over other measures of spectral variation is expected [65,85].…”