A study on "Paddy Field mapping in Nalgonda District India using multi-temporal Sentinel-2 Imagery in Google Earth Engine" was carried out during the Rabi season of 2019. Rabi is the winter crop season in India that typically occurs between October and May. Winter crop cultivation is critical to the Indian agricultural sector because it drives economic growth and ensures food security for the country’s population. Nonetheless, due to a variety of factors such as cloud cover, landscape heterogeneity, and small field sizes, monitoring these crops poses significant challenges. Theseobstacles contribute to the scarcity of optical satellite data, making it difficult to obtain accurate information on Winter crop growth and development. The main goal of this study aimed to develop a reliable and accurate methodology for mapping paddy fields in the Nalgonda district, India using Normalized Difference Vegetation Index (NDVI) data obtained from Sentinel-2 optical imagery. To achieve this, advanced machine learning algorithms, namely Random Forest (RF) and Support Vector Machine (SVM), were employed within the Google Earth Engine platform. The methodology involved measuring NDVI values from Sentinel-2 data and generating an NDVI mask to extract rice pixels while masking out other features such as water, forest, long-term crops like lemon/tangerine, and other dynamic features. The methodology was successfully validated in three villages, specifically Gurajala, Ramagiri, and Eaduloor, located in Shaligouraram and Kattangur Mandal of the Nalgonda district. To measure the NDVI values, ten fields were selected from the aforementioned villages, and collected ground reference points through field surveys. The classification of crops was performed on Sentinel-2 time series data using a Random Forest (RF) classifier and Support Vector Machine (SVM). The results indicated an overall accuracy of 97%, 90% and the kappa coefficient was found to be 0.96, 0.85. Additionally, validation is done by using the data split method 70% & 30%. RF was found to be more effective than SVM in terms of accuracy. Overall, this study successfully developed a methodology for mapping paddy fields in the Nalgonda district using multi-temporal S-2 imagery and ML algorithms.
Uno studio sulla "Mappatura dei campi di riso nel distretto di Nalgonda in India utilizzando l’immagine Sentinel-2 multi-temporale in Google Earth Engine" è stato condotto durante la stagione Rabi del 2019. La coltivazione di piante durante la stagione invernale è fondamentale per il settore agricolo indiano perché stimola la crescita economica e garantisce la sicurezza alimentare per la popolazione del paese. Tuttavia, a causa di diversi fattori come la copertura nuvolosa, l’eterogeneità del paesaggio e le dimensioni ridotte dei campi, monitorare queste colture rappresenta una sfida significativa. Questi ostacoli contribuiscono alla scarsità dei dati satellitari ottici, rendendo difficile ottenere informazioni accurate sulla crescita e lo sviluppo delle colture stagionali. L’obiettivo principale di questo studio era quello di sviluppare una metodologia affidabile e accurata per mappare i campi di riso nel distretto di Nalgonda, in India, utilizzando i dati dell’Indice di Vegetazione a Differenza Normalizzata (NDVI) ottenuti dalle immagini ottiche Sentinel-2. Per raggiungere questo obiettivo, sono stati impiegati algoritmi avanzati di apprendimento automatico, ovvero il Random Forest (RF) e il Support Vector Machine (SVM), all’interno della piattaforma Google Earth Engine. La metodologia prevedeva la misurazione dei valori NDVI dai dati Sentinel-2 e la generazione di una maschera NDVI per estrarre i pixel di riso, mascherando al contempo altre classi di copertura del suolo come l’acqua, la foresta, le colture a lungo termine come il limone/mandarino e altre caratteristiche dinamiche. La metodologia è stata validata con successo in tre villaggi, Gurajala, Ramagiri ed Eaduloor, situati nei mandali di Shaligouraram e Kattangur del distretto di Nalgonda. Per misurare i valori NDVI, sono stati selezionati dieci campi dai suddetti villaggi, e sono stati raccolti punti di riferimento sul campo attraverso rilevamenti sul campo. La classificazione delle colture è stata effettuata sui dati delle serie temporali di Sentinel-2 utilizzando un classificatore Random Forest (RF) e un Support Vector Machine (SVM). I risultati hanno indicato un’accuratezza complessiva del 97%, 90% e un coefficiente kappa di 0.96, 0.85. Inoltre, lo studio ha convalidato l’area del riso utilizzando il metodo di suddivisione dei dati 70% & 30%. RF è risultato essere più efficace di SVM in termini di accuratezza.
Paddy Field mapping in Nalgonda District, India using multi-temporal Sentinel-2 Imagery in Google Earth Engine
PUTTA, PRASHANTH REDDY
2021/2022
Abstract
A study on "Paddy Field mapping in Nalgonda District India using multi-temporal Sentinel-2 Imagery in Google Earth Engine" was carried out during the Rabi season of 2019. Rabi is the winter crop season in India that typically occurs between October and May. Winter crop cultivation is critical to the Indian agricultural sector because it drives economic growth and ensures food security for the country’s population. Nonetheless, due to a variety of factors such as cloud cover, landscape heterogeneity, and small field sizes, monitoring these crops poses significant challenges. Theseobstacles contribute to the scarcity of optical satellite data, making it difficult to obtain accurate information on Winter crop growth and development. The main goal of this study aimed to develop a reliable and accurate methodology for mapping paddy fields in the Nalgonda district, India using Normalized Difference Vegetation Index (NDVI) data obtained from Sentinel-2 optical imagery. To achieve this, advanced machine learning algorithms, namely Random Forest (RF) and Support Vector Machine (SVM), were employed within the Google Earth Engine platform. The methodology involved measuring NDVI values from Sentinel-2 data and generating an NDVI mask to extract rice pixels while masking out other features such as water, forest, long-term crops like lemon/tangerine, and other dynamic features. The methodology was successfully validated in three villages, specifically Gurajala, Ramagiri, and Eaduloor, located in Shaligouraram and Kattangur Mandal of the Nalgonda district. To measure the NDVI values, ten fields were selected from the aforementioned villages, and collected ground reference points through field surveys. The classification of crops was performed on Sentinel-2 time series data using a Random Forest (RF) classifier and Support Vector Machine (SVM). The results indicated an overall accuracy of 97%, 90% and the kappa coefficient was found to be 0.96, 0.85. Additionally, validation is done by using the data split method 70% & 30%. RF was found to be more effective than SVM in terms of accuracy. Overall, this study successfully developed a methodology for mapping paddy fields in the Nalgonda district using multi-temporal S-2 imagery and ML algorithms.È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: unitesi@unipv.it.
https://hdl.handle.net/20.500.14239/15544