Transfer learning in convolutional neural networks for fire scene classification in the Pantanal Biome (Brazil)
Submitted: 2026-02-19
|Accepted: 2026-05-11
|Published: 2026-05-27
Copyright (c) 2026 Jack Pastrana, Tássia , Pamela, Ana, Deyvis, Camila , Daniel
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Downloads
Keywords:
Convolutional Neural Networks (CNNs); Tropical ecosystems; Transfer learning; Remote sensing; forest fires.
Supporting agencies:
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)
Abstract:
During 2024, wildfires in the Pantanal, Brazil, posed a significant threat to the world’s largest wetland. This study presents a scene classification approach using Sentinel-2 imagery from fire-affected and non-affected areas, based on deep learning techniques. First, a model was experimentally trained using the EuroSAT dataset to classify grasslands and herbaceous vegetation. Subsequently, a transfer learning technique was applied to images of the fire-affected Pantanal. Finally, the VGG-19 architecture was trained from scratch, without using pretrained parameters. Considering these three experiments, training with the EuroSAT dataset achieved a loss of 0.202 after 50 epochs. The model with transfer learning, using the weights learned during the experimental training, obtained a validation loss of 0.15 and an accuracy of 95%, whereas training from scratch reached a validation loss of 0.07 and an accuracy of 97%. The results indicate similar performance between the two strategies, demonstrating that in-domain training can be considered an effective method for training CNNs applied to satellite imagery, without compromising classification accuracy, even in the presence of significant differences in the contextual information of the images. This approach is presented as a methodological alternative for environmental monitoring in complex tropical regions, characterized by high spatial and spectral heterogeneity.
References:
Alcaras, E., Costantino, D., Guastaferro, F., Parente, C., & Pepe, M. (2022). Normalized burn ratio plus (NBR+): A new index for Sentinel-2 imagery. Remote Sensing, 14(8), 1727. https://doi.org/10.3390/rs14071727
Alkan, H., Akıncı, H., & Zeybek, M. (2024). Comparison of diverse machine learning algorithms for forest fire susceptibility mapping in Antalya, Türkiye. Advances in Space Research. https://doi.org/10.1016/j.asr.2024.04.018
European Space Agency. (2025). Copernicus Sentinel-2 mission overview. European Space Agency.
Ballut-Dajud, G. A., Sandoval, C., Lambert, G., Muñiz, L., Méndez, M., & Torres, E. (2022). Factors affecting wetland loss: A review. Land, 11(3), 434. https://doi.org/10.3390/land11030434
Basu, S., Basu, K., Ghosh, S., Ganguly, S., & Mukhopadhyay, S. (2015). DeepSat: A learning framework for satellite imagery. In Proceedings of the 23rd SIGSPATIAL International Conference on Advances in Geographic Information Systems (pp. 1–10). https://doi.org/10.1145/2820783.2820812
Braga, D., Aragão, L.E.O.C., Leão, H., Osborn, F., Wagner, F.H., Cabral, B.F., Dutra, D.J., Izidoro, B., Anderson, L.O., Chaves, J.H., Pantoja, N., Saatchi, S., Dalagnol, R. 2025. Monitoring sustainable logging in an Amazon forest under management using deep learning and Planet NICFI imagery. International Journal of Remote Sensing. https://doi.org/10.1080/01431161.2025.2603696
Cho, A., Park, E., Kim, J., Kim, J., Li, C., & Song, J. (2023). Burned area mapping using unitemporal PlanetScope imagery with a deep learning-based approach. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 16, 242–253. https://doi.org/10.1109/JSTARS.2022.3225070
Chollet, F. (2017). Deep learning with Python. Manning.
