Retrieval of vegetation properties using Top of Atmosphere radiometric data: A multi-sensor approach

Abstract

Vegetation is the main source of primary production and plays an important role inmodeling the exchanges of energy and mass of carbon, oxygen and water between the earth and the atmosphere. Mapping andmonitoring of vegetation canopies are crucial for various applications including agro-ecosystem models, climatology modeling, crop growth modeling and vegetation productivity modeling. The reflected radiance from vegetation carries information about chemical and bio-geophysical vegetation characteristics. Leaf area index (LAI), chlorophyll content and fraction of vegetation cover (fCover) are among the most important vegetation properties that can be retrieved from satellite observations. Over the past decades, considerable effort has been given toward improving vegetation properties retrieval fromremotely sensed data. A number of satellite observations are currently being utilized to retrieve vegetation characteristics in a diverse range of spatial and temporal resolutions and new ones will become available soon with an unprecedented capacity to quantify vegetation properties. The accurate vegetation variable retrieval depends on three issues: radiometric data quality, forward radiance modeling and parametrization and regularization. The major goal of this contribution is to improve vegetation properties retrieval from top of atmosphere optical satellite observations, given the richness of current and near-future sources of optical remote sensing data. The first part of this contribution deals with model parameterization for variable retrieval and determining retrievable variables through sensitivity analysis. The sensitivity of top of atmosphere radiance and surface reflectance of a soil-vegetation systemto input biophysical and biochemical variables is estimated using the coupled Soil-Leaf-Canopy radiative transfer model SLC and MODTRAN. This study also proposes an improvement to the design and sampling of screening methods for efficient sensitivity analysis of computationally expensive models. The results demonstrated high correlation between the proposed improvement (with only modest computational demands) against variance-based global sensitivity analysis in determining the most influential and non-influential variables. The second part of this contribution provides a proof of concept for a multi–temporal, multi–sensor approach to retrieve biophysical and biochemical vegetation variables using spectral–directional radiometric data. The approach is designed for the exploitation of a temporal sequence constructed by combining data acquired by different sensors over time. Focus is given to the retrieval of three important vegetation variables; LAI, fCover and chlorophyll content over the agricultural test site in Barrax, Spain. A variety of different satellite observations including, CHRIS–Proba, Landsat TM and ASTER are used to invert the coupled surface–atmosphere SLC–MODTRAN radiative transfer model. This thesis presents an overview of the results and challenges in utilizing the multi–temporal, multi–sensor approach and the Bayesian inversion technique in the retrieval of terrestrial vegetation properties. It was shown that the approach is capable of dealing with a diversity of optical sensors by exploiting heterogeneous radiometric data and allows for having frequent updates of vegetation biophysical and biochemical products over time. Moreover, the results showed that integrating information from different sensors improves the retrieval of vegetation variables compared to the case of single sensor retrievals. The last part of this contribution is devoted to the investigation of the topographic effects on top of atmosphere radiance modeling and variable retrieval. To account for such effects, this study presents an extension of “the four-stream radiative transfer theory” previously proposed by Verhoef and Bach (2003b, 2007, 2012). The extension is mainly focused on two aspects in top of atmosphere radiance modeling. First, the study gives a detailed account of themain topography-induced effects on top of atmosphere radiance modeling and the relevant equations are given. Second, a new formulation is proposed for the derivation of the six atmospheric coefficients using MODTRAN in order to make it more computationally efficient, as well as to avoid using zero surface albedo which causes some miscalculations in the results. It was demonstrated that the topographic effect has to be taken into account in variable retrieval and the approach presented in this thesis can be used tomodel such effects on top of atmosphere radiance over rugged terrain.