A new perspective on vegetation water dynamics
Assimilating ASCAT observations to constrain soil and vegetation states using a data-driven observation operator
X. Shan (TU Delft - Water Resources)
S.C. Steele-Dunne – Promotor (TU Delft - Mathematical Geodesy and Positioning)
F.J. Lopez Dekker – Promotor (TU Delft - Mathematical Geodesy and Positioning)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
In the current generation, most land surface models (LSMs) do not explicitly model the plant hydraulic states or fluxes, which limits the ability of LSMs to model evapotranspiration [1], stomatal conductance [2], and monitor and predict drought [3]. Therefore it is necessary to constrain the canopy water dynamics in LSMs. Advanced SCATterometer (ASCAT) provides a long record of C-band backscatter since 2007. A key advantage of the ASCAT instrument is the ability to obtain measurements of the Earth’s surface from different incidence angles. The dependence of ASCAT backscattering coefficient (hereafter referred to as backscatter) on incidence angle provides valuable information about vegetation water dynamics via normalized backscatter (σo 40), and vegetation parameters (slope (σ′), and curvature (σ′′)) of the Taylor expansions of backscatter to incidence angle [4–7]. In this thesis, the ASCAT normalized backscatter and slope are assimilated into the "Interactions between soil, biosphere and atmosphere" (ISBA-A-gs, hereafter referred to as ISBA) LSM.
In order to assimilate ASCAT observables, an observation operator is needed to link between LSM states to radar observations. Radiative transfer models (RTMs) are often used to assimilate radar backscatter into LSM [8, 9]. However, RTMs require moisture content or dielectric properties of soil and vegetation cover which are not simulated by the LSM. Therefore, to directly link land surface states and ASCAT observables, a Deep Neural Network (DNN) was trained and validated in Chapter 3. The performances and sensitivity of theDNNwere evaluated tomake sure the observation operator is physically plausible...