Wildfire-affected Area Extraction Method Using Sentinel-2 Imagery and U-Net

Hanseong Chae; Jinmu Choi

doi:10.22776/kgs.2024.59.2.283

All Issue

2024 Vol.59, Issue 2 Preview Page Next Page

Research Article

Wildfire-affected Area Extraction Method Using Sentinel-2 Imagery and U-Net Sentinel-2 위성영상과 U-Net을 이용한 산불 피해지 추출 방법 연구

30 April 2024. pp. 283-294

PDF XML

Abstract

With the recent trend of climate warming, wildfires are expected to become more severe. It is essential to promptly and accurately extract spatial information on wildfire-affected areas for effective response, including impact assessment and facilitating recovery efforts. This study aims to train U-Net, a deep learning-based semantic segmentation model, using Sentinel-2 satellite imagery and reference data, and propose a method for extracting wildfire-affected areas based on it. To achieve this, we generated reference data through visual inspection for 303 wildfires and 353 wildfire images in Korea from 2016 to 2022, subsequently training the U-Net model. The extracted results from U-Net were then compared and evaluated for accuracy against those obtained using ISODATA, a traditional satellite image classification method. Our findings indicate that U-Net demonstrates superior accuracy in extracting wildfire-affected areas compared to ISODATA. By augmenting the training data to further enhance the U-Net model’s performance and automating the process, we can readily generate time-series spatial data of wildfire- affected areas in Korea. Such data can prove invaluable for disaster management and related research endeavors.

Keywords

Wildfire

U-Net

Sentinel-2

Semantic Segmentation

Wildfire-affected areas

최근 기후 온난화 경향이 심해짐에 따라 산불의 피해가 더욱 심해질 것으로 예상된다. 산불 피해 지역의 영향과 복구 등 효과적인 대응을 위해 산불 피해 지역의 공간 정보를 신속하고 정확하게 추출할 필요가 있다. 본 연구는 Sentinel-2 위성영상과 참조자료를 이용하여 딥러닝 기반 의미론적 분할모델인 U-Net을 훈련하고, 이를 바탕으로 산불 피해 지역을 추출하는 방법을 제시하고자 하였다. 이를 위해 2016년부터 2022년까지 국내 발생 산불 중 303건의 산불과 353장의 산불 영상에 대해 육안 판독에 기반하여 참조자료를 생성하고, 의미론적 분할모델인 U-Net을 훈련하였다. U-Net을 이용하여 추출한 결과는 전통적 위성영상 분류법 중 하나인 ISODATA 기법을 이용한 결과와 함께 정확도를 비교 및 평가하였다. 비교 결과 U-Net을 이용하는 것이 ISODATA보다 더 정확하게 산불 피해 지역을 추출하는 것으로 나타났다. U-Net 모델의 성능을 더 높일 수 있도록 학습자료를 보충하고, 일련의 과정을 자동화하면 시계열 국내 산불 피해 지역의 공간자료를 쉽게 생산할 수 있어 재난 관리 및 관련 연구에서 활용할 수 있을 것이다.

키워드

산불

U-Net

Sentinel-2

의미론적 분할모델

산불 피해 지역

References

박성욱, 2020, Lansat 위성영상과 딥러닝을 이용한 산불 피해지 탐지, 부경대학교 석사학위논문.

이수진･김광진･김영호･김지원･이양원, 2017, "Sentinel-2 영상 기반의 FBI(Fire Burn Index) 개발 및 우리나라 산불피해지역 탐지 실험," 한국사진지리학회지, 27(4), 187-202.

10.35149/jakpg.2017.27.4.012

Buda, M., Maki, A. and Mazurowski, M. A., 2018, A systematic study of the class imbalance problem in convolutional neural networks, Neural Networks, 106, 249-259.

10.1016/j.neunet.2018.07.011

Chen, W., Moriya, K., Sakai, T., Koyama, L. and Cao, C. X., 2016, Mapping a burned forest area from Landsat TM data by multiple methods, Geomatics, Natural HaZards and Risk, 7(1), 384-402.

10.1080/19475705.2014.925982

Chiaraviglio, N., Artes, T., Bocca, R., Lopez, J., Gentile, A., Ayanz, J. S. M., Cortes, A. and Margelef, T., 2016, Automatic fire perimeter determination using MODIS Hotspots information, In 2016 IEEE 12th International Conference on e-Science, 414-423.

10.1109/eScience.2016.7870928

De Bem, P. P., Carvalho J. O. A., Carvalho, O. L. F., Gomes, R. A. T. and Guimaraes, R. F., 2020, Performance Analysis of Deep Convolutional Autoencoders with Different Patch Sizes for Change Detection from Burnt Areas, Remote Sensing, 12(16), 2576.

10.3390/rs12162576

Farasin, A., Colomba, L. and Garza, P., 2020, Double-step U-Net: a deep learning-based approach for the estimation of wildfire damage severity through Sentinel-2 satellite data, Applied Sciences, 10(12), 4332.

