Published
Ensemble Deep Learning for the Classification of Strategic Crops Using Sentinel-2 Time Series
Aprendizaje profundo en conjunto para la clasificación de cultivos estratégicos utilizando series temporales de Sentinel-2
DOI:
https://doi.org/10.15446/ing.investig.121669Keywords:
crop type classification, time series, ensemble learning, TempCNN, BiLSTM, Transformer-Encoder (en)clasificacion de tipos de cultivos, series temporales, aprendizaje por conjunto, TempCNN, BiLSTM, Transformer-Encoder (es)
Downloads
Reliable and up-to-date information on crop distribution is essential for ensuring food security in the face of rapid population growth and climate change, and with recent advancements in remote sensing and artificial intelligence, large-scale analyses of agricultural landscapes can now be performed more quickly and accurately. The main objective of this study is to improve the classification of strategic crops (wheat, barley, and corn) through an ensemble learning approach that integrates three complementary neural network architectures—TempCNN, BiLSTM, and Transformer-Encoder—using Sentinel-2 time series data. To achieve this objective, several vegetation indices were first computed from the spectral bands of Sentinel-2 time series data. A Random Forest (RF) model was then employed to identify and select the most relevant features. Finally, these features were used to train the individual models, and their prediction probabilities were combined using soft voting to produce the final ensemble predictions. Model performance was evaluated across four spatial cross-validation configurations, in which the data from each of the four departments of Brittany was held out in turn as an independent test region, ensuring a rigorous assessment of the model’s spatial generalization capability. The ensemble model achieves a mean Overall Accuracy (OA) of 98.14% across all configurations, with per-class F1-scores exceeding 0.93 for barley, 0.97 for wheat, and 0.99 for corn. Compared to TempCNN and LSTM baseline models, as well as Recurrent Convolutional Neural Network (R-CNN), RF, and Support Vector Machine (SVM), the proposed ensemble model consistently achieves higher OA and per-class metrics. It also outperforms the same ensemble trained without vegetation indices. These findings highlight the effectiveness of combining complementary deep learning architectures and demonstrate the added value of vegetation indices for crop type classification using satellite time series.
La información fiable y actualizada sobre la distribución de los cultivos es esencial para garantizar la seguridad alimentaria ante el rápido crecimiento demográfico y el cambio climático. Gracias a los recientes avances en teledetección e inteligencia artificial, los análisis a gran escala de los paisajes agrícolas pueden realizarse ahora con mayor rapidez y precisión. El objetivo principal de este estudio es mejorar la clasificación de cultivos estratégicos (trigo, cebada y maíz) mediante un enfoque de aprendizaje por conjuntos (ensemble learning) que integra tres arquitecturas de redes neuronales complementarias: TempCNN, BiLSTM y Transformer-Encoder, utilizando series temporales de datos de Sentinel-2. Para alcanzar este objetivo, primero se calcularon varios índices de vegetación a partir de las bandas espectrales de las series temporales. Posteriormente, se empleo un modelo de Random Forest (RF) para identificar y seleccionar las características mas relevantes. Finalmente, estas características se utilizaron para entrenar los modelos individuales, y sus probabilidades de predicción se combinaron mediante votación blanda (soft voting) para generar las predicciones finales del conjunto. El rendimiento del modelo se evalúo mediante cuatro configuraciones de validación cruzada espacial, en las que los datos de cada uno de los cuatro departamentos de Bretaña se mantuvieron sucesivamente como región de prueba independiente, garantizando una evaluación rigurosa de la capacidad de generalización espacial del modelo. El modelo conjunto alcanza una Precisión Global (OA) media del 98.14 % en todas las configuraciones, con puntuaciones F1 por clase que superan el 0.93 para la cebada, el 0.97 para el trigo y el 0.99 para el maíz. En comparación con los modelos base TempCNN y LSTM, así como con redes neuronales convolucionales recurrentes (R-CNN), RF y maquinas de vectores de soporte (SVM), el modelo propuesto logra consistentemente métricas de OA y por clase superiores. Asimismo, supera al mismo conjunto entrenado sin índices de vegetación. Estos hallazgos resaltan la eficacia de combinar arquitecturas de aprendizaje profundo complementarias y demuestran el valor añadido de los índices de vegetación para la clasificación de tipos de cultivos a partir de series temporales satelitales.
