Firnanda, Tsalits Arifuddin Firnanda (2025) AquaSAR: Model Berbasis Algoritma Neural Network untuk Estimasi Klorofil-a Perairan menggunakan Citra Syntethic Aperture Radar (SAR) Sentinel-1. Other thesis, Institut Teknologi Sepuluh Nopember.
![]() |
Text
5016211075-Undergraduate_ThesIs.pdf Restricted to Repository staff only Download (5MB) | Request a copy |
Abstract
Danau merupakan ekosistem penting yang berperan dalam siklus hidrologi, irigasi, serta penyediaan sumber daya air bagi kehidupan. Namun, danau sering menghadapi tantangan serius seperti eutrofikasi akibat kelebihan nutrien yang mendorong pertumbuhan fitoplankton dan dapat memicu ledakan alga berbahaya yang menyebabkan degradasi kualitas air. Pemantauan kualitas air secara konvensional dengan pengambilan sampel memiliki keterbatasan dalam hal biaya dan juga waktu, sehingga digunakan metode alternatif dengan penginderaan jauh yang lebih efisien dan efektif. Akan tetapi, keterbatasan penginderaan jauh pasif di daerah tropis yang sering berawan menjadi hambatan tersendiri. Oleh karena itu, penelitian ini bertujuan untuk mengestimasi konsentrasi klorofil-a di Danau Laguna, Filipina, menggunakan citra Synthetic Aperture Radar(SAR) Sentinel-1 yang merupakan citra aktif sehingga tidak terpengaruh oleh kondisi cuaca berawan karena bekerja pada panjang gelombang mikro, dengan menggunakan algoritma neural network (NN) berbasis regresi. Pendekatan ini bertujuan untuk meningkatkan akurasi prediksi dibandingkan metode klasifikasi yang tidak memberikan nilai kontinu di setiap titiknya akan tetapi digambarkan dengan tingkatan kelas seperti penelitian sebelumnya menggunakan support vector mahine (SVM) yang hanya menghasilkan overall akurasi 67,74%. Proses penelitian mencakup pre-processing citra SAR untuk ekstraksi nilai backscatter, pelatihan model berbasis NN menggunakan skema 10-fold cross-validation dengan parameter evaluasi MSE, RMSE dan R². Hasil penelitian menunjukkan bahwa Model AquaSAR menunjukkan performa cukup stabil dengan rata-rata MSE sebesar 1,2956 µg2/L2, RMSE sebesar 1,1364 µg/L dan R² sebesar 0,3929. Perbandingan dengan model regresi linear dan arsitektur feedforward neural network lainnya menunjukkan bahwa model AquaSAR memberikan akurasi prediksi terbaik, dengan nilai evaluasi yang lebih unggul pada MSE, RMSE, dan R². Hasil pemetaan prediksi menunjukkan distribusi musiman klorofil-a pada Danau Laguna lebih tinggi pada musim kemarau yang disebabkan oleh intensitas cahaya matahari dan juga suhu perairan. Penelitian ini menunjukkan potensi pemanfaatan citra SAR dan machine learning untuk pemantauan kualitas air di wilayah tropis dengan kondisi cuaca yang memiliki banyak awan, serta kontribusi nyata dalam pengelolaan sumber daya air secara berkelanjutan, tetapi juga menegaskan perlunya peningkatan akurasi melalui integrasi data multisensor dan diversifikasi data latih.
============================================================================================================================================
Lakes are vital ecosystems that play a crucial role in the hydrological cycle, irrigation, and the provision of water resources for life. However, they often face serious challenges such as eutrophication, which results from excess nutrients that promote the overgrowth of phytoplankton and can trigger harmful algal blooms, leading to water quality degradation. Remote sensing has become a common solution for water quality monitoring, but passive remote sensing faces significant limitations in tropical regions due to frequent cloud cover. This study aims to estimate chlorophyll-a concentrations in Laguna Lake, Philippines, using Sentinel-1 Synthetic Aperture Radar(SAR) imagery. SAR is an active sensor system that operates in the microwave spectrum, allowing it to penetrate cloud cover and function under all weather conditions. The approach employs a regression-based neural network (NN) algorithm, aiming to improve prediction accuracy compared to classification-based methods, which only provide categorical levels rather than continuous values. One such method, support vector machine (SVM), previously yielded an overall accuracy of only 67,74%. The research process involved SAR image pre-processing to extract backscatter values, followed by training the NN model using a 10-fold cross-validation scheme. Model performance was evaluated using metrics such as mean squared error (MSE), root mean square error (RMSE), and the coefficient of determination (R²). Results showed that the AquaSAR model achieved relatively stable performance, with an average MSE of 1,2956 µg2/L2, RMSE of 1,1364 µg/L, and R² of 0,3929. However, the accuracy was still considered low for precise quantitative estimation. The predicted chlorophyll-a maps showed a seasonal distribution pattern in Laguna Lake, with higher concentrations observed during the dry season, likely due to increased sunlight intensity. This study highlights the potential of SAR imagery combined with machine learning for water quality monitoring in tropical regions with frequent cloud cover. It also offers meaningful contributions to sustainable water resource management. Nonetheless, it emphasizes the need for improved accuracy through the integration of multisensor data and the diversification of training datasets.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | Backscatter SAR, Eutrofikasi, Machine Learning, Life Below Water, SAR Backscatter, Eutrophication, Machine Learning, Life Below Water |
Subjects: | G Geography. Anthropology. Recreation > G Geography (General) > G70.5.I4 Remote sensing G Geography. Anthropology. Recreation > GC Oceanography > GC101.2 Seawater--Analysis Q Science > QA Mathematics > QA336 Artificial Intelligence T Technology > TD Environmental technology. Sanitary engineering > TD890 Global Environmental Monitoring System |
Divisions: | Faculty of Civil Engineering and Planning > Geomatics Engineering > 29202-(S1) Undergraduate Thesis |
Depositing User: | Tsalits Arifuddin Firnanda |
Date Deposited: | 22 Jul 2025 09:24 |
Last Modified: | 22 Jul 2025 09:24 |
URI: | http://repository.its.ac.id/id/eprint/120685 |
Actions (login required)
![]() |
View Item |