Mapping Urban Tree Species by Integrating Canopy Height Model with Multi-Temporal Sentinel-2 Data
(1) Background: Urban tree species mapping is crucial for ecosystem service evaluation and sustainable urban strategy development. However, due to the spectral similarity among dominant urban tree species, spectral data alone are insufficient for high-accuracy classification. (2) Methods: We present...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-02-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/17/5/790 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | (1) Background: Urban tree species mapping is crucial for ecosystem service evaluation and sustainable urban strategy development. However, due to the spectral similarity among dominant urban tree species, spectral data alone are insufficient for high-accuracy classification. (2) Methods: We present an approach that integrates the high-precision Canopy Height Model (CHM), generated from Ziyuan-3 (ZY3) stereo images, with multi-temporal Sentinel-2 data, for mapping 23 dominant urban tree species in Shenzhen. We primarily employed a random forest classifier using combinations of spectral bands, vegetation indices, and structural features, with subsets refined through Variance Inflation Factor (VIF) screening. We compared different models with different combinations of features, with or without the inclusion of CHM data. (3) Results: This study found that integrating VIF-screened seasonal Sentinel-2 spectral data with vegetation indices and structural metrics (Cop_DEM and ZY3_Cop_CHM) yielded an overall accuracy of 89.2%. Notably, ZY3_Cop_CHM emerged as the most influential predictor in the model. Additionally, the incorporation of ZY3_CHM data enhanced the classification accuracy by 7.1% and improved the accuracy by 4.8% compared with the use of ALOS_CHM. The species-specific F1 accuracy of a tree varies under different models and feature combinations, which underscores the need for tailored model tuning and an increase in overall model performance. Conclusions: These results indicate that integrating the ZY3_CHM data with multi-temporal Sentinel-2 data can accurately map the dominant urban tree species, suggesting its potential applicability in other urban environments. |
|---|---|
| ISSN: | 2072-4292 |