In Press (Winter)
Back to the articles list |
Back to browse issues page
Department of Natural Resources Engineering, Faculty of Agriculture and Natural Resources, University of Hormozgan, Bandar-Abbas, Hormozgan, Iran. (Corresponding author) , a.r.nafarzadegan@hormozgan.ac.ir
Abstract: (622 Views)
1- Introduction
Vegetation cover is a fundamental component of ecosystems, covering a large portion of the Earth's surface and playing a crucial role in biogeochemical and hydrological cycles. It is closely linked with soil, water, and the atmosphere, while also providing essential ecosystem services such as carbon sequestration, water and soil conservation, air purification, and biodiversity preservation. Monitoring vegetation dynamics over time is therefore vital for assessing ecosystem health and sustainability. Land use and land cover (LULC) analyses are also critical in environmental studies, reflecting both human activities and natural processes that significantly influence biodiversity, environmental quality, and sustainable development. Rapid urban expansion, agricultural changes, and deforestation highlight the need for accurate and up-to-date LULC data to support effective resource management and planning. Recent advances in remote sensing and geographic information systems, combined with machine learning algorithms such as Support Vector Machine (SVM) and Random Forest (RF), have enhanced the precision of LULC classification and change detection. Earth observation satellite imagery, including Landsat and Sentinel-2, provides extensive spatial and temporal data for monitoring vegetation cover changes over large and inaccessible areas. This study focuses on the Takht–Qaleh Qazi watershed, an important agricultural region in Hormozgan Province, with the aim of analyzing spatiotemporal land use and vegetation cover changes using spectral indices (BSI, DSWI, EVI, GNDVI, MSI, NDMI, NDVI, SAVI) and machine learning classification models. The research seeks to address the lack of up-to-date environmental monitoring in this area and to provide valuable insights that support sustainable natural resource management.
2- Material and Methods
This study aimed to investigate land use changes and temporal vegetation dynamics in the Takht–Qaleh Qazi watershed over the period 2014 to 2024. The research process was designed in two main stages: The first stage involved generating classified land use maps for the target years (2014 and 2024) using satellite imagery and machine learning algorithms. The second stage consisted of analyzing temporal vegetation changes using spectral indices and multi-temporal Landsat and Sentinel-2 images within the Google Earth Engine environment. Data processing, spectral index calculation, and machine learning implementations were performed using the R programming language with specialized packages including caret, e1071, randomForest, raster, rasterVis, and Boruta. Satellite images were loaded based on required spectral bands and clipped to the study area using spatial boundary layers to retain relevant data only. Training and validation sample points were imported as separate spatial layers into R. Spectral indices such as NDVI, DSWI, and NDMI were calculated for each image and integrated with band data to create a comprehensive dataset for sample points. To reduce dimensionality and remove highly correlated variables, feature selection was conducted in two steps: Pearson correlation analysis via SPSS was first used to exclude highly correlated variables, followed by the Boruta algorithm in R to select the most important variables for modeling. Classification modeling was performed using Support Vector Machine (SVM) and Random Forest (RF) algorithms. Models were trained on training data and validated with evaluation data. Performance was assessed using appropriate statistical metrics. Finally, trained models were applied to the entire study area to produce and save final land use/land cover classification maps.
3- Results
This study evaluated the performance of two machine learning algorithms, Support Vector Machine (SVM) and Random Forest (RF), for land use classification and spatiotemporal analysisin the Takht–Ghaleh Ghazi watershed from 2014 to 2024. Both models showed improved classification accuracy over time, with SVM achieving superior results, particularly in identifying agricultural and orchard lands. By 2024, SVM accuracy reached 83% (Kappa = 0.75), outperforming RF’s 78% accuracy (Kappa = 0.67). Analysis of land use changes revealed a decrease in agricultural land and water bodies, accompanied by increases in orchards and barren lands, reflecting dynamic land management and environmental pressures in the region. Vegetation health was assessed through spectral indices (EVI, NDVI, SAVI) derived from Landsat 8 and Sentinel-2 imagery. These indices peaked in 2020, indicating optimal vegetation density and vigor, followed by a declining trend attributed to factors such as urban expansion, land use change, and climate variability. Although some recovery was noted post-2020, vegetation conditions have yet to return to their peak state. The results demonstrate that machine learning models can effectively classify different kinds of land use over the study area and that satellite-derived spectral indices provide reliable temporal insights into vegetation dynamics. These findings underscore the potential of integrating advanced remote sensing and machine learning techniques for sustainable natural resource management and environmental monitoring.
