Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/164486
Title: Pemodelan Spatio-Temporal PM2.5 di Jakarta Berdasarkan Faktor Meteorologi Menggunakan Modifikasi Generalized Lasso
Other Titles: Spatio-Temporal Modeling of PM2.5 in Jakarta Using a Modified Generalized Lasso Based on Meteorological Factors
Authors: Rahardiantoro, Septian
Masjkur, Mohammad
Rofiana, Cahyani Dyah
Issue Date: 2025
Publisher: IPB University
Abstract: Udara bersih merupakan komponen penting bagi kesehatan manusia, terutama di kota besar seperti Jakarta dengan mobilitas yang tinggi. Aktivitas tersebut menghasilkan polutan, salah satunya partikel halus PM2.5 yang sangat berbahaya. Penelitian ini menerapkan model spatio-temporal generalized lasso termodifikasi dengan matriks penalti D berbasis queen’s contiguity untuk mengidentifikasi pengaruh yang sama pada peubah meteorologi terhadap PM2.5 di setiap waktu dan wilayah. Penentuan parameter regularisasi dilakukan dengan membandingkan nilai lambda optimum dari dua metode validasi, yaitu Approximate Leave-One-Out Cross-Validation (ALOCV) dan Generalized Cross-Validation (GCV). Lambda optimum dari GCV sebesar 34,673 dipilih dengan nilai RMSE terkecil sebesar 57,220. Hasil analisis menunjukkan bahwa pengaruh faktor meteorologi cenderung bersifat heterogen berdasarkan waktu atau musim tertentu. Kelembaban (X2) merupakan peubah paling berpengaruh secara spasial dan temporal, lalu diikuti oleh curah hujan (X4) dan kecepatan angin (X3) dengan pengaruh berlawanan arah, serta suhu (X1) dengan pengaruh searah terhadap PM2.5. Generalized lasso mampu mengidentifikasi pola spasial dan temporal dari faktor meteorologi terhadap PM2.5 secara efektif, serta memberikan kontribusi penting dalam pemahaman dinamika polusi udara berbasis wilayah dan waktu.
Clean air is a crucial component for human health, particularly in metropolitan areas such as Jakarta, where mobility is high. Such activity contributes to the emission of pollutants, including fine particulate matter (PM2.5), which poses significant health risks. This study applies a modified spatio-temporal generalized lasso model with a penalty matrix ?? based on queen’s contiguity to identify consistent effects of meteorological variables on PM2.5 across both time and space. The regularization parameter D is determined by comparing two validation methods: Approximate Leave-One-Out Cross-Validation (ALOCV) and Generalized Cross-Validation (GCV). The optimal lambda from GCV is 34.673 was selected due to the lowest RMSE of 57.220. The analysis reveals that meteorological influences on PM2.5 are heterogeneous depending on time or seasonal patterns. Humidity (X2) is identified as the most influential variable both spatially and temporally, followed by rainfall (X4) and wind speed (X3), which show inverse effects, and temperature (X1), which shows a positive association with PM2.5. The generalized lasso effectively captures spatio-temporal patterns of meteorological impacts on PM2.5 and contributes significantly to understanding the dynamics of air pollution across regions and time.
URI: http://repository.ipb.ac.id/handle/123456789/164486
Appears in Collections:UT - Statistics and Data Sciences

Files in This Item:
File Description SizeFormat 
cover_G1401211005_68e140b23aa94d2da7516c11c509cda2.pdfCover551.09 kBAdobe PDFView/Open
fulltext_G1401211005_1c12ae3fecd74343bf9963b157731c53.pdf
  Restricted Access
Fulltext1.29 MBAdobe PDFView/Open
lampiran_G1401211005_95d7876a319f42b09437ea6e1a3fe314.pdf
  Restricted Access
Lampiran376.74 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.