Irawan, Randi (2025) PREDIKSI TINGGI MUKA AIR DI LAHAN GAMBUT PROVINSI JAMBI BERDASARKAN CURAH HUJAN, SUHU DAN KELEMBAPAN UDARA MENGGUNAKAN METODE RANDOM FOREST REGRESSION. S1 thesis, Universitas Jambi.
![[img]](https://repository.unja.ac.id/style/images/fileicons/text.png) |
Text
BAB I.pdf Download (117kB) |
|
Text
BAB V.pdf Download (93kB) |
|
Text
COVER.pdf Download (86kB) |
|
Text
ABSTRAK.pdf Download (116kB) |
|
Text
DAFTAR PUSTAKA.pdf Download (115kB) |
|
Text
HALAMAN PENGESAHAN.pdf Download (346kB) |
|
Text
FULL SKRIPSI.pdf Restricted to Repository staff only Download (8MB) |
Abstract
Kebakaran lahan di Indonesia selama 2015–2019 mencapai 4,4 juta hektare, dengan 50% di antaranya terjadi di lahan gambut. Provinsi Jambi menjadi salah satu wilayah yang terdampak besar, di mana sekitar 67% area terbakar merupakan lahan gambut. Rendahnya tinggi muka air (TMA) disebut sebagai salah satu faktor utama pemicu kebakaran lahan gambut, sebagaimana dinyatakan oleh BRIN. Kondisi fluktuasi TMA di Provinsi Jambi menunjukkan bahwa perubahan TMA sangat dipengaruhi oleh unsur-unsur iklim seperti curah hujan, suhu, dan kelembapan udara, namun tidak mengikuti pola linier yang tetap. Sebagai contoh, pada tahun 2019 bulan Februari, curah hujan mencapai 21,824 mm dengan suhu 26,321°C dan kelembapan udara 89,25%, menghasilkan TMA positif sebesar 0,157 meter. Namun pada bulan Juni, meskipun kelembapan masih tinggi (85,5%), rendahnya curah hujan (6,300 mm) dan meningkatnya suhu (27,467°C) menyebabkan TMA turun signifikan menjadi -0,292 meter. Hal serupa terjadi pada bulan Oktober, di mana curah hujan naik sedikit menjadi 6,909 mm dan suhu turun menjadi 26,748°C, namun TMA tetap negatif yaitu -0,303 meter. Variasi ini memperlihatkan bahwa perubahan TMA tidak mengikuti pola hubungan linier terhadap satu variabel tertentu, melainkan merupakan hasil dari interaksi kompleks dan bersifat non-linier antara curah hujan, suhu, dan kelembapan udara. Sehingga kontribusi masing-masing unsur iklim terhadap TMA saling bergantung satu sama lain dan tidak dapat dijelaskan secara terpisah, maka diperlukan pendekatan analisis yang mampu menangkap dinamika interaktif tersebut secara menyeluruh. Hubungan antar variabel iklim dan TMA yang bersifat kompleks dan tidak linier ini menuntut metode analisis yang mampu menangkap interaksi dan dinamika yang saling memengaruhi secara simultan. Salah satu pendekatan yang sesuai adalah machine learning, karena tidak memerlukan asumsi linieritas atau distribusi data tertentu. Pada penelitian ini, metode Random Forest Regression adalah salah satu pendekatan yang mampu menangani hubungan non-linier, mendeteksi interaksi antar variabel prediktor, serta memberikan hasil prediksi yang stabil dan akurat. Pada penelitian ini, proses pemodelan dilakukan menggunakan bahasa pemrograman Python dengan bantuan pustaka scikit-learn, yang menyediakan berbagai fungsi untuk implementasi algoritma machine learning, termasuk Random Forest Regression dan teknik validasi seperti Grid Search dan Cross Validation dalam membangun model prediksi tinggi muka air. Sebelum model dibangun secara penuh, dilakukan Hyperparameter Tuning terhadap empat parameter utama, yaitu max_depth, min_samples_split, min_samples_leaf, dan n_estimators. Nilai kandidat yang diuji masing-masing adalah max_depth = (2, 3, 4), min_samples_split = (2, 5), min_samples_leaf = (1, 2), dan n_estimators = (100, 200, 300, 400, 500), sehingga menghasilkan total 60 kombinasi. Evaluasi terhadap kombinasi tersebut dilakukan menggunakan pendekatan Grid Search dan 5-Fold Cross Validation, yang menghasilkan total 300 kali pelatihan model. Berdasarkan proses hyperparameter tuning hasil evaluasi diperoleh kombinasi terbaik pada max_depth = 3, min_samples_split = 5, min_samples_leaf = 2, dan n_estimators = 500, dengan nilai rata-rata Mean Squared Error (MSE) terendah sebesar 0.01043. Evaluasi Model menunjukkan kinerja yang baik, dengan nilai RMSE sebesar 0.1176 dan R² sebesar 0.7169 pada data testing. Hasil ini menunjukkan kemampuan model dalam menjelaskan variasi data dan memprediksi tinggi muka air dengan tingkat kesalahan yang rendah. Analisis Feature Importance menunjukkan bahwa curah hujan memiliki kontribusi terbesar terhadap TMA sebesar 91.4%, diikuti oleh kelembapan udara sebesar 5.1%, dan suhu sebesar 3.5%. Temuan ini menegaskan bahwa curah hujan merupakan faktor paling