References
[1] D. Elavarasan and P. M. D. Vincent, "Crop yield prediction using deep reinforcement learning model for sustainable agrarian applications," IEEE Access, vol. 8, pp. 86886-86901, 2020, https://doi.org/10.1109/ACCESS.2020.2992480.
[2] M. Ustuner, F. B. Sanli, S. Abdikan, M. T. Esetlili, and Y. Kurucu, "Crop type classification using vegetation indices of RapidEye imagery," Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., vol. XL-7, pp. 195-198, Sep. 2014, https://doi.org/10.5194/isprsarchives-XL-7-195-2014.
[3] A. Joshi, B. Pradhan, S. Gite, and S. Chakraborty, "Remote-sensing data and deep-learning techniques in crop mapping and yield prediction: A systematic review," Remote Sens., vol. 15, no. 8, p. 2014, Apr. 2023, https://doi.org/10.3390/rs15082014.
[4] Food and Agriculture Organization of the United Nations, The future of food and agriculture: trends and challenges. Rome: FAO, 2017. [Online]. Available: https://www.fao.org/3/i6583e/i6583e.pdf.
[5] M. O. Turkoglu, S. D’Aronco, G. Perich, F. Liebisch, C. Streit, K. Schindler and J. D. Wegner, "Crop mapping from image time series: Deep learning with multi-scale label hierarchies," Remote Sens. Environ., vol. 264, p. 112603, Oct. 2021, https://doi.org/10.1016/j.rse.2021.112603.
[6] W. F. Laurance, J. Sayer, and K. G. Cassman, "Agricultural expansion and its impacts on tropical nature," Trends Ecol. Evol., vol. 29, no. 2, pp. 107-116, Feb. 2014, https://doi.org/10.1016/j.tree.2013.12.001.
[7] B. Asadi and A. Shamsoddini, "Crop mapping through a hybrid machine learning and deep learning method," Remote Sens. Appl.: Soc. Environ., vol. 33, p. 101090, Jan. 2024, https://doi.org/10.1016/j.rsase.2023.101090.
[8] S. Meng, X. Wang, X. Hu, C. Luo, and Y. Zhong, "Deep learning-based crop mapping in the cloudy season using one-shot hyperspectral satellite imagery," Comput. Electron. Agric., vol. 186, p. 106188, Jul. 2021, https://doi.org/10.1016/j.compag.2021.106188.
[9] J. A. Foley, R. DeFries, G. P. Asner, C. Barford and G. Bonan, "Global consequences of land use," Science, vol. 309, no. 5734, pp. 570-574, Jul. 2005, https://doi.org/10.1126science.1111772.
[10] J. Xu, J. Yang, X. Xiong, H. Li, J. Huang, K. C. Ting, Y. Ying and T. Lin, "Towards interpreting multi- temporal deep learning models in crop mapping," Remote Sens. Environ., vol. 264, p. 112599, Oct. 2021, https://doi.org/10.1016/j.rse.2021.112599.
[11] L. Elmansouri, "Multiple classifier combination for crop types phenology based mapping," in IEEE Int. Conf. Adv. Technol. Signal Image Process. (ATSIP), Fez, Morocco, May 2017, pp. 1-6, https://doi.org/10.1109/ATSIP.2017.8075529.
[12] M. Alami Machichi, L. Elmansouri, Y. Imani, O. Bourja, O. Lahlou, Y. Zennayi, F. Bourzeix, I. Hanadé Houmma and R. Hadria, "Crop mapping using supervised machine learning and deep learning: A systematic literature review," Int. J. Remote Sens., vol. 44, no. 8, pp. 2717-2753, Apr. 2023, https://doi.org/10.1080/01431161.2023.2205984.
[13] A. O. Ok, O. Akar, and O. Gungor, "Evaluation of random forest method for agricultural crop classification," Eur. J. Remote Sens., vol. 45, no. 1, pp. 421-432, Jan. 2012, https://doi.org/10.5721/eujrs20124535.
[14] J. M. Peña-Barragán, M. K. Ngugi, R. E. Plant, and J. Six, "Object-based crop identification using multiple vegetation indices, textural features and crop phenology," Remote Sens. Environ., vol. 115, no. 6, pp. 1301-1316, Jun. 2011, https://doi.org/10.1016/j.rse.2011.01.009.
[15] S. Feng, J. Zhao, T. Liu, H. Zhang, Z. Zhang, and X. Guo, "Crop type identification and mapping using machine learning algorithms and Sentinel-2 time series data," IEEE J. Sel. Top. Appl. Earth Observ. Remote Sens., vol. 12, no. 9, pp. 3295-3306, Sep. 2019, https://doi.org/10.1109/JSTARS.2019.2922469.
