Wildfire Susceptibility Assessment in Southern China: A Comparison of Multiple Methods

Abstract

Wildfire is a primary forest disturbance. A better understanding of wildfire susceptibility and its dominant influencing factors is crucial for regional wildfire risk management. This study performed a wildfire susceptibility assessment using multiple methods, including logistic regression, probit regression, an artificial neural network, and a random forest (RF) algorithm. Yunnan Province, China was used as a case study area. We investigated the sample ratio of ignition and nonignition data to avoid misleading results due to the overwhelming number of nonignition samples in the models. To compare model performance and the importance of variables among the models, the area under the curve of the receiver operating characteristic plot was used as an indicator. The results show that a cost-sensitive RF had the highest accuracy (88.47%) for all samples, and 94.23% accuracy for ignition prediction. The identified main factors that influence Yunnan wildfire occurrence were forest coverage ratio, month, season, surface roughness, 10 days minimum of the 6 h maximum humidity, and 10 days maxima of the 6 h average and maximum temperatures. These seven variables made the greatest contributions to regional wildfire susceptibility. Susceptibility maps developed from the models provide information regarding the spatial variation of ignition susceptibility, which can be used in regional wildfire risk management.

Keywords

China / Random forest / Variable importance rank / Wildfire susceptibility / Yunnan forest

References

[1] Aldersley A, Murray SJ, Cornell SE. Global and regional analysis of climate and human drivers of wildfire. Science of the Total Environment, 2011, 409(18): 3472-3481

[2] Bradley AP. The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognition, 1997, 30(7): 1145-1159

[3] Braun, W.J., B.L. Jones, J.S.W. Lee, D.G. Woolford, and B.M. Wotton. 2010. Forest fire risk assessment: An illustrative example from Ontario, Canada. Journal of Probability and Statistics 2010: Article No. 823018.

[4] Bonazountas M, Kallidromitou D, Kassomenos PA, Passas N. Forest fire risk analysis. Human and Ecological Risk Assessment: An International Journal, 2005, 11(3): 617-626

[5] Breiman L. Bagging predictors. Machine Learning, 1996, 24(2): 123-140.

[6] Breiman L. Random forests. Machine Learning, 2001, 45(1): 5-32

[7] Casella G, Robert CP, Wells MT. Generalized accept–reject sampling schemes. Lecture Notes-Monograph Series, 2004, 45: 342-347

[8] Catani F, Lagomarsino D, Segoni S, Tofani V. Landslide susceptibility estimation by random forests technique: Sensitivity and scaling issues. Natural Hazards and Earth System Science, 2013, 13(11): 2815-2831

[9] Chen F, Lin X-D, Niu S-K, Wang S, Li D. Influence of climate change on forest fire in Yunnan Province, southwestern China. Journal of Beijing Forestry University, 2012, 34(6): 7-15 (in Chinese)

[10] Chen F, Fan ZF, Niu SK, Zheng JM. The influence of precipitation and consecutive dry days on burned areas in Yunnan Province, southwestern China. Advances in Meteorology, 2014

[11] Del Río S, López V, Benítez JM, Herrera F. On the use of MapReduce for imbalanced big data using random forest. Information Sciences, 2014, 285: 112-137

[12] Domingos, P. 1999. Metacost: A general method for making classifiers cost-sensitive. Proceedings of the Fifth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Diego, CA, USA, 15–18 August 1999. 155–164. New York: ACM.

[13] Ebisuzaki W, Zhang L. Assessing the performance of the CFSR by an ensemble of analyses. Climate Dynamics, 2011, 37(11–12): 2541-2550

[14] Elkan, C. 2001. The foundations of cost-sensitive learning. Proceedings of the Seventeenth International Joint Conference on Artificial Intelligence. Seattle, Washington, USA, 4–10 August 2001. ACM Digital Library. San Francisco, CA: Morgan Kaufmann Publishers.

[15] Fang KNB, Jian-Bina WU. A review of technologies on random forests. Statistics & Information Forum, 2011, 2011(3): 33-39.

[16] Finney MA. The challenge of quantitative risk analysis for wildland fire. Forest Ecology and Management, 2005, 211(1): 97-108

[17] Hagan MT, Demuth HB, Beale MH. Neural network design, 1996, Boston: PWS Publishing

[18] Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology, 1982, 143(1): 29-36

[19] Hsu KL, Gupta HV, Sorooshian S. Artificial neural network modeling of the rainfall–runoff process. Water Resources Research, 1995, 31(10): 2517-2530

[20] Huang BW. The comprehensive regionalization compendium of Chinese nature. Collection of Geographical Publications, 1989, 21: 10-20 (in Chinese)

[21] IPCC (Intergovernmental Panel on Climate Change). 2013. Climate change 2013. The physical science basis. Contribution of Working Group I to the fifth assessment report of the Intergovernmental Panel on Climate Change, ed. T.F. Stocker, D. Qin, G.K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P.M. Midgley. Cambridge, UK: Cambridge University Press.

[22] Li XW, Fu GB, Zeppel MJB, Yu XB, Zhao G, Eamus D, Yu Q. Probability models of fire risk based on forest fire indices in contrasting climates over China. Journal of Resources and Ecology, 2012, 3(2): 105-117

[23] Liaw A, Wiener M. Classification and regression by randomForest. R News, 2002, 2(3): 18-22.

[24] Massada AB, Syphard AD, Stewart SI, Radeloff VC. Wildfire ignition-distribution modelling: A comparative study in the Huron-Manistee National Forest, Michigan, USA. International Journal of Wildland Fire, 2013, 22(2): 174-183

[25] McCullagh P, Nelder JA. Generalized linear models, 1989, London: CRC Press

[26] McCune B, Grace JB. Analysis of ecological communities, 2002, Gleneden Beach, OR: MjM Software Design

[27] Metropolis, N. 1987. The beginning of the monte-carlo method. Los Alamos Science No. 15(Special Issue): 125–130.

[28] Miller C, Ager AA. A review of recent advances in risk analysis for wildfire management. International Journal of Wildland Fire, 2013, 22(1): 1-14

[29] 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. Forest Ecology and Management, 2012, 275: 117-129

[30] Peng GF, Liu Y, Zhang YP. Research on characteristics of drought and climatic trend in Yunnan Province. Journal of Catastrophology, 2009, 24(4): 40-44 (in Chinese)

[31] Posada D, Buckley TR. Model selection and model averaging in phylogenetics: Advantages of Akaike information criterion and Bayesian approaches over likelihood ratio tests. Systematic Biology, 2004, 53(5): 793-808

[32] Robnik-Sikonja, M. Improving random forests. Paper presented at the ECML 2004, Pisa, Italy, 20–24 September 2004.

[33] Rodrigues M, de la Riva J. An insight into machine-learning algorithms to model human-caused wildfire occurrence. Environmental Modelling & Software, 2014, 57: 192-201

[34] Saha S, Moorthi S, Pan H-L, Wu XR, Wang J, Nadiga S, Tripp P, Kistler R NCEP climate forecast system reanalysis (CFSR) selected hourly time-series products, January 1979 to December 2010, 2010, Boulder, CO: Research Data Archive at the National Center for Atmospheric Research, Computational and Information Systems Laboratory

[35] SAS Institute. 2002–2003. SAS system version 9.2 for Windows. Cary, NC: SAS Institute.

[36] State Forestry Administration. 1990–2015. China forestry statistical yearbook. Beijing: China Forestry Press (in Chinese).

[37] Sun YM, Kamel MS, Wong AKC, Wang Y. Cost-sensitive boosting for classification of imbalanced data. Pattern Recognition, 2007, 40(12): 3358-3378

[38] Symonds MRE, Moussalli A. A brief guide to model selection, multimodel inference and model averaging in behavioural ecology using Akaike’s information criterion. Behavioral Ecology and Sociobiology, 2011, 65(1): 13-21

[39] Thompson JR, Spies TA. Vegetation and weather explain variation in crown damage within a large mixed-severity wildfire. Forest Ecology and Management, 2009, 258(7): 1684-1694

[40] Venables WN, Ripley BD. Modern applied statistics with S-PLUS, 2002, New York: Springer

[41] Wang, S.C. 2003. Artificial neural network. In Interdisciplinary Computing in Java Programming, 81–100. Boston, MA: Springer.

[42] Weinstein D, Woodbury P. Review of methods for developing probabilistic risk assessments. Part 1: Modeling fire. Advances in threat assessment and their application to forest and rangeland management, 2010, 2: 285-302.

[43] Ye T, Wang Y, Guo ZX, Li YJ. Factor contribution to fire occurrence, size, and burn probability in a subtropical coniferous forest in East China. PLoS ONE, 2017, 12(2): e0172110

[44] Zhang MJ, He JY, Wang BL, Wang SJ, Li SS, Liu WL, Ma XN. Extreme drought changes in Southwest China from 1960 to 2009. Journal of Geographical Sciences, 2013, 23(1): 3-16

[45] Zhang XS. Vegetation map of the People’s Republic of China (1:1000000), 2007, Beijing: Geology Press (in Chinese)

[46] Zhou XH, McClish DK, Obuchowski NA. Statistical methods in diagnostic medicine, 2009, New York: John Wiley & Sons