The PhD research aimed to develop novel strategies able to better retrieve and interpret the chlorophyll Solar-Induced Fluorescence (SIF) signal emitted by terrestrial vegetation and inland waters at ground level, to advance the understanding of ecosystems structure and functioning. SIF metrics were defined taking advantage of the full SIF spectrum available from the recently developed “spectrum-fitting” algorithm (SpecFit). The metrics were designed to characterize the SIF spectrum, in terms of red and far-red peaks maximum values (SIFred, SIFfar-red), corresponding wavelengths and the spectrally integrated value (SIFINT). SIF typically evaluated in the O2-A (SIF760) and O2-B (SIF687) bands and reflectance indices (used as proxies for canopy biophysical parameters) have been compared to the SIF spectrum. The reflectance indices selected are the NDVIred-edge, CIred-edge, NIRv and PRI. The analysis has been carried out at seasonal/diurnal scales, exploiting top-of canopy (TOC) spectral measurements acquired over three crops. The SIF evaluated at the peaks always show a strong correlation with the corresponding O2 bands values, while the SIFINT represents a more complete parameter and shows peculiar dynamics. At diurnal scale, the combined use of reflectance indices and TOC SIF metrics allows to gain a better knowledge of the crops dynamics. Seasonally, the SIF and reflectance indices show more similar temporal evolution along the growth-phases because they are mainly driven by changes in the overall canopy biomass, chlorophyll content and incident light. The reabsorption of the SIF within the canopy-leaf system affects the overall SIF spectral shape and magnitude at this temporal scale. As demonstrated on the synthetic dataset, the reabsorption effect prevents an accurate evaluation of the fluorescence quantum yield (SIFyield). Correcting the TOC SIF spectrum for the reabsorption is pivotal. In this regard, two different approaches have been developed and tested. The parametric method enables to correct SIF for the reabsorption (SIFRC) establishing parametric relationships with spectral variables routinely measured at TOC. The method accuracy depends on the plant growth phase, showing better results for medium-dense canopies. This behavior compromises the application of the method on the full seasonal analysis. The second approach based on Fourier-Machine Learning algorithm retrieves the SIFRC, and biophysical parameters (LAI, Cab, SIFyield, aPAR) with a better accuracy for all the conditions. The two approaches have been compared by considering synthetic simulations and real field measurements. Two methods were developed and tested starting from different assumptions: the parametric method can be used in a simpler way but it lacks accuracy for sparse conditions; while the Fourier-Machine Learning algorithm is more complex but offer better results. Regarding clear lake waters, a novel version of the Fluorescence Line Height approach has been implemented. The SIF proxy obtained agree with the temporal evolution of other conventional spectral indices (EPAR, R550 and [Chl-a]). Novel phytoplankton primary production models have been defined and tested adapting the vegetation Light Use Efficiency model for inland waters. Promising results have been obtained when the SIFFLH and a novel photosynthesis efficiency proxy here introduced are considered. In conclusion, the results obtained highlight the relevance to retrieve the SIF spectrum and the importance to employ SIF reabsorption correction methods to obtain relevant parameters better related with terrestrial vegetation functioning and less affected from canopy structure. This study has demonstrated that the hyperspectral and frequency measurements allow to follow the phytoplankton dynamics, particularly in clear sky days. Furthermore, the use of parameters linked to the SIF represents a promising approach for monitoring the phytoplankton primary production in lakes.

