Huesca Martinez, Margarita: University of Twente (Project Lead)
Project Funding:
2017 - 2019
NRA: 2016 NASA: Utilization of Airborne Visible/Infrared Imaging Spectrometer- Next Generation Data from an Airborne Campaign in India-AVRSNG
Funded by NASA
Abstract:
Biodiversity is a critical indicator of ecosystem functioning and health. Greater native biodiversity (i.e. not considering invasive species) enhances ecosystem function, productivity, resilience, variability of the genetic pool, and a multitude of other ecosystem services. Recent studies have started to explore the potential of remote sensing to estimate biodiversity attributes and patterns in natural and managed ecosystems, as spectral variability is expected to be correlated with biodiversity richness, its traits and function. Most studies to date have used moderate resolution satellites to investigate species richness and alpha biodiversity using NDVI and other indices, with somewhat mixed results. That is, significant correlations frequently exist but with poor or low predictive power. Few studies have had access to image spectroscopy data, which allows greater power in determining the relationships between spectral variability and biodiversity. India is characterized by high biodiversity forests that provide an opportunity to assess the extent that biodiversity attributes and patterns can be measured with imaging spectrometry. The Mudumalai National Park in Tamil Nadu (Western Ghats mountains) with about 200 forest species, provides a good test for estimating biodiversity because forest types range from moist tropical deciduous forests (both intact and degraded), to moderately diverse dry deciduous forests, and low diversity tropical dry thorn forests. We will use species data from the 19 long-term research plots in the park that are maintained by the Centre for Ecological Sciences, Indian Institute of Science and one 50-ha plot (72 forest species) from the Smithsonian Institution’s Forest Global Earth Observatory, to test and validate our results. Alpha biodiversity (α) is measured by the species number and abundance at the local level and beta diversity (β) as the variance among sites within a local area, or the overall variance at the regional scale is measured by gamma diversity (γ). Of the three diversity indicators, the one that most directly be measured by remote sensing is α-plant diversity (i.e. species richness in a local area). This measure takes advantage of the spectral variability hypothesis, which states that greater spectral variability is correlated with higher biodiversity. We will test methods that have been used in previous studies for quantifying alpha diversity based on spectral heterogeneity but we will use a broader range of methods such as Euclidian distances, univariate and multivariate regression using individual bands and PCA bands. We will use narrow-band vegetation indices to calculate spatial heterogeneity but also full spectrum methods (e.g., spectral matching). We propose to test a new method to model α-diversity that takes advantage of the large number of bands in AVIRIS-NG imagery, using agglomerative hierarchical clustering analysis to identify the number of classes within a standard plot area. It is a highly flexible algorithm that makes no assumptions about data structure or number of clusters prior to clustering and can be used for any form of similarity or distance metric. Clustering will be repeated for several plot sizes to determine how the α-diversity changes with changing plot size. The advantage of this method over previous methods is that it does not fix the maximum species richness in the area a priori. We will compare heterogeneity between plots (β-diversity) based on distance estimation among pair of plots in terms of α-diversity using several dissimilarity coefficients found in the ecological literature. Lastly, we will assess biodiversity at landscape scale (γ-diversity) in different tropical deciduous forests types and by measuring it across the AVIRIS-NG coverage of the park. Our results will be compared to field data from our collaborator’s sites to validate the biodiversity predictions.
More details may be found in the following project profile(s):
