Perbandingan Metode Partisi Machine Learning untuk Kerentanan Kebakaran Lahan Gambut Kalimantan pada Variasi ENSO

Intan Bryliana Putri; Latifah Nurul Qomariyatuzzamzami; Purwanti Lelly Sabrina; Yahya Darmawan

doi:10.30865/json.v8i1.10022

Authors

Intan Bryliana Putri Sekolah Tinggi Meteorologi Klimatologi dan Geofisika
Latifah Nurul Qomariyatuzzamzami Sekolah Tinggi Meteorologi Klimatologi dan Geofisika
Purwanti Lelly Sabrina Sekolah Tinggi Meteorologi Klimatologi dan Geofisika
Yahya Darmawan Sekolah Tinggi Meteorologi Klimatologi dan Geofisika

DOI:

https://doi.org/10.30865/json.v8i1.10022

Keywords:

machine learning, kNNDM, kerentanan kebakaran, ENSO, gambut tropis

Abstract

Lahan gambut Kalimantan merupakan wilayah rawan kebakaran yang polanya dipengaruhi fase El Niño -Southern Oscillation (ENSO). Penelitian ini membandingkan metode partisi acak KFOLD k=5 dan metode partisi spasial BANDS dalam pemodelan kerentanan kebakaran lahan gambut Kalimantan periode 2015 – 2024 , menggunakan algoritma machine learning tunggal melalui paket ENMTML pada tiga fase ENSO. Random Forest dengan KFOLD mencapai AUC 0,999 pada El Niño, namun diagnostik kNNDM menunjukkan statistik Wasserstein KFOLD selalu lebih tinggi dibanding BANDS yang berkisar 90.218-121.109 meter berbanding hanya 19.096-77.225 meter, sehingga performa tinggi KFOLD kurang merepresentasikan kondisi prediksi sesungguhnya. Pada metode BANDS, algoritma terbaik berbeda tiap fase yaitu Maximum Entropy default dengan AUC 0,929 pada El Niño, Boosted Regression Trees dengan AUC 0,983 pada Netral, dan Random Forest dengan AUC 0,870 pada La Niña. Kalimantan Tengah dan Kalimantan Selatan menunjukkan proporsi kerentanan tinggi hingga sangat tinggi terbesar dibanding wilayah lain pada fase ENSO yang berbeda. Variasi ENSO tidak berperan signifikan pada kerentanan kebakaran lahan gambut di Kalimantan dibandingkan dengan aktivitas konveksi lokal dan prediktor kelembaban lingkungan. Pemilihan skema partisi yang memperhitungkan struktur spasial terbukti penting untuk menghasilkan peta kerentanan yang lebih realistis dan berpotensi mendukung upaya konservasi lahan gambut yang lebih sesuai dengan karakteristik lingkungan setempat.

References

[1] S. L. Wilkinson, R. Andersen, P. A. Moore, S. J. Davidson, G. Granath, and J. M. Waddington, “Wildfire and degradation accelerate northern peatland carbon release,” Nat. Clim. Chang., vol. 13, no. 5, pp. 456–461, May 2023, doi: 10.1038/s41558-023-01657-w.

[2] L. Ma et al., “A globally robust relationship between water table decline, subsidence rate, and carbon release from peatlands,” Commun Earth Environ, vol. 3, no. 1, p. 254, Oct. 2022, doi: 10.1038/s43247-022-00590-8.

[3] J. Koarashi et al., “Progressive release of long-stored carbon from tropical peatland disturbances,” Nat Commun, vol. 17, no. 1, p. 4369, May 2026, doi: 10.1038/s41467-026-72890-y.

[4] L. Tacconi, “Preventing fires and haze in Southeast Asia,” Nature Clim Change, vol. 6, no. 7, pp. 640–643, Jul. 2016, doi: 10.1038/nclimate3008.

