This thesis research examined the potential of using existing High-Resolution Land Cover (HRLC) maps as a low-cost source of reference Land Cover (LC) data and investigated methods for increasing their reliability. The aim was to address the challenge of reference data collection for global LC mapping that is particularly time and cost intensive compared to local or low-resolution LC mapping. The main outcome of the research is a methodology for combining existing HRLC maps to create a new HRLC dataset with high accuracy, named Map Of Land Cover Agreement (MOLCA). This dataset is created by intersecting existing HRLC maps, keeping only the areas where all maps agree and setting pixels to null where at least one map shows a different value. Accuracy assessment of MOLCA was done in one case study, and it showed a very accuracy of 96%, which is suitable for a reference dataset. The effectiveness of reference data obtained from MOLCA methodology was tested in one use case and two case studies, demonstrating that it has the potential to assist in HRLC production as the resulting dataset – MOLCA - can be used for training or validation. The use case and case studies were not done on a global level. However, an estimate based on these studies shows that MOLCA could provide around a trillion of reference LC samples worldwide in 28 days with the currently employed methodology. Obtaining a comparable number of samples through traditional methods, such as photo interpretation, within the same time frame would be impossible. The thesis also identifies some limitations of the approach and offers possible solutions to them.
La ricerca presentata nella tesi ha esaminato il potenziale dell'utilizzo di mappe di copertura del suolo ad alta risoluzione esistenti come fonte a basso costo di dati di copertura del suolo di riferimento e ha investigato metodi per aumentarne l'affidabilità. L'obiettivo era affrontare la sfida della raccolta dei dati di riferimento per la mappatura globale della copertura del suolo, che è particolarmente dispendiosa in termini di tempo e costi rispetto alla mappatura locale o a bassa risoluzione. Il principale risultato della ricerca è una metodologia per combinare le mappe esistenti di copertura del suolo ad alta risoluzione per creare un nuovo dato ad alta risoluzione e ad alta accuratezza, chiamato Map Of Land Cover Agreement (MOLCA). Questo dato viene creato intersecando le mappe esistenti di copertura del suolo ad alta risoluzione, mantenendo solo le aree in cui tutte le mappe sono d'accordo e impostando i pixel a nullo dove almeno una mappa mostra un valore diverso. La valutazione dell'accuratezza di MOLCA ha mostrato un'accuratezza molto elevata del 96%, adatta per un dato di riferimento. L'efficacia dei dati di riferimento ottenuti dalla metodologia MOLCA è stata testata in un caso d'uso e in due studi di caso, dimostrando che ha il potenziale per assistere nella produzione di dati HRLC poiché i dati risultanti - MOLCA - possono essere utilizzati per l'addestramento di algoritmi di apprendimento automatico o per la valutazione dell'accuratezza. Il caso d'uso e gli studi di caso non sono stati effettuati a livello globale. Tuttavia, una stima basata su questi studi mostra che MOLCA potrebbe fornire circa un trilione di campioni di copertura del suolo di riferimento in tutto il mondo in 28 giorni con la metodologia attualmente impiegata. Ottenere un numero comparabile di campioni attraverso metodi tradizionali, come foto-restituzione di immagini satellitari, nello stesso periodo di tempo sarebbe impossibile. La tesi identifica anche alcune limitazioni dell'approccio e offre possibili soluzioni per risolverle.
Reusing existing land cover maps to reinforce reference data for future land cover maps production
BRATIC, GORICA
2022/2023
Abstract
This thesis research examined the potential of using existing High-Resolution Land Cover (HRLC) maps as a low-cost source of reference Land Cover (LC) data and investigated methods for increasing their reliability. The aim was to address the challenge of reference data collection for global LC mapping that is particularly time and cost intensive compared to local or low-resolution LC mapping. The main outcome of the research is a methodology for combining existing HRLC maps to create a new HRLC dataset with high accuracy, named Map Of Land Cover Agreement (MOLCA). This dataset is created by intersecting existing HRLC maps, keeping only the areas where all maps agree and setting pixels to null where at least one map shows a different value. Accuracy assessment of MOLCA was done in one case study, and it showed a very accuracy of 96%, which is suitable for a reference dataset. The effectiveness of reference data obtained from MOLCA methodology was tested in one use case and two case studies, demonstrating that it has the potential to assist in HRLC production as the resulting dataset – MOLCA - can be used for training or validation. The use case and case studies were not done on a global level. However, an estimate based on these studies shows that MOLCA could provide around a trillion of reference LC samples worldwide in 28 days with the currently employed methodology. Obtaining a comparable number of samples through traditional methods, such as photo interpretation, within the same time frame would be impossible. The thesis also identifies some limitations of the approach and offers possible solutions to them.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD_thesis.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: This thesis research examined the potential of using existing High-Resolution Land Cover (HRLC) maps as a low-cost source of reference Land Cover (LC) data and investigated methods for increasing their reliability. The aim was to address the challenge of reference data collection for global LC mapping that is particularly time and cost intensive compared to local or low-resolution LC mapping. The main outcome of the research is a methodology for combining existing HRLC maps to create a new HRLC dataset with high accuracy, named Map Of Land Cover Agreement (MOLCA). This dataset is created by intersecting existing HRLC maps, keeping only the areas where all maps agree and setting pixels to null where at least one map shows a different value. Accuracy assessment of MOLCA was done in one case study, and it showed a very accuracy of 96%, which is suitable for a reference dataset. The effectiveness of reference data obtained from MOLCA methodology was tested in one use case and two case studies, demonstrating that it has the potential to assist in HRLC production as the resulting dataset – MOLCA - can be used for training or validation. The use case and case studies were not done on a global level. However, an estimate based on these studies shows that MOLCA could provide around a trillion of reference LC samples worldwide in 28 days with the currently employed methodology. Obtaining a comparable number of samples through traditional methods, such as photo interpretation, within the same time frame would be impossible. The thesis also identifies some limitations of the approach and offers possible solutions to them.
Dimensione
29.56 MB
Formato
Adobe PDF
|
29.56 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/198418