Claim Missing Document
Check
Articles

Found 4 Documents
Search

PEMANFAATAN TEKNIK RGB PADA CITRA SATELIT HIMAWARI-8 UNTUK ANALISIS DINAMIKA ATMOSFER KEJADIAN BANJIR LAMPUNG 20 - 21 FEBRUARI 2017 Paski, Jaka Anugrah Ivanda; Sepriando, Alpon; Pertiwi, Dyah Ajeng Sekar
Jurnal Meteorologi Klimatologi dan Geofisika Vol 4 No 3 (2017): Jurnal Meteorologi Klimatologi dan Geofisika
Publisher : Unit Penelitian dan Pengabdian Masyarakat Sekolah Tinggi Meteorologi Klimatologi dan Geofisika

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (789.46 KB) | DOI: 10.36754/jmkg.v4i3.48

Abstract

Teknik RGB (Red-Green-Blue) merupakan salah satu teknik intepretasi citra satelit dengan mengombinasikan beberapa kanal secara tumpang tindih warna merah, hijau dan biru untuk menyajikan informasi yang lebih mudah dipahami. Teknik RGB dapat digunakan dalam kajian analisis cuaca, terutama untuk mengidentifikasi kondisi khusus seperti bencana hidrometeorologi. Kejadian banjir pada tanggal 20 - 21 Februari 2017 yang merendam sekurangnya 7 kabupaten dan 1 kota di Provinsi Lampung yang dindikasikan terjadi karena hujan ekstrim yang merata di wilayah Lampung. Hasil pelaporan curah hujan di stasiun Klimatologi Masgar terukur 107.0 mm/hari, Pos Pengamatan Politeknik Negeri Lampung terukur 159.6 mm/hari dan Pos Pengamatan Kemiling Bandar Lampung terukur 154.0 mm/hari dimana curah hujan termasuk dalam kategori hujan sangat lebat BMKG (> 100 mm/hari). Hasil analisis kondisi regional menunjukan adanya tekanan udara rendah di barat lampung dan daerah konvergensi serta shearline di Lampung bagian barat dan tengah. Analisis citra satelit menunjukan adanya kumpulan awan dengan suhu puncak yang sangat dingin, teknik RGB menggunakan identifikasi mikrofisis atmosfer pada malam hari (Night Microphysics) dan sebaran massa udara (Air Mass) menunjukan adanya proses mikrofisis yang intensif serta aliran massa udara penyebab awan hujan yang tumbuh dan meluas di wilayah Lampung sebelum dan saat terjadinya banjir. Hasil produk olahan HCAI (Highresolution Cloud Analysis Information) menunjukan awan didominasi oleh awan Comulonimbus (Cb) dan awan konvektif padat (Dense Cloud).
DETEKSI KEBAKARAN HUTAN DAN LAHAN MENGGUNAKAN CITRA SATELIT HIMAWARI-8 DI KALIMANTAN TENGAH Sepriando, Alpon; Hartono, Hartono; Jatmiko, Retnadi Heru
Jurnal Sains & Teknologi Modifikasi Cuaca Vol 20, No 2 (2019): December 2019
Publisher : BPPT

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1431.029 KB) | DOI: 10.29122/jstmc.v20i2.3884