[16] J. Yao, J. Wu, C. Xiao, Z. Zhang, and J. Li, "The classification method study of crops remote sensing with deep learning, machine learning, and Google Earth Engine," Remote Sens., vol. 14, no. 12, p. 2758, Jun. 2022, https://doi.org/10.3390/rs14122758.
[17] C. Pelletier, G. Webb, and F. Petitjean, "Temporal convolutional neural network for the classification of satellite image time series," Remote Sens., vol. 11, no. 5, p. 523, Mar. 2019, https://doi.org/10.3390/rs11050523.
[18] H. R. Khan, Z. Gillani, M. H. Jamal, A. Athar and M. T. Chaudhry, "Early identification of crop type for smallholder farming systems using deep learning on time-series Sentinel-2 imagery," Sensors, vol. 23, no. 4, pp. 1779-1779, Feb. 2023, https://doi.org/10.3390/s23041779.
[19] M. Rußwurm and M. Korner, "Temporal vegetation modelling using long short-term memory networks for crop identification from medium-resolution multi-spectralsatellite images," in IEEE Conf. Comput. Vis. Pattern Recognit. Workshops (CVPRW), Honolulu, HI, USA, Jul. 2017, pp. 1496-1504, https://doi.org/10.1109/cvprw.2017.193.
[20] F. Feng, M. Gao, R. Liu, S. Yao, and G. Yang, "A deep learning framework for crop mapping with reconstructed Sentinel-2 time series images," Comput. Electron. Agric., vol. 213, p. 108227, Sep. 2023, https://doi.org/10.1016/j.compag.2023.108227.
[21] F. Weilandt, R. Behling, R. Goncalves, A. Madadi, L. Richter, T. Sanona, D. Spengler, and J. Welsch, "Early crop classification via multi-modal satellite data fusion and
temporal attention," Remote Sens., vol. 15, no. 3, pp. 799-799, Jan. 2023, https://doi.org/10.3390/rs15030799.
[22] Y. Xu, Y. Ma, and Z. Zhang, "Self-supervised pre-training for large-scale crop mapping using Sentinel-2 time series," ISPRS J. Photogramm. Remote Sens., vol. 207, pp. 312-325, Jan. 2024, https://doi.org/10.1016/j.isprsjprs.2023.12.005.
[23] M. Rußwurm, C. Pelletier, M. Zollner, S. Lefèvre and M. Korner, "BreizhCrops: A time series dataset for crop type mapping." Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., vol. XLIII-B2-2020, 14 Aug. 2020, pp. 1545-1551, https://doi.org/10.5194/isprs-archivesx-liii-b2-2020-1545-2020.
[24] C. Eisfelder, B. Boemke, U. Gessner, P. Sogno, G. Alemu, R. Hailu, C. Mesmer and J. Huth, "Cropland and crop type classification with Sentinel-1 and Sentinel-2 time series using Google Earth Engine for agricultural monitoring in Ethiopia," Remote Sens., vol. 16, no. 5, p. 866, Jan. 2024, https://doi.org/10.3390/rs16050866.
[25] R. Tufail, P. Tassinari, and D. Torreggiani, "Assessing feature extraction, selection, and classification combi- nations for crop mapping using Sentinel-2 time series: A case study in northern Italy," Remote Sens. Appl.: Soc. Environ., vol. 38, p. 101525, Mar. 2025, https://doi.org/10.1016/j.rsase.2025.101525.
[26] P. Wei, H. Ye, S. Qiao, R. Liu, C. Nie, B. Zhang, L. Song and S. Huang, "Early crop mapping based on Sentinel- 2 time-series data and the random forest algorithm," Remote Sens., vol. 15, no. 13, pp. 3212-3212, Jun. 2023, https://doi.org/10.3390/rs15133212.
[27] Z. M. Mobarakeh, S. Pourmanafi, and M. Ahmadi, "Employing Sentinel-2 time-series and noisy data quality control enhance crop classification in arid environments: A comparison of machine learning and deep learning methods," Int. J. Appl. Earth Observ. Geoinf., vol. 142, p. 104678, Aug. 2025, https://doi.org/10.1016/j.jag.2025.104678.
