A comparative study of machine learning and convolutional neural network approaches for forest fire occurrence prediction

Abstract

The increasing frequency and severity of forest fires, driven by climate change and human activities, highlight the urgent need for accurate forest fire prediction models. This study developed a deep learning model based on a convolutional neural network (CNN) that integrates topographic and meteorological information to predict forest fire occurrences in South Korea. The CNN’s performance was compared against four traditional machine learning (ML) models: Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF) and XGBoost. Using a dataset of Korean forest fire records from 2011 to 2021, monthly meteorological data, and seven topographic variables, the models were validated through fivefold cross-validation. The results demonstrate that the CNN model, which effectively learns spatial patterns from gridded data, achieved superior predictive power with an accuracy of 0.962 and an F1-score of 0.962. However, ensemble models such as RF and XGBoost also showed excellent performance with accuracies of 0.931 and 0.937, respectively, while SVM and LR showed relatively low performance. Variable importance analysis with LR, RF and XGBoost consistently identified meteorological factors, particularly relative humidity, precipitation and temperature, as the primary drivers of fire occurrence. Among topographic variables, the Topographic Wetness Index (TWI) was the most influential. This study demonstrates that a prediction model considering both spatial patterns and climatic factors can significantly improve accuracy. These findings provide a robust basis for developing advanced early warning systems and targeted fire prevention strategies, which are the most direct methods for mitigating fire-related carbon emissions.

Keywords

Forest fire prediction / Convolutional neural network / Machine learning / Feature importance

References

[1] Adab H, Kanniah KD, Solaimani K. Modeling forest fire risk in the northeast of Iran using remote sensing and GIS techniques. Nat Hazards, 2013, 65(3): 1723-1743

[2] Aghajani H, Fallah A, Emadian SF. Modelling and analyzing the surface fire behaviour in Hyrcanian forest of Iran. J Forest Sci, 2014, 60(9): 353-362

[3] Alonso-Betanzos A, Fontenla-Romero O, Guijarro-Berdiñas B, Hernández-Pereira E, Inmaculada Paz Andrade M, Jiménez E, Luis Legido Soto J, Carballas T. An intelligent system for forest fire risk prediction and fire fighting management in Galicia. Expert Syst Appl, 2003, 25(4): 545-554

[4] Araveeporn A, Wanitjirattikal P (2024) Comparison of machine learning methods for binary classification of multicollinearity data. In: Proceedings of the 2024 7th international conference on mathematics and statistics. ACM 44–49. https://doi.org/10.1145/3686592.3686600

[5] Bhatt S, Chouhan U. An enhanced method for predicting and analysing forest fires using an attention-based CNN model. J for Res, 2024, 35(1 67

[6] Breiman L. Random forests. Mach Learn, 2001, 451): 5-32

[7] Browne MW. Cross-Validation Methods J Math Psychol, 2000, 44(1): 108-132

[8] Chuvieco E, Congalton RG. Application of remote sensing and geographic information systems to forest fire hazard mapping. Remote Sens Environ, 1989, 29(2): 147-159

[9] Clark JS, Royall PD, Chumbley C. The role of fire during climate change in an eastern deciduous forest at devil’s bathtub. N Y Ecol, 1996, 77(7): 2148-2166

[10] Drake JM, Randin C, Guisan A (2006) Modelling ecological niches with support vector machines. J Appl Ecol 43(3):424–432. https://doi.org/10.1111/j.1365-2664.2006.01141.x

[11] Dutta R, Aryal J, Das A, Kirkpatrick JB. Deep cognitive imaging systems enable estimation of continental-scale fire incidence from climate data. Sci Rep, 2013, 3: 3188

[12] Farrell A, Wang GM, Rush SA, Martin JA, Belant JL, Butler AB, Godwin D. Machine learning of large-scale spatial distributions of wild turkeys with high-dimensional environmental data. Ecol Evol, 2019, 9(10): 5938-5949

[13] Friedman JH. Greedy function approximation: a gradient boosting machine. Ann Stat, 2001, 29(5): 1189-1232

