| dc.contributor.advisor | Ridha, Ahmad | |
| dc.contributor.advisor | Agmalaro, Muhammad Asyhar | |
| dc.contributor.author | ALFIANDI, RIDWAN CAHYA | |
| dc.date.accessioned | 2026-08-14T06:32:15Z | |
| dc.date.available | 2026-08-14T06:32:15Z | |
| dc.date.issued | 2026 | |
| dc.identifier.uri | http://repository.ipb.ac.id/handle/123456789/178831 | |
| dc.description.abstract | Konsentrasi PM2.5 tahunan di Jakarta jauh melampaui pedoman WHO, namun pemantauan berbasis stasiun darat masih terbatas. Aerosol Optical Depth (AOD) dari produk MAIAC berpotensi melengkapi keterbatasan tersebut tetapi mengalami proporsi missing value yang sangat tinggi akibat tutupan awan di wilayah tropis. Penelitian ini bertujuan membangun dan mengevaluasi model imputasi AOD MAIAC serta model statistical downscaling konsentrasi PM2.5 menggunakan XGBoost di Jakarta selama periode Januari 2022 hingga Desember 2025. Proporsi missing value AOD berkisar antara 80,29% hingga 84,39%. Model imputasi AOD menggunakan prediktor aerosol MERRA-2 dan meteorologi ERA5 menghasilkan R² pada data pengujian sebesar 0,9518 hingga 0,9807. Data AOD hasil imputasi selanjutnya digunakan sebagai prediktor utama pada model statistical downscaling PM2.5 yang dikembangkan dengan sepuluh perlakuan konfigurasi fitur mencakup representasi spasial, fitur temporal, dan reduksi dimensi PCA. Perlakuan terbaik menghasilkan R² sebesar 0,7459 pada SPKU DKI1 Bundaran HI, 0,6841 pada SPKU DKI2 Kelapa Gading, dan 0,4861 pada SPKU DKI3 Jagakarsa. Penambahan fitur lag dan rolling mean meningkatkan performa pada dua stasiun, sedangkan PCA secara konsisten menurunkan performa. XGBoost terbukti efektif untuk imputasi AOD namun performa downscaling PM2.5 bervariasi antarstasiun dan dipengaruhi oleh tingginya proporsi data hasil imputasi. | |
| dc.description.abstract | Jakarta's annual PM2.5 concentration far exceeds the WHO guideline, yet ground-based monitoring remains limited. Aerosol Optical Depth (AOD) from the MAIAC product can potentially supplement this limitation but suffers from very high missing value proportions due to cloud cover in tropical regions. This study aimed to develop and evaluate MAIAC AOD imputation and PM2.5 statistical downscaling models using XGBoost in Jakarta from January 2022 to December 2025. Missing value proportions of AOD ranged from 80,29% to 84,39%. The AOD imputation model was built using MERRA-2 aerosol and ERA5 meteorological predictors, achieving test R² values of 0,9518 to 0,9807. The imputed AOD data were subsequently used as the primary predictor in the PM2.5 statistical downscaling model, which was developed with ten feature configuration treatments encompassing spatial representation, temporal features, and PCA dimensionality reduction. The best treatments yielded R² values of 0,7459 at DKI1 Bundaran HI, 0,6841 at DKI2 Kelapa Gading, and 0,4861 at DKI3 Jagakarsa. Lag and rolling mean features improved performance at two stations, while PCA consistently degraded performance. XGBoost proved effective for AOD imputation, but PM2.5 downscaling performance varied across stations, influenced by high imputed data proportions. | |
| dc.description.sponsorship | | |
| dc.language.iso | id | |
| dc.publisher | IPB University | id |
| dc.title | Statistical Downscaling Konsentrasi PM2.5 di Jakarta Menggunakan XGBoost Berbasis Imputasi AOD MAIAC | id |
| dc.title.alternative | | |
| dc.type | Skripsi | |
| dc.subject.keyword | AOD MAIAC | id |
| dc.subject.keyword | Imputasi | id |
| dc.subject.keyword | PM2.5 | id |
| dc.subject.keyword | Statistical Downscaling | id |
| dc.subject.keyword | XGBoost | id |
| dc.subtype | Undergraduate Theses | |