[5] A. Schmidt et al., “Fire frequency, intensity, and burn severity in Kalimantan’s threatened Peatland areas over two Decades,” Front. For. Glob. Change, vol. 7, p. 1221797, Feb. 2024, doi: 10.3389/ffgc.2024.1221797.

[6] W. Cai et al., “Changing El Niño–Southern Oscillation in a warming climate,” Nat Rev Earth Environ, vol. 2, no. 9, pp. 628–644, Aug. 2021, doi: 10.1038/s43017-021-00199-z.

[7] S. Nurdiati, A. Sopaheluwakan, and P. Septiawan, “Joint Pattern Analysis of Forest Fire and Drought Indicators in Southeast Asia Associated with ENSO and IOD,” Atmosphere, vol. 13, no. 8, p. 1198, Jul. 2022, doi: 10.3390/atmos13081198.

[8] R. Salmayenti, A. J. Baird, J. Holden, and D. V. Spracklen, “Drainage density and land cover interact to affect fire occurrence in Indonesian peatlands,” Environ. Res. Lett., vol. 20, no. 5, p. 054036, May 2025, doi: 10.1088/1748-9326/adc755.

[9] M. W. Jones et al., “Global and Regional Trends and Drivers of Fire Under Climate Change,” Reviews of Geophysics, vol. 60, no. 3, p. e2020RG000726, Sep. 2022, doi: 10.1029/2020RG000726.

[10] W. Thuiller, “Ecological niche modelling,” Current Biology, vol. 34, no. 6, pp. R225–R229, Mar. 2024, doi: 10.1016/j.cub.2024.02.018.

[11] N. Sillero et al., “Want to model a species niche? A step-by-step guideline on correlative ecological niche modelling,” Ecological Modelling, vol. 456, p. 109671, Sep. 2021, doi: 10.1016/j.ecolmodel.2021.109671.

[12] S. Aldiansyah and F. Wardani, “Assessment of resampling methods on performance of landslide susceptibility predictions using machine learning in Kendari City, Indonesia,” Water Practice & Technology, vol. 19, no. 1, pp. 52–81, Jan. 2024, doi: 10.2166/wpt.2024.002.

[13] D. Theng and K. K. Bhoyar, “Feature selection techniques for machine learning: a survey of more than two decades of research,” Knowl Inf Syst, vol. 66, no. 3, pp. 1575–1637, Mar. 2024, doi: 10.1007/s10115-023-02010-5.

[14] L. Nguyen Van and G. Lee, “Optimizing Stacked Ensemble Machine Learning Models for Accurate Wildfire Severity Mapping,” Remote Sensing, vol. 17, no. 5, p. 854, Feb. 2025, doi: 10.3390/rs17050854.

[15] C. Li et al., “Integrated spatial generalized additive modeling for forest fire prediction: a case study in Fujian Province, China,” J. For. Res., vol. 36, no. 1, p. 30, Feb. 2025, doi: 10.1007/s11676-025-01822-1.

[16] S. Janizadeh et al., “Combination four different ensemble algorithms with the generalized linear model (GLM) for predicting forest fire susceptibility,” Geomatics, Natural Hazards and Risk, vol. 14, no. 1, p. 2206512, Dec. 2023, doi: 10.1080/19475705.2023.2206512.

[17] L. B. Prasetyo et al., “Assessing Sumatran Peat Vulnerability to Fire under Various Condition of ENSO Phases Using Machine Learning Approaches,” Forests, vol. 13, no. 6, p. 828, May 2022, doi: 10.3390/f13060828.

[18] A. F. A. D. Andrade, S. J. E. Velazco, and P. De Marco Júnior, “ENMTML: An R package for a straightforward construction of complex ecological niche models,” Environmental Modelling & Software, vol. 125, p. 104615, Mar. 2020, doi: 10.1016/j.envsoft.2019.104615.

[19] V. Bahn and B. J. McGill, “Testing the predictive performance of distribution models,” Oikos, vol. 122, no. 3, pp. 321–331, Mar. 2013, doi: 10.1111/j.1600-0706.2012.00299.x.