dominan dalam pengendalian tinggi muka air, yang sangat penting untuk mendukung strategi mitigasi kebakaran lahan gambut secara berkelanjutan di Provinsi Jambi. Land fires in Indonesia during 2015–2019 reached 4.4 million hectares, with 50% of them occurring on peatlands. Jambi Province is one of the areas that is most affected, where around 67% of the burned area is peatland. Low water level (TMA) is cited as one of the main factors triggering peatland fires, as stated by BRIN. The fluctuating conditions of TMA in Jambi Province show that changes in TMA are strongly influenced by climatic elements such as rainfall, temperature, and air humidity, but do not follow a fixed linear pattern. For example, in February 2019, rainfall reached 21,824 mm with a temperature of 26,321°C and an air humidity of 89.25%, resulting in a positive TMA of 0,157 meters. However, in June, despite the high humidity (85.5%), low rainfall (6,300 mm) and rising temperatures (27,467°C) caused the TMA to drop significantly to -0.292 meters. A similar thing happened in October, where rainfall rose slightly to 6.909 mm and the temperature dropped to 26.748°C, but the TMA remained negative at -0.303 meters. This variation shows that changes in TMA do not follow a pattern of linear relationships to one particular variable, but are the result of complex and non-linear interactions between precipitation, temperature, and air humidity. So that the contribution of each climate element to TMA is interdependent with each other and cannot be explained separately, an analytical approach is needed that is able to capture these interactive dynamics comprehensively. The relationship between climate variables and TMA is complex and non- linear in nature and requires analytical methods that are able to capture interactions and dynamics that affect each other simultaneously. One appropriate approach is machine learning, as it does not require certain assumptions of linearity or distribution of data. In this study, the Random Forest Regression method is one of the approaches that is able to handle non-linear relationships, detect interactions between predictor variables, and provide stable and accurate prediction results. In this study, the modeling process was carried out using the Python programming language with the help of the scikit-learn library, which provides various functions for the implementation of machine learning algorithms, including Random Forest Regression and validation techniques such as Grid Search and Cross Validation in building water level prediction models. Before the model is fully built, Hyperparameter Tuning is carried out on four main parameters, namely max_depth, min_samples_split, min_samples_leaf, and n_estimators. The values of the tested candidates were max_depth = (2, 3, 4), min_samples_split = (2, 5), min_samples_leaf = (1, 2), and n_estimators = (100, 200, 300, 400, 500), resulting in a total of 60 combinations. The evaluation of the combination was carried out using the Grid Search and 5-Fold Cross Validation approaches, which resulted in a total of 300 model trainings. Based on the hyperparameter tuning process, the results of the evaluation obtained the best combination at max_depth = 3, min_samples_split = 5, min_samples_leaf = 2, and n_estimators = 500, with the lowest average Mean Squared Error (MSE) value of 0.01043. The Model Evaluation showed good performance, with an RMSE value of 0.1176 and an R² of 0.7169 in the testing data. These results demonstrate the model's ability to explain data variations and predict water table height with a low error rate. Feature Importance analysis showed that rainfall had the largest contribution to TMA at 91.4%, followed by air humidity at 5.1%, and temperature at 3.5%. These findings confirm that rainfall is the most dominant factor in water level control, which is critical to support sustainable peatland fire mitigation strategies in Jambi Province.
| Type: | Thesis (S1) |
|---|---|
| Uncontrolled Keywords: | Gambut, Tinggi Muka Air, Random Forest Regression |
| Subjects: | L Education > L Education (General) |
| Divisions: | Fakultas Sains dan Teknologi > Matematika |
| Depositing User: | IRAWAN |
| Date Deposited: | 10 Jul 2025 07:22 |
| Last Modified: | 03 Aug 2025 05:31 |
| URI: | https://repository.unja.ac.id/id/eprint/83073 |
Actions (login required)
 |
View Item |
