Scholarlink

<p>Upwelling is the rising of cold, deep water to the surface, often triggered by wind patterns or shifts in thermal stratification. This study aims to develop an upwelling prediction model based on historical climate data in Lake Laut Tawar, Central Aceh, Aceh Province, to support the preparedness of floating net cage (KJA) farmers in facing upwelling risks. Upwelling due to climate change impacts mass fish mortality, so predictions are needed to provide early warning. This study design uses an observational approach with time series analysis, involving 2,556 observational data from six climate variables collected from NASA from January 1, 2017 to December 31, 2023. Given the enormous volume and high temporal resolution of the climate data collected over seven years, which constitutes a form of big data, these computational methods are particularly well-suited for capturing the complex, dynamic relationships inherent in the dataset. The analysis process includes modelling with Vector Autoregressive Moving Average (VARMA) and Vector Autoregressive (VAR) methods, as well as Support Vector Machine (SVM) based classification. The best model is selected based on the smallest Akaike Information Criterion (AIC) value and autocorrelation-freeness assumption. Model evaluation uses error metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Squared Error (MSE). The analysis results show that the Seasonal VARMA (1,0,2)(0,0,1)30 and VAR(6) models provide accurate predictions with the lowest AIC of -23.04 and − 8.43, respectively. At the same time, the SVM algorithm with Polynomial and RBF kernels produces an optimal F1-score of 0.985. This study concludes that integrating time series and machine learning methods in the prediction model can effectively mitigate the risk of upwelling. The developed model was implemented in a Streamlit-based interactive dashboard to facilitate monitoring and decision-making by KJA farmers. This study provides a proactive and preventive approach that can reduce economic losses due to the upwelling phenomenon.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integration of machine learning and time series analysis for upwelling prediction dashboard in lake Laut tawar, indonesia: A study based on climate forecasting

  • 期刊论文

  • 发表日期：

    2025年8月15日

查看全文
  • Muhammad Zia Ulhaq,
  • Muhammad Farid,
  • Zahra Ifma Aziza,
  • Teuku Muhammad Faiz Nuzullah,
  • Fakhrus Syakir,
  • Novi Reandy Sasmita

摘要

Upwelling is the rising of cold, deep water to the surface, often triggered by wind patterns or shifts in thermal stratification. This study aims to develop an upwelling prediction model based on historical climate data in Lake Laut Tawar, Central Aceh, Aceh Province, to support the preparedness of floating net cage (KJA) farmers in facing upwelling risks. Upwelling due to climate change impacts mass fish mortality, so predictions are needed to provide early warning. This study design uses an observational approach with time series analysis, involving 2,556 observational data from six climate variables collected from NASA from January 1, 2017 to December 31, 2023. Given the enormous volume and high temporal resolution of the climate data collected over seven years, which constitutes a form of big data, these computational methods are particularly well-suited for capturing the complex, dynamic relationships inherent in the dataset. The analysis process includes modelling with Vector Autoregressive Moving Average (VARMA) and Vector Autoregressive (VAR) methods, as well as Support Vector Machine (SVM) based classification. The best model is selected based on the smallest Akaike Information Criterion (AIC) value and autocorrelation-freeness assumption. Model evaluation uses error metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Squared Error (MSE). The analysis results show that the Seasonal VARMA (1,0,2)(0,0,1)30 and VAR(6) models provide accurate predictions with the lowest AIC of -23.04 and − 8.43, respectively. At the same time, the SVM algorithm with Polynomial and RBF kernels produces an optimal F1-score of 0.985. This study concludes that integrating time series and machine learning methods in the prediction model can effectively mitigate the risk of upwelling. The developed model was implemented in a Streamlit-based interactive dashboard to facilitate monitoring and decision-making by KJA farmers. This study provides a proactive and preventive approach that can reduce economic losses due to the upwelling phenomenon.