This thesis describes some of the scientific research projects I took part in during my doctoral studies regarding the applications of the electromagnetic spectrum in precision agriculture. Four complementary research lines are presented in this manuscript: (i) the assessment of soil texture distribution through gamma spectrometric measurements, (ii) the study of metal uptake in tomato crops with the subsequent production of a Cadmium content prediction map, (iii) the identification of Flavescence Dorée symptoms through RGB imagery analysed using vegetation indices, and (iv) the monitoring of vineyard water stress levels via infrared data. After a first introductory part on the electromagnetic spectrum and gamma ray spectroscopy, the latter is utilized to study soil texture in the Mezzano Lowland (Italy). Since different soil textures adsorb cations in different quantities, the abundances of K and Th radioelements can be linked to the clay and sand soil fractions through a correlation study. A Machine Learning algorithm is employed to enhance the correlation study and the derived relations between soil texture and radioelements abundances are used to produce an updated and spatially enhanced version of the soil texture map provided by the Emilia-Romagna regional administration. This map highlights traces of abandoned riverbeds crossing the study area until VII century CE, confirming the accuracy of the methodology. The Mezzano Lowland is further analysed for tomato crop metal uptake during the various steps of the production chain. Current regulations impose strict limits on metal content in food products, especially in baby food. This pushes food companies to monitor the metal concentration in their production chain to analyse its most critical steps. Through the analysis of soil, plants, berries, fertilizers and finished product samples, the criticality of Aluminium (Al), Arsenic (As), Cadmium (Cd) and Mercury (Hg) is assessed. The only potentially critical element ends up being Cd, for which has been developed a predictive model resulting in a Cd content prediction map delivered to the local company this study has been made in collaboration with. The electromagnetic spectrum is furthermore exploited by utilizing RGB images to locate and identify symptoms of Flavescence Dorée in vineyards. The developed methodology consists of airborne photogrammetric surveys (performed over a vineyard near Forlì, Italy), the orthorectification of survey images and a subsequent orthomosaic production, a vegetation indices analysis, a spatial convolution filter analysis and a GIS environment analysis. The processing ends with the production of Instance, Density and Incidence maps to aid farmers plan plant uprooting to prevent disease spread. The methodology has been validated via True Positives, True Negatives, False Positives and False Negatives values and specificity, sensitivity, Positive Predicted Value and Negative Predicted Value metrics which show promising results. Plants, due to their transpiration rate being tied to canopy temperature through stomatal opening of leaves, send signals of their water stress status in the infrared region of the spectrum. The reflectance of leaves in this frequency range is converted into Crop Water Stress Index values after a canopy-soil pixel discrimination. The CWSI map of the studied area, a vineyard near Serra De’ Conti (Italy), is produced for the months of June, July and August highlighting the evolution of water stress status of the cultivars, together with peculiar local soil conditions and plant health status due to ongoing diseases. The methodology presented acts therefore as a valuable monitoring tool to enhance yield quality and reduce water wastes.
Questa tesi descrive alcuni dei progetti di ricerca scientifica a cui ho partecipato durante i miei studi di dottorato riguardanti le applicazioni dello spettro elettromagnetico nell'agricoltura di precisione. In questo manoscritto sono presentate quattro linee di ricerca complementari: (i) la valutazione della distribuzione della tessitura del suolo tramite misurazioni di spettroscopia gamma, (ii) lo studio dell’assorbimento di metalli nei pomodori con successiva produzione di una mappa di previsione del contenuto di Cadmio, (iii) l’identificazione dei sintomi di Flavescenza Dorata tramite immagini RGB e indici di vegetazione, e (iv) il monitoraggio dello stress idrico nei vigneti tramite dati a infrarossi. Dopo un’introduzione sullo spettro elettromagnetico e la spettroscopia gamma, quest'ultima è usata per studiare la tessitura del suolo nella Bassa del Mezzano (Italia). Dato che le diverse tessiture adsorbono cationi in quantità variabili, le abbondanze di K e Th sono correlate alle frazioni di argilla e sabbia tramite uno studio di correlazione. Un algoritmo di Machine Learning è impiegato per migliorare lo studio di correlazione, e le relazioni derivate producono una mappa aggiornata della tessitura fornita dalla regione Emilia-Romagna, che evidenzia antichi letti fluviali presenti fino al VII secolo d.C., confermando l'accuratezza della metodologia. Nella Bassa del Mezzano viene inoltre analizzato l’assorbimento di metalli nei pomodori durante le fasi della produzione. Le normative vigenti impongono limiti sul contenuto di metalli negli alimenti, specie per il baby food, spingendo le aziende a monitorare la concentrazione di metalli per analizzare le fasi più critiche della produzione. Attraverso l'analisi di campioni di suolo, piante, bacche, fertilizzanti e prodotto finito, viene valutata la criticità di Alluminio (Al), Arsenico (As), Cadmio (Cd) e Mercurio (Hg). L’unico elemento potenzialmente critico è il Cd, per cui è stato sviluppato un modello predittivo che ha portato alla creazione di