Data Reconstruction Of Sea Surface Temperature In Fish Management Area 713 (FMA-713) in Indonesia Waters by Using Machine Learning Susanna Nurdjaman, Rheno A Wicaksono, Aditya R Kartadikaria, Muhammad Rais Abdullah
Faculty of Earth Science and Technology, Bandung Institute of Technology
Jalan Ganesha 10, Bandung 40132, Indonesia
rheno.armand07[at]gmail.com
Abstract
The FMA-713 in Indonesia is water that has dynamic of temperature changes due to interactions with the Pacific Ocean and cause upwelling at several points. Sea surface temperature data can be obtained by measuring with satellite imagery. However, there are drawbacks to measurements using satellite imagery, namely missing data when measuring sea surface temperature due to cloud cover. In this study, a machine learning method was used to reconstruct sea surface temperature data using a backpropagation neural network algorithm. The data used in this research is data captured with MODIS Satellite with spatial resolution of 4,63 km from 2003 to 2021. the data are separated into test data (4 years) and training data with the variant length of 5 years, 10 years and 15 years. Reconstruction of empty data was carried out by means of single-step prediction and an RMSE value of 0.7 oC was obtained. After reconstructing the blank data, the reconstructed data is used as input for reconstructing the occurrence of sea surface temperature data. This program is made with 4 scenarios, these scenarios are scenario 1 (reconstructed data is empty data), scenario 2 (empty data is filled with zero values), and scenario 3 (empty data is filled with average values) and Scenario 4 (empty data is filled with a value of 28). Accurate results were obtained in reconstructing sea surface temperature where the 4 scenarios had a correlation value of r = 0.96. The results show that scenarios 3 and 4 are the most accurate scenarios compared to the other 2 scenarios with RMSE values ranging from 0.7 oC (scenarios 3 and 4), 12.76 oC (scenario 2), while scenario 1 resulting in the most poor performance compared to the other scenarios.
Keywords: The FMA-713, Sea Surface Temperature, Machine learning, Backpropagation Neural Network, data reconstruction
