Vegetation Dynamics Forecasting Using Multispectral Time Series: A Comparative Study of Statistical, Machine Learning, and Deep Learning Models

• José Pirangaba Neto UPE
• João Fausto Lorenzato UPE
• João Vitor Ataíde UPE

DOI: https://doi.org/10.5753/eniac.2026.31516

Resumo

Due to the high data dimensionality and the availability of chaotic atmospherical data, monitoring and forecasting vegetation dynamics using multispectral pictures presents computational challenges. Using data from Sentinel-2 with more than 3.8 million registrations from January 2018 to December 2025, this paper suggests a method for spatiotemporal analysis of vegetative cover in the Usina União e Indústria, in the state of Pernambuco, Brazil. A preprocessing pipeline based on the HANTS algorithm was created for cloud gap filling and noise reduction, with regularization occurring every 15 days. A comparison study of statistical techniques (ARIMA and SARIMA), machine learning (LightGBM and XGBoost), and deep learning spatiotemporal (Conv3D and ConvLSTM) was then developed and evaluated based on a multistep forecast scenario for 2025. The findings show that machine learning and deep learning are superior, with LightGBM and XGBoost standing out for their superior ability to grasp spatial and temporal patterns. HANTS preprocessing proved crucial for improving temporal consistency and lowering problem complexity. This study’s contribution is an integrated evaluation of various predictive techniques in long multispectral time series and a demonstration of the effectiveness of tree-based machine learning models for accurate and computationally efficient vegetation monitoring using remote sensing data.

Referências

Aakash, D., K., J., and V., T. (2023). Environmental analysis using remote sensing data. In 2023 International Conference on Recent Advances in Science and Engineering Technology (ICRASET), Nagara, India. IEEE.

Atanasov, A., Petrov, P., and Yankova, R. (2025). Exploring the potential of NDVI derivatives for monitoring in precision agriculture. In 2025 International Conference Automatics and Informatics (ICAI), Varna, Bulgaria. IEEE.

Baodong, M., Lixin, W., and Shanjun, L. (2009). Analysis to Shenyang urban expansion by using multi-source remote sensing images. In 2009 Joint Urban Remote Sensing Event, Shanghai, China. IEEE.

Chang, S., Chi, Z., Chen, H. c., Hu, T., Gao, C., Meng, J., and Han, L. (2024). Development of a multiscale XGBoost-based model for enhanced detection of potato late blight using Sentinel-2, UAV, and ground data. IEEE Transactions on Geoscience and Remote Sensing, 62.

Drusch, M., Bello, U., Carlier, S., Colin, O., Fernandez, V., Gascon, F., Hoersch, B., Isola, C., Laberinti, P., Martimort, P., Meygret, A., Spoto, F., Sy, O., Marchese, F., and Bargellini, P. (2012). Sentinel-2: ESA’s optical high-resolution mission for GMES operational services. Remote Sensing of Environment, 120:25–36.

Gu, Y., Brown, J. F., Verdin, J. P., and Wardlow, B. (2007). A five-year analysis of modis ndvi and ndwi for grassland drought assessment over the central great plains of the united states. Geophysical Research Letters, 34(6).

Lillesand, T., Kiefer, R. W., and Chipman, J. (2015). Remote Sensing and Image Interpretation. John Wiley & Sons, Hoboken, NJ, USA, 7th edition.

Łochowski, P., Cieżkowski, W., and Chormański, J. (2024). Green tracks in urban landscapes: Optimizing vegetation health assessment with UAS NDVI. In IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium, Athens, Greece. IEEE.

Sivakumar, S. S. and Periasamy, S. (2024). Damage assessment of cyclone on the mangrove forest using remote sensing techniques. In 2024 IEEE Mediterranean and Middle-East Geoscience and Remote Sensing Symposium (M2GARSS), Oran, Algeria. IEEE.

Srivastava, H. and Pant, T. (2020). A time series based study of modis ndvi for vegetation cover. In 2020 IEEE India Geoscience and Remote Sensing Symposium (InGARSS). IEEE.

Wang, J., Chen, T., and He, Y. (2024). Application of satellite remote sensing and GIS technology in natural disaster risk assessment. In 2024 IEEE 2nd International Conference on Sensors, Electronics and Computer Engineering (ICSECE), Jinzhou, China. IEEE.

Wen, L., Li, P., He, Z., Ma, Z., Luo, D., Jiang, Z., and Liu, W. (2025). An algorithm for satellite remote sensing image compression based on neural network architecture evolution. In Proceedings of the 2025 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Brisbane, Australia. IEEE.

Xu, L., Cai, R., Yu, H., Du, W., Chen, Z., and Chen, N. (2024). Monthly ndvi prediction using spatial autocorrelation and nonlocal attention networks. 17.

Zhou, J., Jia, L., and Menenti, M. (2015). Reconstruction of global modis ndvi time series: Performance of harmonic analysis of time series (hants). Remote Sensing of Environment.