una mappa di previsione della sua concentrazione consegnata alla ditta locale partner di progetto. Lo spettro elettromagnetico viene sfruttato inoltre tramite immagini RGB per identificare sintomi di Flavescenza Dorata nei vigneti. La metodologia comprende i rilievi fotogrammetrici aerei (eseguiti su un vigneto vicino Forlì, Italia), l’ortorettifica delle immagini rilevate, la produzione di un ortomosaico, l’analisi degli indici di vegetazione, un filtro di convoluzione spaziale e un’analisi in ambiente GIS. L’elaborazione termina con la produzione di mappe di Istanza, Densità e Incidenza per aiutare i coltivatori a pianificare l’estirpazione delle piante per prevenire la diffusione della malattia. La metodologia è stata validata con valori di Veri Positivi, Veri Negativi, Falsi Positivi e Falsi Negativi e metriche di specificità, sensibilità, Valore Predittivo Positivo e Negativo, con risultati promettenti. Le piante, poiché il tasso di traspirazione è legato alla temperatura della chioma tramite apertura stomatica, inviano segnali di stress idrico nella regione infrarossa. La riflettanza delle foglie in questo intervallo di frequenze è convertita in valori di Crop Water Stress Index (CWSI) dopo una discriminazione tra pixel associati a chioma e a suolo. La mappa CWSI dell’area di studio, un vigneto vicino Serra De’ Conti (Italia), è prodotta per i mesi di giugno, luglio e agosto evidenziando l’evoluzione dello stress idrico delle colture, insieme a particolari condizioni locali del suolo e lo stato di salute delle piante dovuto a malattie in corso. La metodologia presentata agisce quindi come uno strumento di monitoraggio prezioso per migliorare la qualità della resa e ridurre gli sprechi d'acqua.
Harnessing sensing technologies for a smarter agriculture
MAINO, Andrea
2025
Abstract
This thesis describes some of the scientific research projects I took part in during my doctoral studies regarding the applications of the electromagnetic spectrum in precision agriculture. Four complementary research lines are presented in this manuscript: (i) the assessment of soil texture distribution through gamma spectrometric measurements, (ii) the study of metal uptake in tomato crops with the subsequent production of a Cadmium content prediction map, (iii) the identification of Flavescence Dorée symptoms through RGB imagery analysed using vegetation indices, and (iv) the monitoring of vineyard water stress levels via infrared data. After a first introductory part on the electromagnetic spectrum and gamma ray spectroscopy, the latter is utilized to study soil texture in the Mezzano Lowland (Italy). Since different soil textures adsorb cations in different quantities, the abundances of K and Th radioelements can be linked to the clay and sand soil fractions through a correlation study. A Machine Learning algorithm is employed to enhance the correlation study and the derived relations between soil texture and radioelements abundances are used to produce an updated and spatially enhanced version of the soil texture map provided by the Emilia-Romagna regional administration. This map highlights traces of abandoned riverbeds crossing the study area until VII century CE, confirming the accuracy of the methodology. The Mezzano Lowland is further analysed for tomato crop metal uptake during the various steps of the production chain. Current regulations impose strict limits on metal content in food products, especially in baby food. This pushes food companies to monitor the metal concentration in their production chain to analyse its most critical steps. Through the analysis of soil, plants, berries, fertilizers and finished product samples, the criticality of Aluminium (Al), Arsenic (As), Cadmium (Cd) and Mercury (Hg) is assessed. The only potentially critical element ends up being Cd, for which has been developed a predictive model resulting in a Cd content prediction map delivered to the local company this study has been made in collaboration with. The electromagnetic spectrum is furthermore exploited by utilizing RGB images to locate and identify symptoms of Flavescence Dorée in vineyards. The developed methodology consists of airborne photogrammetric surveys (performed over a vineyard near Forlì, Italy), the orthorectification of survey images and a subsequent orthomosaic production, a vegetation indices analysis, a spatial convolution filter analysis and a GIS environment analysis. The processing ends with the production of Instance, Density and Incidence maps to aid farmers plan plant uprooting to prevent disease spread. The methodology has been validated via True Positives, True Negatives, False Positives and False Negatives values and specificity, sensitivity, Positive Predicted Value and Negative Predicted Value metrics which show promising results. Plants, due to their transpiration rate being tied to canopy temperature through stomatal opening of leaves, send signals of their water stress status in the infrared region of the spectrum. The reflectance of leaves in this frequency range is converted into Crop Water Stress Index values after a canopy-soil pixel discrimination. The CWSI map of the studied area, a vineyard near Serra De’ Conti (Italy), is produced for the months of June, July and August highlighting the evolution of water stress status of the cultivars, together with peculiar local soil conditions and plant health status due to ongoing diseases. The methodology presented acts therefore as a valuable monitoring tool to enhance yield quality and reduce water wastes.File | Dimensione | Formato | |
---|---|---|---|
PhD_Thesis_Andrea_Maino.pdf
accesso aperto
Dimensione
15.54 MB
Formato
Adobe PDF
|
15.54 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/220150
URN:NBN:IT:UNIFE-220150