Abstract

IntisariKebakaran hutan dan lahan terjadi hampir setiap tahun di Indonesia, terutama di wilayah Sumatera dan Kalimantan saat musim kemarau. Deteksi kebakaran hutan dan lahan dengan citra satelit menggunakan indikator yang disebut titik panas. Titik panas yang digunakan saat ini di Indonesia diperoleh dari pengolahan data citra satelit berorbit polar (MODIS dan VIIRS) dengan resolusi temporal yang rendah, yaitu hanya 6 kali dalam sehari. Tujuan dari penelitian ini adalah memanfaatkan data citra satelit Himawari-8 untuk deteksi kebakaran hutan dan lahan yang menghasilkan titik panas dengan resolusi temporal 10 menit, dimana hasilnya di validasi dengan citra polar dan data kebakaran lapangan. Lokasi penelitian berada di Provinsi Kalimantan Tengah dan waktu penelitian adalah bulan September 2019. Data yang digunakan untuk pengolahan adalah 5 saluran Advanced Himawari Imager, peta batas administrasi dan tutupan lahan. Pemrosesan data citra satelit mencakup pemilihan piksel penutup lahan dan batas administrasi, penentuan waktu pengamatan, eliminasi piksel awan, Algoritma Pemantau Kebakaran Aktif, dan validasi hasil. Data citra Himawari-8 dapat diolah menjadi titik panas dengan temporal 10 menit. Validasi terhadap citra polar memiliki tingkat akurasi 66,2%-75,4%, comission error 28,2-46,9% dan omission error 24,6-33,8%. Tingginya comision error terhadap citra VIIRS dikarenakan citra VIIRS memiliki resolusi spasial yang jauh lebih tinggi dibandingkan dengan citra Himawari-8.  AbstractForest and land fires occur almost every year in Indonesia, especially in Sumatra and Kalimantan during the dry season. Detection of forest and land fires with satellite imagery uses an indicator called a hotspot. The hotspots used today in Indonesia are obtained from the processing of polar orbital satellite image data (MODIS and VIIRS) with a low temporal resolution, which is only six times a day. The purpose of this study is to utilize Himawari-8 satellite imagery data for the detection of forest and land fires that produce hotspots with a temporal resolution of 10 minutes, where the results are validated with polar imagery and field fire data. The research location is in Central Kalimantan Province, and the time of the study is September 2019. Data used for processing are 5 Advanced Himawari Imager channels, administrative boundary maps, and land cover. Processing of satellite imagery data includes the selection of cover pixels and administrative boundaries, determination of observation time, elimination of cloud pixels, Active Fire Monitoring Algorithm, and validation of results. Himawari-8 image data can be processed into hotspots with a temporal 10 minutes. Validation of polar images has an accuracy rate of 66.2% -75.4%, commission error 28.2-46.9% and omission error 24.6-33.8%. The high commission error on the VIIRS image is because the VIIRS image has a much higher spatial resolution compared to the Himawari-8 image. 
Evaluasi Kinerja Algoritma Random Forest dan KNN untuk Prediksi Cuaca di Jakarta Wibowo, Nurdin; Hardyanto, Wahyu; Sepriando, Alpon; Djuniadi, Djuniadi; Nurmayati, Tri Nurmayati; Alam, Fakhrul
Jurnal Pendidikan Matematika : Judika Education Vol. 9 No. 3 (2026): Jurnal Pendidikan Matematika:Judika Education
Publisher : Institut Penelitian Matematika, Komputer, Keperawatan, Pendidikan dan Ekonomi (IPM2KPE)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.31539/wzj0r898

Abstract

This study aims to compare the performance of Random Forest and K-Nearest Neighbors (KNN) algorithms in classifying daily weather categories at Kemayoran Meteorological Station, Jakarta. The data used were BMKG observation data from 2017 to 2023, with classification targets consisting of No Significant Weather, RA, TS, and TS,RA. The variables included temperature, rainfall, air pressure, humidity, wind speed, sunshine duration, wind direction, and month. The data were preprocessed and divided into training and testing sets using an 80:20 ratio. The results showed that Random Forest achieved an accuracy of 78% with a weighted average F1-score of 0.75, while KNN achieved an accuracy of 65% with a weighted average F1-score of 0.59. Random Forest performed better in classifying dominant weather categories, although both algorithms still had limitations in identifying minority categories, particularly TS. These findings indicate that Random Forest is more suitable for daily weather classification based on BMKG observation data in urban areas.   Keywords: K-Nearest Neighbors; Random Forest; Weather Classification
Prediksi Fine Fuel Moisture Code Menggunakan Algoritma Random Forest untuk Sistem Peringatan Dini Kebakaran Hutan Mulyaningsih, Sri; Hardyanto, Wahyu; Sepriando, Alpon; Djuniadi, Djuniadi
Jurnal Pendidikan Matematika : Judika Education Vol. 8 No. 6 (2025): Jurnal Pendidikan Matematika:Judika Education
Publisher : Institut Penelitian Matematika, Komputer, Keperawatan, Pendidikan dan Ekonomi (IPM2KPE)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.31539/xy82p541

Abstract

This study aims to predict the Fine Fuel Moisture Code (FFMC) as an indicator of the ignitability level of natural fuels using the Random Forest algorithm. FFMC is an essential component of fire danger rating systems, used to describe the dryness level of fine fuels; therefore, accurate prediction is crucial for forest and land fire mitigation efforts. The research utilizes meteorological variables—including maximum temperature, relative humidity, wind speed, and rainfall—as input parameters. The model implementation is carried out in Google Colab using the Python programming language, accompanied by hyperparameter tuning through GridSearchCV to obtain the most optimal model configuration.The results indicate that relative humidity and rainfall are the most influential meteorological features in predicting FFMC. The developed Random Forest model demonstrates excellent performance, as reflected by a Mean Squared Error of 8,875 and an R-squared value of 0,974, which signifies a very high predictive capability. These findings confirm that the machine-learning approach based on Random Forest is able to produce precise FFMC estimations, thereby giving the model strong potential for integration into early warning systems for forest and land fires. Such integration would support decision-making processes, enhance the effectiveness of mitigation strategies, and strengthen preparedness in managing fire-related risks.   Keywords: Fine Fuel Moisture Code, Fire Danger Prediction, Meteorological Features, Machine Learning, Random Forest.