La seguente ricerca di dottorato ha come obiettivo lo sviluppo di nuove strategie in grado di migliorare la stima e l’interpretazione del segnale di fluorescenza indotto dalla luce solare (SIF) emesso dalla clorofilla. In particolare la SIF emessa a livello del suolo è stata utilizzata per migliorare la comprensione del funzionamento della vegetazione terrestre ed acque interne. Nuove metriche di fluorescenza sono state definite a partire dello spettro completo di SIF ottenuto tramite l’algoritmo SpecFit applicato a misure spettrali di campo Top-Of-Canopy. Tali metriche sono state concepite per caratterizzare lo spettro totale di SIF, in termini di valori massimi di emissione (SIFred, SIFfar-red), posizione dei picchi ed integrale di fluorescenza (SIFINT). Il loro comportamento su scale stagionale e giornaliera è stato dunque confrontato con la SIF valutata nelle bande di assorbimento dell’ossigeno atmosferico (SIF760, SIF687) ed indici di riflettanza (NDVIred-edge, CIred-edge, NIRv, PRI), questi ultimi usati come proxy di parametri biofisici di vegetazione. Il SIF valutato ai picchi mostra sempre una forte correlazione con i corrispondenti valori delle bande O2, mentre il SIFINT rappresenta un parametro più completo e mostra dinamiche peculiari. A scala diurna, l'uso combinato di indici di riflettanza e metriche SIF permette una migliore caratterizzazione delle dinamiche delle diverse colture investigate. Stagionalmente, gli indici di SIF e di riflettanza sono caratterizzati dalla medesima evoluzione temporale in quanto entrambi guidati da cambiamenti della luce incidente e nella biomassa della canopy, del suo contenuto di clorofilla. Tuttavia, a tale scala, il riassorbimento della fluorescenza che avviene a livello fogliare e di chioma affligge la forma spettrale e intensità della SIF. In accordo con l’analisi condotta su misure spettrali simulate, il riassorbimento impedisce la corretta stima del rendimento di fluorescenza (SIFyield): pertanto, correggere lo spettro SIF per questo processo è cruciale. Con questo obiettivo, sono stati sviluppati due possibili approcci. Il metodo parametrico consente di correggere la SIF per il riassorbimento sfruttando relazioni parametriche con variabili spettrali misurate a livello TOC. Tuttavia, l’accuratezza di questo metodo dipende dallo stadio di sviluppo della pianta: migliori risultati sono ottenuti per chiome di media densità. Tale comportamento preclude l’utilizzo del metodo a scale stagionale. Il secondo approccio accoppia l’analisi della trasformata di Fourier con tecniche di Machine Learning. In questo caso, sia lo spettro SIF corretto che parametri biofisici quali LAI, Cab, SIFyield e aPAR, sono calcolati con maggiore accuratezza, indipendentemente dalla vegetazione considerata. In generale, il metodo parametrico può essere utilizzato in modo più semplice, ma manca di accuratezza quando applicato a vegetazioni rade, mentre l'algoritmo Fourier-ML è più complesso, ma offre risultati migliori. Per quanto concerne le acque interne, l’algoritmo Fluorescence Line Height è stato modificato per poter essere applicato a tali ambienti sfruttando misure iperspettrali di campo. L’evoluzione temporale del proxy di SIF così ottenuto concorda con l’andamento osservato per indici spettrali convenzionali (EPAR, R550, [Chl-a]), specialmente nei giorni caratterizzati da cielo sereno. Nuovi modelli di produzione primaria del fitoplancton sono stati inoltre definiti adattando per le acque interne il modello LUE (Light Use Efficiency) sviluppato per la vegetazione. Risultati promettenti sono stati ottenuti quando la SIFFLH e un nuovo proxy di efficienza di fotosintesi sono contemplati nel modello.

(2022). Solar-induced chlorophyll fluorescence signal retrieval in terrestrial vegetation and inland waters from hyperspectral proximal sensing. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2022).

