%0 Conference Paper
%A Krieger, Vera
%A Matveeva, Maria
%A Rascher, Uwe
%T Systematic Assessment Of Airborne Sun-Induced Fluorescence Maps By The Application Of Quality Criteria
%I Rheinische Friedrich-Wilhelms-Universität Bonn
%M FZJ-2020-00156
%D 2019
%X When plants  absorb light,  not  all energy is  converted by photosynthesis, but  excess  energy is released as heat or emitted as sun-induced chlorophyll fluorescence (F). This signal, related to the photosynthetic  efficiency  of  plants,  has  been  intensively  studied  and  measured  from  ground, airborne and satellite. However, retrieving sun-induced fluorescence (F) from remote sensing data is challenging because accurate modeling of atmospheric influences is required.. The advent of the airborne imaging spectrometer HyPlant made possible to produce F maps in high-spatial resolution (1-3 meters), which is a valuable tool to better understand F at relevant ecosystem scale. Currently, two  different  algorithms  are  used  routinely  to  retrieve  red  and  far-red  F  from  HyPlant.  Both methods are based on the O 2  absorption bands, but while iFLD method employs a semi-empirical atmospheric correction (i.e., bare-soils), the SFM makes use of a physically-based atmospheric modeling (MODTRAN5 code). A common method of testing the reliability of a remotely sensed F product (in this study airborne F maps) is the comparison with “ground truth” data where the atmosphere can be neglected. In this work we tested another possibility of assessing the quality of the airborne F maps, which does not require ground reference measurements. For this purpose we have developed so-called ’quality criteria’, which should help to find errors and artefacts that have arisen during F retrieval. This method was used to test the quality of the airborne F maps of 2016 campaign.  By applying the quality criteria, clear differences in the performance of two retrievals were found. Although it was shown that both retrievals performed well in F 760  retrieval, even at places with changes from vegetated to non-vegetated sites on pixel scale, iFLD was more robust for retrieving correct absolute values for F 760  and F 687 , while SFM performed less accurate in this term, over- and  underestimating  F  values.  Furthermore,  previously  reported  problems  with  image  pre-processing (deconvolution for correcting PSF) of SFM became clear here. This was causing strong artefacts in F 687  retrievals from SFM. However, SFM proved to be the more suitable method for identifying small differences on pixel scale. Moreover, this algorithm did not show systematic variations over entire flight lines as observed by the use of iFLD. The physically-based approach of atmospheric correction used with SFM thus provided more interference-free F maps than the semi-empirical correction using non-fluorescent surfaces as in iFLD retrieval. Testing F retrievals on vegetation under different illumination conditions showed the necessity to calculate F yield  for quantification of photosynthesis rates. The application of the proposed quality features proved to be a valuable tool for assessing the performance of F retrieval on airborne maps. Therefore we propose to use the quality criteria even when sufficient ground references are available, because even if the quality criteria do not replace ground-truth data, they provide important additional information about the quality of the F product of the respective retrieval method.
%B EARSel SIG Imaging Spectroscopy Workshop
%C 6 Feb 2019 - 8 Feb 2019, Brno (Czech)
Y2 6 Feb 2019 - 8 Feb 2019
M2 Brno, Czech
%F PUB:(DE-HGF)6
%9 Conference Presentation
%U https://juser.fz-juelich.de/record/872665