Model description

​

To produce leaf area index (LAI) maps at each site, we utilized the well-established PFT-specific hybrid method to retrieve LAI from Sentinel-2 imagery. The hybrid method consists of random forest models trained on simulated datasets generated by PROSAIL-5B with two refinements: variable canopy fraction of fully senescent leaves and soil bidirectional reflectance factor (BRF) simulated by Brightness-Shape-Moisture model. We corrected canopy BRF using near-infrared reflectance of vegetation and vegetation cover within mixed pixels. We also used 30-m land cover data, resampled to a 10-m resolution to match Sentinel-2 imagery, as the input for different PFTs. This land cover data is available up to 2022 and can be used for LAI retrieval in 2023, assuming minor land cover changes. The resampled data may influence the accuracy of real land cover types, particularly in heterogeneous regions.​

​

Model scripts

 

-MATLAB​

​

Corresponding to the LAI paper published in the RSE journal (HERE), we developed MATLAB codes to retrieve LAI form Sentinel-2 imagery. The details can be found in the fold of ‘MATLAB code’ (HERE). We provided the trained RF model data and examples for land cover and Sentinel-2 imagery at CRK site in South Korea.

​

-Google Earth Engine​

​

Based on MATLAB codes, we developed GEE codes to produce LAI maps. However, due to the difference between GEE and MATLAB codes (such as random forest model training), there may exist minor variations (R2 close to 0.99) in LAI maps.
For different tasks, we provide several different versions as:

(1) All LAI maps within the specified time range (site level)
https://code.earthengine.google.com/708a78ca7f237d37c57036862e06603c
For this script, it requires some time (several minutes) to run and so please wait a while.

(2) Median LAI map within the specified time range (site level)
https://code.earthengine.google.com/97091165e06c2a4165dd85dff58561ba

(3) Median LAI map within the specified time range (South Korea)
https://code.earthengine.google.com/48de846329e5bcf53fde8367a716016c

We are still working on improving the model codes. If you have any suggestions and questions about the GEE code, please feel free to contact us.

​

Data availability ​​

Sentinel-2 imagery can be downloaded from the GEE platform (COPERNICUS/S2_SR_HARMONIZED). Land cover data are available from https://doi.org/10.5194/essd-16-1353-2024. Ground measurements are derived from the GBOV dataset (https://land.copernicus.eu/global/gbov/dataaccessLP/). When you want to use the GBOV and land cover data for your research, please check the corresponding data policy.

​

Code policy

We require citing below paper whenever you use the above codes in your research.

Wan, L., Ryu, Y.*, Dechant, B., Hwang, Y., Feng, H., Kang, Y., Jeong, S., Lee, J., Choi, C., Bae, J. (2024) Correcting confounding canopy structure, biochemistry and soil background effects improves leaf area index estimates across diverse ecosystems from Sentinel-2 imagery. Remote Sensing of Environment, 2024, 309, 114224. https://doi.org/10.1016/j.rse.2024.114224
 

​

Support and more information are available from Dr. Youngryel Ryu (yryu@snu.ac.kr) and Dr. Liang Wan (liangwanzju@gmail.com).

​