[14] Gao B, Shan YL, Liu XY, Yin SN, Yu B, Cui CX, Cao LL. Prediction and driving factors of forest fire occurrence in Jilin Province, China. J for Res, 2023, 35(1 21

[15] Hong HY, Jaafari A, Zenner EK. Predicting spatial patterns of wildfire susceptibility in the Huichang County, China: an integrated model to analysis of landscape indicators. Ecol Indic, 2019, 101: 878-891

[16] Hosmer DWJr, Lemeshow S, Sturdivant RX. Applied logistic regression, 2013, New York, USA, John Wiley & Sons

[17] Illarionova S, Shadrin D, Gubanov F, Shutov M, Tasuev U, Evteeva K, Mironenko M, Burnaev E. Exploration of geo-spatial data and machine learning algorithms for robust wildfire occurrence prediction. Sci Rep, 2025, 151 10712

[18] Jo HW, Krasovskiy A, Hong MN, Corning S, Kim W, Kraxner F, Lee WK. Modeling historical and future forest fires in South Korea: the FLAM optimization approach. Remote Sens, 2023, 155 1446

[19] Kim J, Kim T, Lee YE, Im S. Spatial and temporal variability of forest fires in the Republic of Korea over 1991–2020. Nat Hazards, 2025, 1218): 9801-9821

[20] Korea Forest Service (2023) Forest fires history. Korea Forest Service. Available online: https://safemap.go.kr/opna/data/dataView.doobjtId=207

[21] Korea Forest Service (2024) Statistical Yearbook of Forestry 2024. Korea Forest Service, No 11-1400000-000001-10

[22] Korea Forest Service, 2025. Recent status of forest fire damage. Korea Forest Service. Available online: https://www.index.go.kr/unity/potal/main/EachDtlPageDetail.dx_cd=1309

[23] Korea Meteorological Administration (2025) Climate Characteristics of Republic of Korea. Korea Meteorological Administration. Available online: https://www.weather.go.kr/w/obs-climate/climate/statistics/korea-char.do

[24] Kottek M, Grieser J, Beck C, Rudolf B, Rubel F. World Map of the Köppen-Geiger climate classification updated. Metz, 2006, 15(3): 259-263

[25] Kwak HB, Lee WK, Lee SY, Won MS, Koo KS, Lee BD, Lee MB. Cause-specific spatial point pattern analysis of forest fire in Korea. J Korean for Soc, 2010, 99(3): 259-266

[26] Lee K, Jo W. The gradient analysis of the Korean Peninsula by using DEM. J Korean Assoc Geogr Inf Stud, 2000, 3(1): 35-43

[27] Lee SY, Lee HP. Analysis of forest fire occurrence in Korea. Fire Sci Eng, 2006, 20(2): 54-63

[28] Lee SY, Han SY, Won MS, An SH, Lee MB. Developing of forest fire occurrence probability model by using the meteorological characteristics in Korea. Korean J Agric for Meteor, 2004, 6(4): 242-249

[29] Lee SW, Won MS, Lee HJ. Relationships between edge formation of burned forests and landscape characteristics with consideration on spatial autocorrelation. J Korean For Soc, 2013, 102(1): 113-121

[30] Lee C, Lim M, Lee Y. Machine learning for big data analytics in development of wildfire prediction models. J Korean Soc Hazard Mitig, 2023, 23(2): 29-39

[31] Li YD, Feng ZK, Chen SL, Zhao ZY. Wang FG (2020) Application of the artificial neural network and support vector machines in forest fire prediction in the Guangxi autonomous region. China Discrete Dyn Nat Soc, 2020, 1: 5612650

[32] Li CH, Su ZW, Ni RY, Wang GY, Ouyang YY, Zeng AC, Guo FT. Integrated spatial generalized additive modeling for forest fire prediction: a case study in Fujian Province. China J for Res, 2025, 36: 30

[33] Moore ID, Grayson RB, Ladson AR. Digital terrain modelling: a review of hydrological, geomorphological, and biological applications. Hydrol Process, 1991, 5(1): 3-30