[28] K. Luo, L. Lu, Y. Xie, F. Chen, F. Yin, and Q. Li, "Crop type mapping in the central part of the North China Plain using Sentinel-2 time series and machine learning," Comput. Electron. Agric., vol. 205, p. 107577, Feb. 2023, https://doi.org/10.1016/j.compag.2022.107577.
[29] H. Crisóstomo de Castro Filho, O. Abı́lio De Carvalho Júnior, O. L. Ferreira De Carvalho, P. Pozzobon De Bem, R. Dos Santos De Moura, A. Olino De Albuquerque, C. Rosa Silva, P. H. Guimarães Ferreira, R. Fontes Guimarães and R. A. Trancoso Gomes, "Rice crop detection using LSTM, Bi-LSTM, and machine learning models from Sentinel-1 time series," Remote Sens., vol. 12, no. 16, p. 2655, Aug. 2020, https://doi.org/10.3390/rs12162655.
[30] Y. Wang, Z. Zhang, L. Feng, Y. Ma, and Q. Du, "A new attention-based CNN approach for crop mapping using time series Sentinel-2 images," Comput. Electron. Agric., vol. 184, p. 106090, May 2021, https://doi.org/10.1016/j.compag.2021.106090.
[31] S. T. Seydi, M. Amani, and A. Ghorbanian, "A dual attention convolutional neural network for crop classification using time-series Sentinel-2 imagery," Remote Sens., vol. 14, no. 3, p. 498, Jan. 2022, https://doi.org/10.3390/rs14030498.
[32] V. Mazzia, A. Khaliq, and M. Chiaberge, "Improvement in land cover and crop classification based on temporal features learning from Sentinel-2 data using recurrent-convolutional neural network (R-CNN)," Appl. Sci., vol. 10, no. 1, p. 238, Dec. 2019, https://doi.org/10.3390/app10010238.
[33] H. Zhao, S. Duan, J. Liu, L. Sun, and L. Reymondin, "Evaluation of five deep learning models for crop type mapping using Sentinel-2 time series images with missing information," Remote Sens., vol. 13, no. 14, p. 2790, Jul. 2021, https://doi.org/10.3390/rs13142790.
[34] Y. Wang, L. Feng, W. Sun, L. Wang, G. Yang and B. Chen, "A lightweight CNN-Transformer network for pixel-based crop mapping using time-series Sentinel-2 imagery." Comput. Electron. Agric., vol. 226, Nov. 2024, p. 109370, https://doi.org/10.1016/j.compag.2024.109370.
[35] Y. Wang, H. Wu, J. Dong, Y. Liu, M. Long, and J. Wang, "Deep time series models: A comprehensive survey and benchmark," IEEE Trans. Pattern Anal. Mach. Intell., pp. 1-20, 2026, https://doi.org/10.1109/TPAMI.2026.3690845.
[36] A. Huete, K. Didan, T. Miura, E. P. Rodriguez, X. Gao, and L. G. Ferreira, "Overview of the radiometric and biophysical performance of the MODIS vegetation indices," Remote Sens. Environ., vol. 83, no. 1-2, pp. 195-213, Nov. 2002, https://doi.org/10.1016/s0034-4257(02)00096-2.
[37] J. W. Rouse, Jr., R. H. Haas, J. A. Schell, and D. W. Deering, "Monitoring vegetation systems in the Great Plains with ERTS," in Third Earth Resources
Technology Satellite-1 Symposium. Washington, DC, USA, Dec. 10-14, 1973, pp. 309-317. [Online]. Available: https://ntrs.nasa.gov/api/citations/19740022614/downloads/19740022614.pdf.
[38] A. A. Gitelson, "Wide dynamic range vegetation index for remote quantification of biophysical characteristics of vegetation." J. Plant Physiol., vol. 161, no. 2, Jan. 2004, pp. 165-173, https://doi.org/10.1078/0176-1617-01176.
[39] A. R. Huete, "A soil-adjusted vegetation index (SAVI)." Remote Sens. Environ., vol. 25, no. 3, Aug. 1988, pp. 295-309, https://doi.org/10.1016/0034-4257(88)90106-x.
[40] J. Qi, A. Chehbouni, A. R. Huete, Y. H. Kerr, and S. Sorooshian, "A modified soil adjusted vegetation index," Remote Sens. Environ., vol. 48, no. 2, pp. 119-126, May 1994, https://doi.org/10.1016/0034-4257(94)90134-1.
[41] G. Rondeaux, M. Steven, and F. Baret, "Optimization of soil-adjusted vegetation indices," Remote Sens. Environ., vol. 55, no. 2, pp. 95-107, Feb. 1996, https://doi.org/10.1016/0034-4257(95)00186-7.
[42] M. A. Hardisky, F. C. Daiber, C. T. Roman, and V. Klemas, "Remote sensing of biomass and annual net aerial primary productivity of a salt marsh," Remote Sens. Environ., vol. 16, no. 2, pp. 91-106, 1984, https://doi.org/10.1016/0034-4257(84)90055-5.
[43] B. Datt, "Remote sensing of water content in eucalyptus leaves," Aust. J. Bot., vol. 47, no. 6, p. 909, 1999, https://doi.org/10.1071/BT98042.
[44] E. M. Barnes, T. R. Clarke, S. E. Richards, P. D. Colaizzi, J. Haberland, M. Kostrzewski, and R. J. Lascano, "Coincident detection of crop water stress, nitrogen status and canopy density using ground based multispectral data," in Proc. 5th Int. Conf. Precis. Agric., Bloomington, MN, USA, 2000, https://www.researchgate.net/publication/ 43256762_Coincident_detection_of_crop_water_stress_nitrogen_status_and_canopy_density_using_ground_based_multispectral_data.
[45] W. J. Frampton, J. Dash, G. Watmough, and E. J. Milton, "Evaluating the capabilities of Sentinel-2 for quantitative estimation of biophysical variables in vegetation," ISPRS J. Photogramm. Remote Sens., vol. 82, pp. 83-92, 2013, https://doi.org/10.1016/j.isprsjprs.2013.04.007.
[46] J. Delegido, J. Verrelst, L. Alonso, and J. Moreno, "Evaluation of Sentinel-2 red-edge bands for empirical estimation of green LAI and chlorophyll content," Sensors, vol. 11, no. 7, pp. 7063-7081, Jul. 2011, https://doi.org/10.3390/s110707063.
[47] C. Daughtry, "Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance," Remote Sens. Environ., vol. 74, no. 2, pp. 229-239, Nov. 2000, https://doi.org/10.1016/S0034-4257(00)00113-9.
[48] A. A. Gitelson, Y. J. Kaufman, R. Stark, and D. Rundquist, "Novel algorithms for remote estimation of vegetation fraction," Remote Sens. Environ., vol. 80, no. 1, pp. 76-87, 2002, https://doi.org/10.1016/S0034-4257(01)00289-9.
[49] G. Metternicht, "Vegetation indices derived from high-resolution airborne videography for precision crop management," Int. J. Remote Sens., vol. 24, no. 14, pp. 2855-2877, 2003, https://doi.org/10.1080/01431160210163074.
[50] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, "Attention is all you need," in Adv. Neural Inf. Process. Syst., Curran Associates, Inc., 2017, https://proceedings.neurips.cc/paper_files/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf.
How to Cite
APA
ACM
ACS
ABNT
Chicago
Harvard
IEEE
MLA
Turabian
Vancouver
Download Citation
License
Copyright (c) 2026 Yacine Ait Ali Yahia, Karima Amrouche, Abir Safta

This work is licensed under a Creative Commons Attribution 4.0 International License.
The authors or holders of the copyright for each article hereby confer exclusive, limited and free authorization on the Universidad Nacional de Colombia's journal Ingeniería e Investigación concerning the aforementioned article which, once it has been evaluated and approved, will be submitted for publication, in line with the following items:
1. The version which has been corrected according to the evaluators' suggestions will be remitted and it will be made clear whether the aforementioned article is an unedited document regarding which the rights to be authorized are held and total responsibility will be assumed by the authors for the content of the work being submitted to Ingeniería e Investigación, the Universidad Nacional de Colombia and third-parties;
2. The authorization conferred on the journal will come into force from the date on which it is included in the respective volume and issue of Ingeniería e Investigación in the Open Journal Systems and on the journal's main page (https://revistas.unal.edu.co/index.php/ingeinv), as well as in different databases and indices in which the publication is indexed;
3. The authors authorize the Universidad Nacional de Colombia's journal Ingeniería e Investigación to publish the document in whatever required format (printed, digital, electronic or whatsoever known or yet to be discovered form) and authorize Ingeniería e Investigación to include the work in any indices and/or search engines deemed necessary for promoting its diffusion;
4. The authors accept that such authorization is given free of charge and they, therefore, waive any right to receive remuneration from the publication, distribution, public communication and any use whatsoever referred to in the terms of this authorization.