Solar-induced chlorophyll fluorescence signal retrieval in terrestrial vegetation and inland waters from hyperspectral proximal sensing

CESANA, ILARIA
2022

Abstract

The PhD research aimed to develop novel strategies able to better retrieve and interpret the chlorophyll Solar-Induced Fluorescence (SIF) signal emitted by terrestrial vegetation and inland waters at ground level, to advance the understanding of ecosystems structure and functioning. SIF metrics were defined taking advantage of the full SIF spectrum available from the recently developed “spectrum-fitting” algorithm (SpecFit). The metrics were designed to characterize the SIF spectrum, in terms of red and far-red peaks maximum values (SIFred, SIFfar-red), corresponding wavelengths and the spectrally integrated value (SIFINT). SIF typically evaluated in the O2-A (SIF760) and O2-B (SIF687) bands and reflectance indices (used as proxies for canopy biophysical parameters) have been compared to the SIF spectrum. The reflectance indices selected are the NDVIred-edge, CIred-edge, NIRv and PRI. The analysis has been carried out at seasonal/diurnal scales, exploiting top-of canopy (TOC) spectral measurements acquired over three crops. The SIF evaluated at the peaks always show a strong correlation with the corresponding O2 bands values, while the SIFINT represents a more complete parameter and shows peculiar dynamics. At diurnal scale, the combined use of reflectance indices and TOC SIF metrics allows to gain a better knowledge of the crops dynamics. Seasonally, the SIF and reflectance indices show more similar temporal evolution along the growth-phases because they are mainly driven by changes in the overall canopy biomass, chlorophyll content and incident light. The reabsorption of the SIF within the canopy-leaf system affects the overall SIF spectral shape and magnitude at this temporal scale. As demonstrated on the synthetic dataset, the reabsorption effect prevents an accurate evaluation of the fluorescence quantum yield (SIFyield). Correcting the TOC SIF spectrum for the reabsorption is pivotal. In this regard, two different approaches have been developed and tested. The parametric method enables to correct SIF for the reabsorption (SIFRC) establishing parametric relationships with spectral variables routinely measured at TOC. The method accuracy depends on the plant growth phase, showing better results for medium-dense canopies. This behavior compromises the application of the method on the full seasonal analysis. The second approach based on Fourier-Machine Learning algorithm retrieves the SIFRC, and biophysical parameters (LAI, Cab, SIFyield, aPAR) with a better accuracy for all the conditions. The two approaches have been compared by considering synthetic simulations and real field measurements. Two methods were developed and tested starting from different assumptions: the parametric method can be used in a simpler way but it lacks accuracy for sparse conditions; while the Fourier-Machine Learning algorithm is more complex but offer better results. Regarding clear lake waters, a novel version of the Fluorescence Line Height approach has been implemented. The SIF proxy obtained agree with the temporal evolution of other conventional spectral indices (EPAR, R550 and [Chl-a]). Novel phytoplankton primary production models have been defined and tested adapting the vegetation Light Use Efficiency model for inland waters. Promising results have been obtained when the SIFFLH and a novel photosynthesis efficiency proxy here introduced are considered. In conclusion, the results obtained highlight the relevance to retrieve the SIF spectrum and the importance to employ SIF reabsorption correction methods to obtain relevant parameters better related with terrestrial vegetation functioning and less affected from canopy structure. This study has demonstrated that the hyperspectral and frequency measurements allow to follow the phytoplankton dynamics, particularly in clear sky days. Furthermore, the use of parameters linked to the SIF represents a promising approach for monitoring the phytoplankton primary production in lakes.
COGLIATI, SERGIO
FINIZIO, ANTONIO
SIF; iperspettrale; acque interne; fitoplancton; produttività
SIF; hyperspectral; inland waters; phytoplakton; produttività
GEO/12 - OCEANOGRAFIA E FISICA DELL'ATMOSFERA
English
27-gen-2022
SCIENZE CHIMICHE, GEOLOGICHE E AMBIENTALI
34
2020/2021
open
(2022). Solar-induced chlorophyll fluorescence signal retrieval in terrestrial vegetation and inland waters from hyperspectral proximal sensing. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2022).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_727177.pdf

accesso aperto

Descrizione: Tesi
Tipologia di allegato: Doctoral thesis
Dimensione 16.38 MB
Formato Adobe PDF
16.38 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/366236
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact