To address the decline in predictive accuracy of existing fuel moisture content models under temporal and cross-species prediction scenarios, an incremental learning-based long short-term memory (LSTM) model is proposed. Taking three typical forest stands—Quercus mongolica, Juglans mandshurica, and Betula platyphylla—at the experimental forest of Northeast Forestry University in Harbin, Heilongjiang Province as study objects, a time-series dataset was constructed using meteorological and moisture content data collected at 20-minute intervals during the high fire-risk period in autumn 2024. The extrapolation performance of the model was evaluated from both temporal and stand-type perspectives. The results showed that, compared with the conventional LSTM model, the proposed method improved prediction accuracy by 46.54% and 47.47% under time extrapolation and cross-species prediction scenarios, respectively, with mean absolute errors (MAE) of 3.61% and 3.97%. In addition, comparisons with a supplementary random forest (RF) baseline model further supported the effectiveness of the proposed approach. The method demonstrates improved adaptability to environmental changes and enhanced predictive capability across different forest stands.
YangX G, YuY, HuH Q, et al. Moisture content estimation of forest litter based on remote sensing data[J]. Environmental Monitoring and Assessment, 2018, 190: 1-10.
HuHaiqing, LuXin, SunLong, et al. Dynamics and prediction models of ground surface dead fuel moisture content for typical stands in Great Xing’an Mountains, Northeast China[J]. Chinese Journal of Applied Ecology, 2016, 27(7): 2212-2224.
[4]
SlijepcevicA, AndersonW R, MatthewsS, et al. An analysis of the effect of aspect and vegetation type on fine fuel moisture content in eucalypt forest[J]. International Journal of Wildland Fire, 2018, 27(3): 190-202.
[5]
PausasJ G, KeeleyJ E. A burning story: The role of fire in the history of life[J]. BioScience, 2009, 59(7): 593-601.
SongYu, HuHaiqing, SunLong, et al. Dynamic change and modeling of moisture content of surface fuel in different slope positions of Daxing’anling[J]. Forest Engineering, 2017, 33(5): 1-7.
[8]
WuX, ZhangG, YangZ G, et al. Machine learning for predicting forest fire occurrence in Changsha: An innovative investigation into the introduction of a forest fuel factor[J]. Remote Sensing, 2023, 15(17): 4208.
[9]
ShmuelA, ZivY, HeifetzE. Machine-learning-based evaluation of the time-lagged effect of meteorological factors on 10-hour dead fuel moisture content[J]. Forest Ecology and Management, 2022, 505: 119897.
LiBaozhong, LiuXinyuan, YuCong, et al. Prediction on moisture content of surface dead fuels in five typical artificial forests in Harbin Urban Forestry Demonstration Base[J]. Journal of Northeast Forestry University, 2022, 50(11): 65-71.
[12]
HouX, WuZ W, ZhuS H, et al. Comparative analysis of machine learning-based predictive models for fine dead fuel moisture of subtropical forest in China[J]. Forests, 2024, 15(5): 736.
[13]
OmarN, Al-ZebariA, SengurA. Deep learning approach to predict forest fires using meteorological measurements[C]//2021 International Informatics and Software Engineering Conference (IISEC). IEEE: Ankara, Turkey, 2021: 1-4.
ZhangJiawei, ZhangSong, PengBo, et al. Prediction model of surface litter moisture content in typical forest in Harbin[J]. Forest Engineering, 2023, 39(4): 110-117.
PengBo. Study on the prediction method of moisture content of fine dead fuels on forest surface based on meteorological-spectral data fusion[D]. Harbin: Northeast Forestry University, 2024.
[18]
刘祺. 帽儿山典型林分地表死可燃物含水率测预模型研究[D]. 哈尔滨: 东北林业大学, 2021.
[19]
LiuQi. Study on the prediction model of surface dead fuel moisture content of typical forest stands in Maoershan[D]. Harbin: Northeast Forestry University, 2021.
MaoWeixing, TongDehai, ZhangCheng, et al. Spatial heterogeneity of moisture content of land surface dead fuel in larch stand and effects of sampling methods on moisture estimation[J]. Journal of Northeast Forestry University, 2012, 40(5): 29-33.
LiHaiyang, HuHaiqing, SunLong. Research on relational models of moisture content of dead forest fuel with meteorological factors and soil factors in China[J]. Forest Engineering, 2016, 32(3): 1-6.
ManZiyuan. Study on the fuel moisture code and dynamic of forest fuel moisture content in Nanchang City, Jiangxi Province[D]. Harbin: Northeast Forestry University, 2020.
[26]
ZhangY L, TianL L. Dynamic changes in moisture content and applicability analysis of a typical litter prediction model in Yunnan Province[J]. PeerJ, 2021, 9: e12206.
[27]
张佐忠. 森林可燃物含水率预测模型研究[D]. 南京: 南京林业大学, 2024.
[28]
ZhangZuozhong. Research on forest combustibles moisture content prediction models[D]. Nanjing: Nanjing Forestry University, 2024.
[29]
LiuS H, WuY D, HuangR. Prediction of drilling plug operation parameters based on incremental learning and CNN-LSTM[J]. Geoenergy Science and Engineering, 2024, 234: 212631.
[30]
KangZ Y, JiaoM, ZhouZ J. Fuel moisture content forecasting using long short-term memory (LSTM) model[C]//IGARSS 2022 IEEE International Geoscience and Remote Sensing Symposium. IEEE: Kuala Lumpur, Malaysia, 2022: 5672-5675.
[31]
LuJ, ShiY, RenZ, et al. Research on flight training prediction based on incremental online learning[J]. Applied Intelligence, 2023, 53(21):25662-25677.
[32]
LosingV, HammerB, WersingH. Incremental on-line learning: A review and comparison of state of the art algorithms[J]. Neurocomputing, 2018, 275: 1261-1274.
[33]
NWCGFire Danger Working Team. Gaining an understanding of the national fire danger rating system[R]. National Interagency Fire Center, NFES 2665, 2002.
[34]
YouJ, CelaA, NatowiczR, et al. Robust time series forecasting with non-heavy-tailed Gaussian loss-weighted sampler[J]. arXiv, 2024: arXiv: