Prediksi Polutan Udara PM10 Menggunakan Model Seasonal Trend Decomposition Procedure Based On LOESS (STL) Dan SARIMA di Kota Surabaya

Yanuar, Mohammad Hammam Andare (2025) Prediksi Polutan Udara PM10 Menggunakan Model Seasonal Trend Decomposition Procedure Based On LOESS (STL) Dan SARIMA di Kota Surabaya. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5014211104-Ungraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Kualitas udara di kota-kota besar seperti Surabaya terus mengalami penurunan seiring meningkatnya aktivitas industri, transportasi, dan pembangunan infrastruktur. Partikulat udara berukuran ≤10 mikron (PM10) menjadi salah satu polutan utama yang berdampak signifikan terhadap kesehatan masyarakat. Untuk mendukung pengendalian pencemaran, penelitian ini bertujuan memprediksi konsentrasi PM10 di Kota Surabaya menggunakan pendekatan deret waktu, yaitu Seasonal Trend Decomposition Procedure based on Loess (STL) dan Seasonal Autoregressive Integrated Moving Average (SARIMA), berdasarkan data harian dari SPKU 6 Wonorejo periode Januari hingga Juli 2024. Model STL digunakan untuk mendekomposisi data menjadi komponen tren, musiman, dan residual, sedangkan SARIMA diterapkan untuk memodelkan pola musiman dan non-musiman. Evaluasi dilakukan menggunakan Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), dan Mean Absolute Error (MAE). Hasil kesesuian evaluasi keakuratan membuktikan bahwa STL secara keseluruhan memiliki tingkat keakuratan yang lebih baik daripada SARIMA. Untuk meningkatkan ketepatan hasil, dilakukan proses kalibrasi antara prediksi dan data aktual. Hasil menunjukkan bahwa setelah kalibrasi, STL menghasilkan MAPE sebesar 11,35% untuk prediksi 3 hari, 14,86% untuk 7 hari, dan 34,23% untuk 14 hari. Ini menunjukkan bahwa kalibrasi STL menghasilkan nilai prediksi yang lebih baik. Kesimpulan penelitian ini model STL terbukti lebih akurat dalam memprediksi konsentrasi PM10, baik sebelum maupun sesudah kalibrasi, serta mampu menangkap pola musiman dan tren yang kompleks. Proses kalibrasi memberikan kontribusi signifikan dalam meningkatkan keandalan prediksi, menjadikannya lebih relevan untuk keperluan pengambilan kebijakan lingkungan. Rekomendasi penelitian ini mencakup integrasi model STL dengan pendekatan machine learning untuk prediksi jangka panjang dan penggabungan data meteorologi serta sumber emisi guna meningkatkan akurasi dan ketepatan strategi mitigasi pencemaran udara berbasis data.
====================================================================================================================================
Air quality in major cities like Surabaya continues to decline due to the increasing intensity of industrial activities, transportation, and infrastructure development. Particulate matter with a diameter ≤10 microns (PM10) has become one of the main pollutants, significantly affecting public health. To support air pollution control efforts, this study aims to predict PM10 concentrations in Surabaya City using time series models: Seasonal Trend Decomposition Procedure based on Loess (STL) and Seasonal Autoregressive Integrated Moving Average (SARIMA), based on daily data from the SPKU 6 Wonorejo station from May to July 2024. The STL model is used to decompose the data into trend, seasonal, and residual components, while SARIMA is applied to model both seasonal and non-seasonal patterns. The evaluation was carried out using Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). To improve accuracy, a calibration process was conducted between the predicted and actual data. The results showed that after calibration, STL yielded a MAPE of 11,35% for 3-day predictions, 14,86% for 7-day predictions, and 34,23% for 14-day predictions. This indicates that the calibration of STL produced better prediction values. In conclusion, the STL model proves to be more accurate for PM10 forecasting, both before and after calibration, thanks to its ability to capture complex seasonal patterns and fluctuating trends. Calibration significantly enhances prediction reliability, making it more suitable for data-driven policy formulation. This study recommends integrating STL with machine learning approaches for long-term forecasting and incorporating meteorological and emission source data to further improve the accuracy and effectiveness of air pollution mitigation strategies.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Pencemaran Udara, Prediksi Polutan , PM10, SARIMA, STL Air pollution, Pollutant prediction, PM10, SARIMA, STL
Subjects:	T Technology > TD Environmental technology. Sanitary engineering > TD883 Air quality management. T Technology > TD Environmental technology. Sanitary engineering > TD883.5 Air--Pollution
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Environmental Engineering > 25201-(S1) Undergraduate Thesis
Depositing User:	Mohammad Hammam Andare Yanuar
Date Deposited:	23 Jul 2025 01:52
Last Modified:	23 Jul 2025 01:52
URI:	http://repository.its.ac.id/id/eprint/120664

Actions (login required)

View Item