THE CC-UNET MODEL OF HEAVY RAIN DETECTION IN MEDAN CITY FOR ANTICIPATING AQUACULTURE PRODUCTION LOSS
Harmful weather has a strong correlation with losses in aquaculture enterprises due to moderate to extreme floods. Such destruction greatly affects aquaculture productivity, lowering total production, production value, income, and Gross Regional Domestic Product. Convective clouds are highly correlated to rain. However, cloud detection as natural disaster mitigation is a challenging task to work on at times. Therefore, artificial intelligence as part of computer science is needed to help simplify the complexity of natural disaster mitigation. By employing secondary data on regional development parameters 2010-21 obtained from the Central Statistics Agency of Medan City 2022, such as Gross Regional Domestic Product (GRDP), Poverty Rate (POV), Human Development Index (HDI), and Aquaculture Production (AP), the present study desired to find the causality relationship between those variables through the Autoregressive Distributed Lag (ARDL) Model on EViews 9.0. Besides, in solving the complexities of cloud detection, this study used a predictive-analytic classification of the CC-Unet model based on deep learning to improve the classification results of convective clouds by taking the advantages of Himawari 8 satellite image data collected on 13 May 2021 and 30 October 2021. The present study found that if the local government in Medan City could increase GRDP and aquaculture production by 1% while also protecting the aquaculture sector from floods, poverty could be reduced by 0.094561% and 0.817079% respectively. In terms of cloud detection, the CC-Unet model in this study has a better accuracy of 97.29% compared to the U-Net model (94.17%). Shrimp farming businesses in Medan City, especially Aur Village, to prevent losses against their shrimp production, can use the prediction of flooding. Consequently, poverty could be eliminated.
Convective cloud, CC-Unet, disaster mitigation, poverty.