4- Discussion & Conclusions
The results derived from vegetation indices (NDVI, EVI, and SAVI) extracted from Landsat 8-9 and Sentinel-2 satellite data revealed consistent trends in vegetation changes over the study periods. Despite some differences in absolute index values due to sensor and processing variations, both datasets showed synchronized increasing and decreasing patterns. The year 2020 was identified as the peak of vegetation density and health, with the highest recorded index values. Subsequent declines may be attributed to climatic factors, natural environmental changes, or human activities such as unsustainable resource exploitation. Land use change analysis using SVM and RF machine learning models on Landsat data from 2014 to 2024 demonstrated significant improvements in classification accuracy by 2024. The SVM model achieved an overall accuracy of 83% and a Kappa coefficient of 0.75, outperforming the RF model which reached 78% accuracy and 0.67 Kappa. This improvement aligns with findings from previous studies highlighting the impact of enhanced data quality and algorithm refinement.
SVM showed particularly high accuracy in classifying agricultural lands, orchards, and water bodies, with balanced accuracy reaching 100% for orchards in 2024. RF performed well in identifying built-up areas in 2014 and orchards in 2024. These results are consistent with similar research but also highlight that model performance can vary depending on regional, temporal, and data-specific factors. Overall, this study emphasizes that combining multiple satellite data sources with advanced classification algorithms enhances the accuracy and reliability of environmental change assessments. The SVM algorithm, in particular, proves highly effective for detailed land use classification and, alongside RF, provides a robust toolset for monitoring land cover dynamics.
Received: 2025/07/13
References
1. Arellano, P., Tansey, K., Balzter, H., & Boyd, D. S. (2015). Detecting the effects of hydrocarbon pollution in the Amazon forest using hyperspectral satellite images. Environmental Pollution, 205, 225-239. [DOI:10.1016/j.envpol.2015.05.041]
2. Bishop, C. M. (2006). Pattern recognition and machine learning. Springer.
3. Caputo, J. (2009). Sustainable forest biomass: Promoting renewable energy and forest stewardship. Washington, DC, USA: Environmental and Energy Study Institute.
4. FAO (Food and Agriculture Organization of the United Nations). (2002). Land use. https://www.fao.org
5. Galvao, L. S., Formaggio, A. R., & Tisot, D. A. (2005). Discrimination of sugarcane varieties in Southeastern Brazil with EO-1 Hyperion data. Remote sensing of Environment, 94(4), 523-534. [DOI:10.1016/j.rse.2004.11.012]
6. Gao, B. C. (1996). NDWI-A normalized difference water index for remote sensing of vegetation liquid water from space. Remote sensing of environment, 58(3), 257-266. [DOI:10.1016/S0034-4257(96)00067-3]
7. García Cárdenas, D. A., Ramón Valencia, J. A., Alzate Velásquez, D. F., & Palacios Gonzalez, J. R. (2018, November). Dynamics of the indices NDVI and GNDVI in a rice growing in its reproduction phase from multi-spectral aerial images taken by drones. In International Conference of ICT for Adapting Agriculture to Climate Change (pp. 106-119). Cham: Springer International Publishing. [DOI:10.1007/978-3-030-04447-3_7]
8. Hersperger, A. M., Oliveira, E., Pagliarin, S., Palka, G., Verburg, P., Bolliger, J., & Grădinaru, S. (2018). Urban land-use change: The role of strategic spatial planning. Global Environmental Change, 51, 32-42. [DOI:10.1016/j.gloenvcha.2018.05.001]
9. Hormozgan Regional Water Company, 2022
10. Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote sensing of environment, 25(3), 295-309. [DOI:10.1016/0034-4257(88)90106-X]
11. Huete, A., Didan, K., Miura, T., Rodriguez, E. P., Gao, X., & Ferreira, L. G. (2002). Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote sensing of environment, 83(1-2), 195-213. [DOI:10.1016/S0034-4257(02)00096-2]
12. Joulaei, H., & Vafaeinejad, A. (2023). Application of machine learning methods for classification of Landsat 9 satellite imagery to assess urban land use area (Case study: West of Tehran). Journal of Remote Sensing and Spatial Information Research, 2(1), 113-126. (in Persian)
13. Kazemi, M., & Jafarpoor, A. (2024). Time series classification of land use using spectral indices, Sentinel-2 imagery, and variable training samples in Google Earth Engine (GEE). Iranian Journal of Watershed Science and Engineering, 18(67), 1-15. (in Persian)
14. Lawrence, R. L., Wood, S. D., & Sheley, R. L. (2006). Mapping invasive plants using hyperspectral imagery and Breiman Cutler classifications (RandomForest). Remote Sensing of Environment, 100(3), 356-362. [DOI:10.1016/j.rse.2005.10.014]
15. Maitima, J. M., Mugatha, S. M., Reid, R. S., Gachimbi, L. N., Majule, A., Lyaruu, H., ... & Mugisha, S. (2009). The linkages between land use change, land degradation and biodiversity across East Africa. African Journal of Environmental Science and Technology, 3(10).
16. Mariye, M., Mariyo, M., Changming, Y., Teffera, Z. L., & Weldegebrial, B. (2022). Effects of land use and land cover change on soil erosion potential in Berhe district: A case study of Legedadi watershed, Ethiopia. International Journal of River Basin Management, 20(1), 79-91. [DOI:10.1080/15715124.2020.1767636]
17. Milton, E. J. (1989). On the suitability of Kodak neutral test cards as reflectance standards. International Journal of Remote Sensing, 10(6), 1041-1047. [DOI:10.1080/01431168908903943]
18. Rong, C., & Fu, W. (2023). A comprehensive review of land use and land cover change based on knowledge graph and bibliometric analyses. Land use, 12(8), 1573. [DOI:10.3390/land12081573]
19. Sheram, K. (1993). The environmental data book: A guide to statistics on the environment and development. Washington, DC: The World Bank.
20. Siasar, H., Salari, A., Bahrami, M., & Hamidifar, H. (2025). Integrating remote sensing and meteorological analysis for monitoring drought conditions in arid regions: a case study from Sistan and Baluchestan province, Iran. Theoretical and Applied Climatology, 156(5), 291. [DOI:10.1007/s00704-025-05527-7]
21. Sivakumar, M. V. K. (2003). Satellite remote sensing and GIS applications in agricultural meteorology. Agricultural and Forest Meteorology, 118(1-2), 1-5.
22. Tucker, C. J. (1979). Red and photographic infrared linear combinations for monitoring vegetation. Remote sensing of Environment, 8(2), 127-150. [DOI:10.1016/0034-4257(79)90013-0]
23. Veraverbeke, S., Gitas, I., Katagis, T., Polychronaki, A., Somers, B., & Goossens, R. (2012). Assessing post-fire vegetation recovery using red-near infrared vegetation indices: Accounting for background and vegetation variability. ISPRS Journal of Photogrammetry and Remote Sensing, 68, 28-39. [DOI:10.1016/j.isprsjprs.2011.12.007]
24. Wang, G., Han, L., Tang, X. Y., & Jin, Z. C. (2012). Temporal and spatial variation of vegetation in the Jinsha River basin. Resources and Environment in the Yangtze Basin, 21(10), 1191-1196.
25. Zhou, Y., Zhang, L., Fensholt, R., Wang, K., Vitkovskaya, I., & Tian, F. (2015). Climate contributions to vegetation variations in central Asian drylands: Pre- and post-USSR collapse. Remote Sensing, 7(3), 2449-2470. [DOI:10.3390/rs70302449]
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