10.3390/app10124332

Hemes, K. S., Norlen, C. A., Wang, J. A., Goulden, M. L. and Field, C. B., 2023, The magnitude and pace of photosynthetic recovery after wildfire in California ecosystems, Proceedings of the National Academy of Sciences, 120(15), e2201954120.

10.1073/pnas.220195412037011220PMC10104501

Hu, X., Ban, Y. and Nascetti, A., 2021, Uni-temporal multispectral imagery for burned area mapping with deep learning, Remote Sensing, 13(8), 1509.

10.3390/rs13081509

Jones, M. W., Abatzoglou, J. T., Veraverbeke, S., Andela, N., Lasslop, G., Forkel, M., Smith, A. J. P., Burton, C., Betts, R. A., van der Werf, G. R., Sitch, S., Canadell, J. G., Santin, C., Kolden, C., Doerr, S. H. and Quere, C. L., 2022, Global and regional trends and drivers of fire under climate change, Reviews of Geophysics, 60(3), e2020RG000726.

10.1029/2020RG000726

Knopp, L., Wieland, M., Rattich, M. and Martinis, S., 2020, A deep learning approach for burned area segmentation with Sentinel-2 data, Remote Sensing, 12(15), 2422.

10.3390/rs12152422

Lee, C. H., Park, S. Y., Kim, T. H., Liu, S., Reba, M. N. M. Oh, J. H. and Han, Y. K., 2022, Machine learning-based forest burned area detection with various input variables: a case study of South Korea, Applied Sciences, 12(19), 10077.

10.3390/app121910077

Liu, S., Zheng, Y., Dalponte, M. and Tong, X., 2020, A novel fire index-based burned area change detection approach using Landsat-8 OLI data, European Journal of Remote Sensing, 53(1), 104-112.

10.1080/22797254.2020.1738900

Lopez-Garcia, M. J. and Caselles, V., 1991, Mapping burns and natural reforestation using thematic mapper data, Geocarto International, 6(1), 31-37.

10.1080/10106049109354290

Martins, V. S., Roy, D. P., Huang, H., Boschetti, L., Zhang, H. K. and Yan, L., 2022, Deep learning high resolution burned area mapping by transfer learning from Landsat-8 to PlanetScope, Remote Sensing of Environment, 280, 113203.

10.1016/j.rse.2022.113203

Maxwell, A. E., Warner, T. A. and Guillen, L. A., 2021, Accuracy assessment in convolutional neural network- based deep learning remote sensing studies-part 1: literature review, Remote Sensing, 13(13), 2450.

10.3390/rs13132450

Muraina, I. O., 2022, Ideal dataset splitting ratios in machine learning algorithms: general concerns for data scientists and data analysis, In 7th International Mardin Artuklu Scientific Researchers Conference, Mardin, Turkey, 496-504.

Oliva, P. and Schroeder, W., 2015, Assessment of VIIRS 375m active fire detection product for direct burned area mapping, Remote Sensing of Environment, 160, 144- 155.

10.1016/j.rse.2015.01.010

Ronneberger, O., Fischer, P., Brox, T., 2015, U-Net: convolutional networks for biomedical image segmentation, in Navab, N., Hornegger, J., Wells, W. and Frangi, A.(eds), Medical Image Computing and Computer-Assisted Intervention - MICCAI 2015, MICCAI 2015, Lecture Notes in Computer Science, Vol 9351. Springer, Cham. doi: 10.1007/978-3-319-24574-4_28.

10.1007/978-3-319-24574-4_28

Seydi, S. T., Hasanlou, M. and Chanussot, J., 2022, Burnt-net: wildfire burned area mapping with single post-fire Sentinel-2 data and deep learning morphological neural network, Ecological Indicators, 140, 108999.

10.1016/j.ecolind.2022.108999

Shirvani, Z., Abdi, O. and Goodman, R. C., 2023, High-resolution semantic segmentation of woodland fires using residual attention UNet and time series of Sentinel-2, Remote Sensing, 15(5), 1342.

10.3390/rs15051342

United Nations Environment Programme, 2023, Emissions Gap Report 2023: Broken Record - Temperatures Hit New Highs, Yet World Fails to Cut Emissions (again), Nairobi. doi: 10.59117/20.500.11822/43922.

10.59117/20.500.11822/43922

Urbanski, S., 2014, Wildland fire emissions, carbon, and climate: emission factors, Forest Ecology and Management, 317, 51-60.

10.1016/j.foreco.2013.05.045

Information

Publisher :The Korean Geographical Society
Publisher(Ko) :대한지리학회
Journal Title :Journal of the Korean Geographical Society
Journal Title(Ko) :대한지리학회지
Volume : 59
No :2
Pages :283-294
Received Date : 2024-04-03
Revised Date : 2024-04-22
Accepted Date : 2024-04-24
DOI :https://doi.org/10.22776/kgs.2024.59.2.283

Journal of the Korean Geographical Society 대한지리학회지 ISSN:1225-6633(Print)

All Issue