Dadkhah, H., Rana, D., Ghaderpour, E., & Mazzanti, P. (2025). Analyzing wildfire patterns and climate interactions in Campania, Italy: A multi-sensor remote sensing study. Ecological Informatics, 103249. https://doi.org/10.1016/j.ecoinf.2025.103249
Dalagnol, R., Wagner, F.H., Chaves, J.H., Pantoja,N.V., Anderson, L.O., Aragão, L.E.O.C., Saatchi, S.S., Souza, C.M. 2023. Mapping tropical forest degradation with deep learning and Planet NICFIdata. Remote Sensing of Environment, 298, 113798.https://doi.org/10.1016/j.rse.2023.113798
Dong, H., Wu, H., Sun, P., & Ding, Y. (2022). Wildfire prediction model based on spatial and temporal characteristics: A case study of a wildfire in Portugal’s Montesinho Natural Park. Sustainability, 14, 10107. https://doi.org/10.3390/su141610107
Dosovitskiy, A., Fischer, P., Ilg, E., Hausser, P., Hazirbas, C., Golkov, V., van der Smagt, P., Cremers, D., & Brox, T. (2015). FlowNet: Learning optical flow with convolutional networks. In Proceedings of the IEEE International Conference on Computer Vision (pp. 2758–2766). https://doi.org/10.1109/ICCV. 2015. 316
Dutra, D. J., Silva, C. V. J., Anderson, L. O., Fonseca, M. G., Silva Junior, C. H. L., & Aragão, L. E. O. C. (2022). Fire dynamics in an emerging deforestation frontier in Southwestern Amazonia, Brazil. Fire, 6(1), 2. https://doi.org/10.3390/fire6010002
EMBRAPA. (2018). Pantanal: características gerais. https://www.embrapa.br/pantanal
Erhan, D., Bengio, Y., Courville, A., Manzagol, P.-A., Vincent, P., & Bengio, S. (2010). Why does unsupervised pre-training help deep learning? In Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics (AISTATS) (pp. 201–208). https://doi.org/10.48550/arXiv.1001.0495
Garcia, G., Laparra, V., Lopez, D., & Gómez, L. (2020). Cross-sensor adversarial domain adaptation of Landsat-8 and Proba-V images for cloud detection. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 747–761. https://doi.org/10.1109/JSTARS.2020.3031741
Gitelson, A.A., Dall’Olmo, G., Moses, W., Rundquist, D.C., Barrow, T., Fisher, T.R., Gurlin, D., Holz, J. 2008. A simple semi-analytical model for remote estimation of chlorophyll-a in turbid waters: Validation. Remote Sensing of Environment,112(9), 3582–3593. https://doi.org/10.1016/j.rse.2008.04.015
Ghali, R., & Akhloufi, M. A. (2023). Deep learning approaches for wildland fires using satellite remote sensing data: Detection, mapping, and prediction. Fire, 6, 192. https://doi.org/10.3390/fire6050192
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT Press.
Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., & Moore, R. (2017). Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment, 202, 18–27. https://doi.org/10.1016/j.rse.2017.06.031
Guo, Y., Liu, Y., Oerlemans, A., Lao, S., Wu, S., & Lew, M. S. (2016). Deep learning for visual understanding. Neurocomputing, 187, 27–48. https://doi.org/10.1016/j.neucom.2015.09.116
Helber, P., Bischke, B., Dengel, A., Borth, D.2019. EuroSAT: A novel dataset and deep learning benchmark for land use and land cover
classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,12(7), 2217–2226. https://doi.org/10.1109/JSTARS.2019.2918242
Hu, X., Li, S., Xu, J., Zhang, Y., Huang, W. 2021. Unitemporal multispectral imagery for burned area mapping with deep learning. Remote Sensing, 13,1509. https://doi.org/10.3390/rs13081509
Hinton, G., Srivastava, N., & Swersky, K. (2012). Neural networks for machine learning – Lecture 6a: Overview of mini-batch gradient descent. University of Toronto.
Huang, J., Rathod, V., Sun, C., Zhu, M., Korattikara, A., Fathi, A., & Murphy, K. (2017). Accuracy trade-offs for modern convolutional object detectors. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 7310–7311). https://doi.org/10.1109/CVPR.2017.351
Instituto Brasileiro de Geografia e Estatística (IBGE).2022. Brasil em síntese: território. IBGE, 2022.Accesible at: https://brasilemsintese.ibge.gov.br/territorio.html.
Kingma, D. P., & Ba, J. (2015). Adam: A method for stochastic optimization. In International Conference on Learning Representations (ICLR).
Klemas, V. 2011. Remote sensing of wetlands: case studies comparing practical techniques. Journal of Coastal Research, 27(3), 418-427. https://doi.org/10.2112/JCOASTRES-D-10-00174.1
Krizhevsky, A., Sutskever, I., Hinton, G.E. 2012.Image Net classification with deep convolutional neural networks. Advances in Neural InformationProcessing Systems, 25, 1097–1105.
