Abstract<\/h3>\n\nThe increasing demand for food and the impact of climate change underscore the need for intensified food production processes to continually address the growing population’s requirements. A critical aspect of food planning involves the identification of cropping frequency, serving as a key strategy for enhancing food production. Remote sensing plays a pivotal role in capturing cropping frequency information by analyzing phenological characteristics recorded in band transformations. Furthermore, the integration of machine learning allows for the categorization of patterns derived from index responses, eliminating the need to individually detect each phenological phase. The study aims to assess the accuracy of multi-sensor data fusion using the STARFM algorithm and machine learning to produce dense time-series images combining Landsat-8 and MODIS downscaled imagery for mapping paddy’s phenology and identifying paddy cropping frequencies. The phenology identification results demonstrate an accuracy range of 3-4 months for the Landsat dataset and less than 1 (one) month for the dataset resulting from its fusion with MODIS. Concurrently, the cropping frequency identification reveals an accuracy of 60%, 42.5%, 95%, 85%, and 100%, respectively, for Landsat phenology, fusion phenology, Landsat Decision Tree, fusion Decision Tree, and Random Forest for both datasets. This underscores the profound impact of data availability and quality on the accuracy of the obtained results. Dense time-series remote sensing data can be used for mapping cropping frequency to indicate the productive paddy areas which should be protected to ensure food security in the future. \u00a9 2024, Asociatia Geographia Technica. All rights reserved.<\/span><\/p>\n <\/p>\n<\/div>\n
The increasing demand for food and the impact of climate change underscore the need for intensified food production processes to continually address the growing population’s requirements. A critical aspect of food planning involves the identification of cropping frequency, serving as a key strategy for enhancing food production. Remote sensing plays a pivotal role in capturing cropping frequency information by analyzing phenological characteristics recorded in band transformations. Furthermore, the integration of machine learning allows for the categorization of patterns derived from index responses, eliminating the need to individually detect each phenological phase. The study aims to assess the accuracy of multi-sensor data fusion using the STARFM algorithm and machine learning to produce dense time-series images combining Landsat-8 and MODIS downscaled imagery for mapping paddy’s phenology and identifying paddy cropping frequencies. The phenology identification results demonstrate an accuracy range of 3-4 months for the Landsat dataset and less than 1 (one) month for the dataset resulting from its fusion with MODIS. Concurrently, the cropping frequency identification reveals an accuracy of 60%, 42.5%, 95%, 85%, and 100%, respectively, for Landsat phenology, fusion phenology, Landsat Decision Tree, fusion Decision Tree, and Random Forest for both datasets. This underscores the profound impact of data availability and quality on the accuracy of the obtained results. Dense time-series remote sensing data can be used for mapping cropping frequency to indicate the productive paddy areas which should be protected to ensure food security in the future. \u00a9 2024, Asociatia Geographia Technica. All rights reserved.<\/span><\/p>\n <\/p>\n<\/div>\n
![\"\"](\"https:\/\/geo.ugm.ac.id\/wp-content\/uploads\/sites\/13\/2017\/10\/Untitled-1.png\")
<\/a><\/p>","protected":false},"excerpt":{"rendered":"