University of Twente Student Theses


Biodiversity Assessment Using Satellite Imagery in San Joaquin Experimental Range, California

Boakye, Ama Serwah (2023) Biodiversity Assessment Using Satellite Imagery in San Joaquin Experimental Range, California.

[img] PDF
Abstract:In the face of human-induced biodiversity loss, biodiversity monitoring serves as a crucial feedback link supporting the development of informed decisions for conservation management. Remote sensing tool has been demonstrated as a valuable technique for modelling biodiversity across various scales of biological organization, utilizing different levels of spatial granularity and temporal resolution compared to the time and labour-intensive field-based studies for collecting reliable biodiversity data. Hyperspectral data provide extensive information in estimating biodiversity by revealing distinct biochemical and biophysical features of plant species but are limited and expensive. In contrast, multispectral sensors such as Sentinel-2 and Landsat allow monitoring for the continuous acquisition of satellite images at a larger spatial extent with higher temporal coverage. However, their lower spatial and spectral resolution limits their effect in discriminating detailed object characteristics. The Spectral Diversity Hypothesis (SDH) is an emerging approach relying on optical remote sensing to assess and monitor spectral diversity. SDH proposes that the spectral diversity in remote sensing images reflects the spatial heterogeneity within the environment. In this study, the SDH was applied to assess the species richness in the Mediterranean region and investigate the potential of the SDH in biodiversity estimation using multispectral data compared to hyperspectral data. As such, this study involved pre-processing remote sensing data, reducing data dimensionality using PCA to capture the spectral variation, and applying the K-means classification on selected PCs to identify distinct spectral species in SJER. Next, the data were validated by comparing the field plot data. The study revealed a weak correlation (R2 = 0.0639 for AVIRIS-NG and R2 = 0.0940 for Landsat 8 data) between spectral diversity and species diversity in the Mediterranean ecosystem, as the application of SDH in this region is characterized by low vegetation cover with strong influence from the soil and non-photosynthetically active vegetation. However, it was revealed that the closed forest areas in the region yielded a better accuracy with R2 of 0.65 for Landsat, with AVIRIS-NG having an accuracy of R2 of 0.68. In contrast, the weakest relationships were observed in open forest areas of the region. Additionally, the study involved spatially and spectrally upscaling the AVIRIS-NG data to simulate Landsat data, aiming to evaluate the impact of the spatial and spectral resolution on species richness derived from remote sensing datasets. The study indicated that the spatial component played a significant role in the discrimination of clusters compared to the spectral component. Reducing the spatial resolution causes the pixels to be more heterogenous as well as to reduce the effect of background information, thereby affecting the spectral mixing of clusters. Consequently, Landsat and AVIRIS-NG images detected consistent patterns of areas with low species richness. Vegetation clusters found in the real Landsat data exhibited spatial consistency with clusters in AVIRIS-NG. Furthermore, The NIR and SWIR spectral domains in both real Landsat and AVIRIS-NG contributed the most to the principal components of AVIRIS-NG and Landsat data. Overall, these research results contribute valuable insights into analysing remote sensing data for biodiversity assessment for upcoming hyperspectral satellite missions like PRISMA and EnMap for monitoring biodiversity in the Mediterranean region
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Subject:38 earth sciences
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page