Li, W., Sun, K., & Wei, J. (2025). Adapting cross-sensor high-resolution remote sensing imagery for land use classification. Remote Sensing, 17(5), 927. https://doi.org/10.3390/rs17050927
Lee, S. H., Lee, M. H., Kang, T. H., Cho, H. R., Yun, H. S., & Lee, S. J. (2025). Comparative analysis of dNBR, dNDVI, SVM kernels, and ISODATA for wildfire-burned area mapping using Sentinel-2 imagery. Remote Sensing. https://doi.org/10.3390/rs17132196
Libonati, R., DaCamara, C. C., Peres, L. F., Sander de Carvalho, L. A., & Garcia, L. C. (2020). Rescue Brazil’s burning Pantanal wetlands. Nature, 588(7837), 217–219. https://doi.org/10.1038/d41586-020-03464-1
Lundberg, S. M., & Lee, S. (2017). A unified approach to interpreting model prediction. In Proceedings of the 31st Conference on Neural Information Processing Systems (NeurIPS 2017).
Priya, R. S., & Vani, K. (2019). Deep learning-based forest fire classification and detection in satellite images. In 2019 11th International Conference on Advanced Computing (ICoAC) (pp. 61–65). https://doi.org/10.1109/ICoAC48881.2019.8986326
Mataveli, G. A. V., Pereira de Oliveira, G., Seixas, H. T., Cardozo, F. D. S., Shimabukuro, Y. E., & Brunsell, N. A. (2021). Pantanal widespread fire: Short- and long-term implications for biodiversity and conservation. Biodiversity and Conservation, 30(11), 3299–3303. https://doi.org/10.1007/s10531-021-02243-2
MapBiomas Project. (2024). MapBiomas Collection 8: Annual land use and land cover maps of Brazil (1985–2022). https://doi.org/10.58053/MapBiomas/IBQPF6
Md Jelas, I., Abu Bakar, M.A., Mahmud, M.R., Yusof,K.W., Sulaiman, N. 2024. Deforestation detectionusing deep learning-based semantic segmentationtechniques: a systematic review. Frontiers inForests and Global Change, 7, 1300060. https://doi.org/10.3389/ffgc.2024.1300060
Instituto Nacional de Pesquisas Espaciais (INPE). 2024.Queimadas program in Brazil. https://terrabrasilis.dpi.inpe.br/queimadas/portal/
Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. (2014). Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15, 1929–1958.
Mojica, J. P., Bernini, H., de Faria, D. G., Hauck Antunes, E., Siqueira, L., Souza Silva, C., & Rodovalho, F. (2023). Spatialization of fire occurrences in Brazil within the context of the fire events detected by the fire panel. RAEGA – O Espaço Geográfico em Análise, 60, 131–158. https://doi.org/10.5380/raega.v60i0.95566
Pazare, H., Raut, P., Dhanorkar, M., & Dhage, S. (2022). Forest fire prediction using deep learning. Forest, 2(1).
Pérez, D., Risco Ramos, R., & Casaverde, L. (2021). Transfer learning for binary classification of thermal images. Ingenius. Revista de Ciencia y Tecnología, 26, 71–86.
Pereira, A. D. M. M., Oliveira, M. D. R., Bao, F., Souza, E. B., Pott, A., Escobar, A. C. D. S., & Damasceno, G. A. (2024). Changes, trends, and gaps in research dynamics after the megafires in the Pantanal. Environment, Development and Sustainability, 1–15.
Santiago Júnior, V. A. (2024). Empirical evidence regarding few-shot learning for scene classification in remote sensing images. Applied Sciences, 14(23), 10776. https://doi.org/10.3390/app142310776
Sacramento, I.F., Michel, R.F.M., Siqueira, R.G. 2020. Bitemporal analysis of burned areas in the Atlantic Forest. Sociedade & Natureza, 32, 540 – 552.https://doi.org/10.14393/SN-v32-2020-53339
Santiago Júnior, V.A. 2024. Empirical evidence regarding few-shot learning for scene classification in remote sensing images. Applied Sciences, 14(23),10776. https://doi.org/10.3390/app142310776
Saha, S., Zhao, S., Shahzad, M., & Zhu, X. X. (2022). Mitigating distribution shift for multi-sensor classification. In IGARSS 2022 – IEEE International Geoscience and Remote Sensing Symposium (pp. 1201–1204). IEEE.
