Paddy field classification with MODIS-terra multi-temporal image transformation using phenological approach in Java Island

• Published in: International journal of electrical and computer engineering (Malacca, Malacca) Vol. 9; no. 2; p. 1346
• Main Authors: Dimyati, Muhammad, Kustiyo, Kustiyo, Dimyati, Ratih Dewanti
• Format: Journal Article
• Language: English
• Published: Yogyakarta IAES Institute of Advanced Engineering and Science 01.04.2019
• Subjects: Classification; Correlation analysis; Elevation; Image classification; MODIS; Reflectance; Indonesia; Java (island)
• ISSN: 2088-8708; 2088-8708
• DOI: 10.11591/ijece.v9i2.pp1346-1358

This paper presents the paddy field classification model using the approach based on periodic plant life cycle events and how these elevations in climate as well as habitat factors, such as elevation. The data used are MODIS-Terra two tiles of H28v09 and H29v09 of 2016, consist of 46 series of 8-daily data, with 500 meter resolution in Java region. The paddy field classification method based on the phenological model is done by Maximum Likelihood on the transformed annual multi-temporal image of the reflectance data, index data, and the combination of reflectance and index data. The results of the study showed that, with the reference of the Paddy Field Map from the Ministry of Agriculture (MoA), the overall accuracies of the paddy field classification results using the combination of reflectance and index data provide the highest (85.4%) among the reflectance data (83.5%) and index data (81.7%). The accuracy levels were varied; these depend on the slope and the types of paddy fields. Paddy fields on the slopes of 0-2% could be well identified by MODIS-Terra data, whereas it was difficult to identify the paddy fields on the slope >2%. Rain-fed lowland paddy field type has a lower user accuracy than irrigated paddy fields. This study also performed correlation (r2) between the analysis results and the statistical data based on district and provincial boundaries were >0.85 and >0.99 respectively. These correlations were much higher than the previous study results, which reached 0.49-0.65 (hilly-flat areas of county-level), and 0.80-0.88 (hilly-flat areas of provincial level) for China, and reached 0.44 for Indonesia.