[34] Noroozi F, Ghanbarian G, Safaeian R, Pourghasemi HR. Forest fire mapping: a comparison between GIS-based random forest and Bayesian models. Nat Hazards, 2024, 120(7): 6569-6592

[35] Oliveira S, Oehler F, San-Miguel-Ayanz J, Camia A, Pereira JMC. Modeling spatial patterns of fire occurrence in Mediterranean Europe using Multiple Regression and Random Forest. For Ecol Manag, 2012, 275: 117-129

[36] Pausas JG, Keeley JE. Wildfires and global change. Front Ecol Environ, 2021, 19(7): 387-395

[37] Pham BT, Jaafari A, Avand M, Al-Ansari N, Du Dinh T, Yen HPH, Van Phong T, Nguyen DH, Van Le H, Mafi-Gholami D, Prakash I, Thi Thuy H, Tuyen TT. Performance evaluation of machine learning methods for forest fire modeling and prediction. Symmetry, 2020, 12(6): 1022

[38] Pourtaghi ZS, Pourghasemi HR, Aretano R, Semeraro T. Investigation of general indicators influencing on forest fire and its susceptibility modeling using different data mining techniques. Ecol Indic, 2016, 64: 72-84

[39] Preeti T, Kanakaraddi S, Beelagi A, Malagi S, Sudi A (2021) Forest fire prediction using machine learning techniques. In: 2021 International Conference on Intelligent Technologies (CONIT). IEEE, pp. 1–6. https://doi.org/10.1109/conit51480.2021.9498448

[40] Randerson JT, Liu H, Flanner MG, Chambers SD, Jin Y, Hess PG, Pfister G, Mack MC, Treseder KK, Welp LR, Chapin FS, Harden JW, Goulden ML, Lyons E, Neff JC, Schuur EAG, Zender CS. The impact of boreal forest fire on climate warming. Science, 2006, 314(5802): 1130-1132

[41] Sakr GE, Elhajj IH, Mitri G, Wejinya UC (2010) Artificial intelligence for forest fire prediction. In: 2010 IEEE/ASME international conference on advanced intelligent mechatronics. Montreal, QC, Canada. IEEE, pp. 1311–1316.

[42] Sánchez-López N, Hudak AT, Boschetti L, Silva CA, Robertson K, Loudermilk EL, Bright BC, Callaham Jr. MA, Taylor MK (2023) A spatially explicit model of tree leaf litter accumulation in fire maintained longleaf pine forests of the southeastern US. Ecol Model 481:110369. https://doi.org/10.1016/j.ecolmodel.2023.110369

[43] Shi XH, Li QM, Qi Y, Huang TT, Li JM (2017) An accident prediction approach based on XGBoost. In: 2017 12th international conference on intelligent systems and knowledge engineering (ISKE). IEEE, pp. 1–7. https://doi.org/10.1109/iske.2017.8258806

[44] Sutanto SJ, Vitolo C, di Napoli C, D’Andrea M, van Lanen HAJ. Heatwaves, droughts, and fires: exploring compound and cascading dry hazards at the pan-European scale. Environ Int, 2020, 134 105276

[45] Vetrita Y, Cochrane MA. Fire frequency and related land-use and land-cover changes in Indonesia’s peatlands. Remote Sensing, 2020, 12(1 5

[46] Vilar del Hoyo L, Martín Isabel MP, Martínez Vega FJ. Logistic regression models for human-caused wildfire risk estimation: analysing the effect of the spatial accuracy in fire occurrence data. Eur J Forest Res, 2011, 1306): 983-996

[47] Wang SS, Wang YX. Quantifying the effects of environmental factors on wildfire burned area in the south central US using integrated machine learning techniques. Atmos Chem Phys, 2020, 20(18): 11065-11087

[48] Zhang GL, Wang M, Liu K. Forest fire susceptibility modeling using a convolutional neural network for Yunnan Province of China. Int J Disaster Risk Sci, 2019, 10(3): 386-403

[49] Zhang GL, Wang M, Liu K. Deep neural networks for global wildfire susceptibility modelling. Ecol Indic, 2021, 127 107735

Funding

Seoul National University

RIGHTS & PERMISSIONS

The Author(s)