Document Type : Original Research Paper
Authors
1 Department of Surveying Engineering, Faculty of Civil Engineering, Shahid Rajaei Teacher Training University. Tehran, Iran
2 Department of Surveying Engineering, Faculty of Civil, Water and Environmental Engineering, Shahid Beheshti University, Tehran, Iran
Abstract
Background and Objectives: Object recognition is a widely discussed topic across various disciplines. However, identifying tree species in forests remains challenging due to their similar appearances and behaviors. This study aims to address this issue by leveraging the temporal-spectral signature model in Google Earth Engine (GEE) to differentiate forest plant species. The innovation of this research lies in determining the temporal-spectral behavior of tree species by calculating the brightness of satellite image bands across different months of the year, creating a unique matrix for each species.
Methods: The study utilized Landsat 8 and 9 satellite imagery from 2016 to 2022, focusing on a section of the southern forests of Gilan province. Initially, spectral behavior curves for vegetation, land, and water were plotted. By coding in GEE, the average brightness values for each band were calculated, producing combined bar and line graphs for the three categories. Temporal-spectral signatures for tree species were then developed using typology maps and field surveys, with 200 data points collected for oak, hornbeam, beech, alder, and Bergan needle species. A matrix of 84×1 was formed, representing the temporal-spectral signature for each species, using Bands 1 to 7 of Landsat 8 across 12 months. MATLAB was employed to visualize the generated matrices.
Findings: The results revealed distinct brightness levels in specific bands and months for different species. For instance, in the first band during the second month, brightness values for oak, hornbeam, beech, Bergan needle, and alder were 0.38, 0.31, 0.27, 0.46, and 0.25, respectively. The highest brightness levels for most species occurred in the fifth band during the tenth month. Classification using the random forest method with both 7-band and 84-band inputs showed that the innovative temporal-spectral signature approach improved the Kappa coefficient to approximately 0.4. This unique signature enables the accurate identification and differentiation of tree species, supported by field observations.
Conclusion: The study demonstrates that temporal-spectral signatures can effectively differentiate tree species in forests, facilitating improved classification and monitoring. This approach holds potential for broader application to other species, paving the way for advanced forest management and monitoring by organizations such as natural resources and environmental agencies. Future research should extend this method to additional species to further enhance forest classification systems.
Keywords
Main Subjects
COPYRIGHTS
© 2024 The Author(s). This is an open-access article distributed under the terms and conditions of the Creative Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) (https://creativecommons.org/licenses/by-nc/4.0/)
)