Global Sensitivity Analysis of the SCOPE Model in Sentinel-3 Bands: Thermal Domain Focus

Egor Prikaziuk,Christiaan Van Der Tol

doi:10.3390/rs11202424

Abstract

Sentinel-3 satellite has provided simultaneous observations in the optical (visible, near infrared (NIR), shortwave infrared (SWIR)) and thermal infrared (TIR) domains since 2016, with a revisit time of 1–2 days. The high temporal resolution and spectral coverage make the data of this mission attractive for vegetation monitoring. This study explores the possibilities of using the Soil Canopy Observation, Photochemistry and Energy fluxes (SCOPE) model together with Sentinel-3 to exploit the two sensors onboard of Sentinel-3 (the ocean and land color instrument (OLCI) and sea and land surface temperature radiometer (SLSTR)) in synergy. Sobol’ variance based global sensitivity analysis (GSA) of top of atmosphere (TOA) radiance produced with a coupled SCOPE-6S model was conducted for optical bands of OLCI and SLSTR, while another GSA of SCOPE was conducted for the land surface temperature (LST) product of SLSTR. The results show that in addition to ESA level-2 Sentinel-3 products, SCOPE is able to retrieve leaf area index (LAI), leaf chlorophyll content (Cab), leaf water content (Cw), leaf senescent material (Cs), leaf inclination distribution (LAD). Leaf dry matter content (Cdm) and soil brightness, despite being important, were not confidently retrieved in some cases. GSA of LST in TIR domain showed that plant biochemical parameters—maximum carboxylation rate (Vcmax) and stomata conductance-photosynthesis slope (Ball-Berry m)—can be constrained if prior information on near-surface weather conditions is available. We conclude that the combination of optical and thermal domains facilitates the constraint of the land surface energy balance using SCOPE.

Highlights

The retrieval of vegetation parameters from remote sensing data in a multi-dimensional parameter space is a challenging task [1]
Several studies have quantified the atmospheric effects by simulating the propagation of top of canopy (TOC) reflectance to top of atmosphere (TOA) radiance using the models MODTRAN [24,25] for Hyperion on EO-1 [26], CHRIS on Proba-1, TM on Landsat 5 and ASTER on Terra [27] and 6S [28] for VEGETATION on SPOT [29], MODIS on Aqua and Terra [30] and vegetation indices derived from TM and ETM+ on Landsat 5 and 7 respectively [31]
In general our results agree with the those studies: leaf area index (LAI) was a dominant factor throughout the spectral range (400–2400 nm), leaf inclination distribution (LAD) showed slight importance in all wavelength with a substantial peak in 700–1300 nm, leaf chlorophyll content (Cab) affected visible range with peaks in green and red-edge regions, leaf dry matter (Cdm) was important in NIR and shortwave infrared (SWIR) region, leaf water content (Cw)—SWIR region, especially above 1300 nm

Summary

Introduction

The retrieval of vegetation parameters from remote sensing data in a multi-dimensional parameter space is a challenging task [1]. In a GSA of PROSAIL and PROGEOSAIL in relation to fuel moisture content, Bowyer and Danson [15] showed that leaf area index (LAI) and fraction of vegetation cover dominated reflectance in shortwave infrared (SWIR) region This dominance of vegetation coverage is reduced if the sensitivity analysis is carried out separately for sparse, intermediate and dense vegetation [16]. Several studies have quantified the atmospheric effects by simulating the propagation of top of canopy (TOC) reflectance to top of atmosphere (TOA) radiance using the models MODTRAN [24,25] for Hyperion on EO-1 (decommissioned in 2017) [26], CHRIS on Proba-1, TM on Landsat 5 and ASTER on Terra [27] and 6S [28] for VEGETATION on SPOT (decomissioned in 2015) [29], MODIS on Aqua and Terra (bands 1–7) [30] and vegetation indices derived from TM and ETM+ on Landsat 5 and 7 respectively [31]. GSA of full range PROSAIL-MODTRAN TOA radiance spectra (400–2500 nm) has recently been reported [32]

Objectives

Methods

Results

Discussion

Conclusion