Hyperspectral reflectance measurements from UAS under intermittent clouds: Correcting irradiance measurements for sensor tilt

Christian J Köppl,Radu Malureanu,Carsten Dam-Hansen,Sheng Wang,Hongxiao Jin,Stefano Barchiesi,Juan M Serrano Sandí,Rafael Muñoz-Carpena,Mark Johnson,Ana M Durán-Quesada,Peter Bauer-Gottwein,Ursula S Mcknight,Monica Garcia

doi:10.1016/j.rse.2021.112719

Abstract

One great advantage of optical hyperspectral remote sensing from unmanned aerial systems (UAS) compared to satellite missions is the possibility to fly and collect data below clouds. The most typical scenario is flying below intermittent clouds and under turbulent conditions, which causes tilting of the platform. This study aims to advance hyperspectral imaging from UAS in most weather conditions by addressing two challenges: (i) the radiometric and spectral calibrations of miniaturized hyperspectral sensors; and (ii) tilting effects on measured downwelling irradiance. We developed a novel method to correct the downwelling irradiance data for tilting effects. It uses a hybrid approach of minimizing measured irradiance variations for constant irradiance periods and spectral unmixing, to calculate the spectral diffuse irradiance fraction for all irradiance measurements within a flight. It only requires the platform's attitude data and a standard incoming light sensor. We demonstrated the method at the Palo Verde National Park wetlands in Costa Rica, a highly biodiverse area. Our results showed that the downwelling irradiance correction method reduced systematic shifts caused by a change in flight direction of the UAS, by 87% and achieving a deviation of 2.78% relative to a on ground reference in terms of broadband irradiance. High frequency (< 3 s) irradiance variations caused by high-frequency tilting movements of the UAS were reduced by up to 71%. Our complete spectral and radiometric calibration and irradiance correction can significantly remove typical striped illumination artifacts in the surface reflectance-factor map product. The possibility of collecting precise hyperspectral reflectance-factor data from UAS under varying cloud cover makes it more operational for environmental monitoring or precision agriculture applications, being an important step in advancing hyperspectral imaging from UAS.

Highlights

Hyperspectral imaging from satellites is a well-established tool, which is used to collect data worldwide and over large areas, but the pixel size of hyperspectral missions such as PRISMA or DESIS is tens of meters (Coppo et al, 2020)
We characterized and calibrated the hyperspectral camera precisely, which allows for better planning of flight operations and to implement data processing algo rithms, tuned to the actual sensor’s capabilities
This information was key to adjust the exposure in field campaigns to avoid digital number (DN) above 1800, to adjust to the actual dy namic range

Summary

Introduction

Hyperspectral imaging from satellites is a well-established tool, which is used to collect data worldwide and over large areas, but the pixel size of hyperspectral missions such as PRISMA or DESIS is tens of meters (Coppo et al, 2020). This resolution is not high enough to investigate small inland water bodies such as narrow streams, or to study vegetation traits or function at the individual scale. As clouds are not transparent to radiation in the solar range, satellites or airplanes flying above the cloud base (typically around 2 km) cannot collect surface reflectance-factor data under overcast conditions.

Objectives

Methods

Results

Discussion

Conclusion