Sausen, T. M. (2007). Sensoriamento remoto e suas aplicações para recursos naturais. INPE.
Silva, I. T., Pessi, D. D., da Silva, N. M., Mioto, C. L., Guaraldo, E., & Paranhos Filho, A. C. (2021). Técnicas de processamento de imagens para treinamento de fotointerpretes na avaliação visual de áreas úmidas e de manchas e cicatrizes de queimadas ou incêndios. Revista de Geociências do Nordeste, 57–67. https://doi.org/10.21680/2447-3359.2021v7n2ID23310
Simonyan, K., & Zisserman, A. (2015). Very deep convolutional networks for large-scale image recognition. arXiv. https://arxiv.org/abs/1409.1556
Simioni, J. P. D., Guasselli, L. A., Belloli, T. F., & Ramos, R. A. (2022). Assessment of fire resilience in subtropical wetlands using high spatial resolution images. Environmental Monitoring and Assessment, 194(6), 417. https://doi.org/10.1007/s10661-022-09841-5
Shimabukuro, Y. E., & Smith, J. A.. (1991). The least-squares mixing models to generate fraction images derived from remote sensing multispectral data. IEEE Transactions on Geoscience and Remote Sensing, 29(1), 16–20. https://doi.org/10.1109/36.97458
Tomas, W.M., Berlinck, C.N., Chiaravalloti, R.M.,Faggioni, G.P., Strüssmann, C., Libonati, R.,Abrahão, C., Alvarenga, G.V., Morato, R.G. 2021.Distance sampling surveys reveal 17 million vertebrates directly killed by the 2020’s wildfires in the Pantanal, Brazil. Scientific Reports, 11, 23547.https://doi.org/10.1038/s41598-021-02844-5
Singh, S.S., Jeganathan, C. 2024. Using ensemble machine learning algorithm to predict forest fire occurrence probability in Madhya Pradesh and Chhattisgarh, India. Advances in Space Research, 73(6), 2969–2987. https://doi.org/10.1016/j.asr.2024.01.014
Unnikrishnan, A., Reddy, C.S. 2020. Characterizing distribution of forest fires in Myanmar using earth observations and spatial statistics tool. Journal of the Indian Society of Remote Sensing, 48(2), 227–234.https://doi.org/10.1007/s12524-019-01023-1
Viganó, H.H. da G., Souza, C.C. de, Cristaldo, M.F.,Jesus, L. de. 2017. Redes neurais artificiais na previsão de queimadas e incêndios no Pantanal. Revista Brasileira de Geografia Física, 10(5),1355–1367. https://doi.org/10.26848/rbgf.v.10.p1355-1367
Wahab, M.A.A., Abdullah, A.F., Shafri, H.Z.M.,Hamedianfar, A., Yusoff, A.R. 2021. Mapping deforestation in permanent forest reserve of
Peninsular Malaysia with multi-temporal SAR imagery and U-Net based semantic segmentation.Remote Sensing, 13, 2430.https://doi.org/10.3390/rs13132430
Yandouzi, M., Grari, M., Idrissi, I., Boukabous, M.,Moussaoui, O.,Azizi, M. 2022. Forest fires detection using deep transfer learning. International Journal of Advanced Computer Science and Applications,13(8), 268–275. https://doi.org/10.14569/IJACSA.2022.0130832
Yin, X., Chen, W., Wu, X., Yue, H. 2017. Fine-tuning and visualization of convolutional neural networks.In 2017 12th IEEE conference on industrial electronics and applications (ICIEA), 1310-1315.https://doi.org/10.1109/ICIEA.2017.8283041
Zhou, W., Newsam, S., Li, C., Shao, Z. 2018. A benchmark dataset for performance evaluation of remote sensing image retrieval. ISPRS Journal of Photogrammetry and Remote Sensing, 145, 197- 209. https://doi.org/10.1016/j.isprsjprs.2018.01.004
