Unlocking vegetation health: optimizing GEDI data for accurate chlorophyll content estimation

Cuifen Xia; Wenwu Zhou; Qingtai Shu; Zaikun Wu; Mingxing Wang; Li Xu; Zhengdao Yang; Jinge Yu; Hanyue Song; Dandan Duan

doi:10.3389/fpls.2024.1492560

Unlocking vegetation health: optimizing GEDI data for accurate chlorophyll content estimation

Front Plant Sci. 2024 Nov 29:15:1492560. doi: 10.3389/fpls.2024.1492560. eCollection 2024.

Authors

Cuifen Xia¹, Wenwu Zhou², Qingtai Shu¹, Zaikun Wu¹, Mingxing Wang¹, Li Xu¹, Zhengdao Yang¹, Jinge Yu³, Hanyue Song⁴, Dandan Duan⁵

Affiliations

¹ College of Forestry, Southwest Forestry University, Kunming, China.
² Guangyuan Forestry Bureau, Guangyuan, China.
³ School of Ecology and Applied Meteorology, Nanjing University of Information Science and Technology, Nanjing, China.
⁴ College of Forestry, Fujian Agriculture and Forestry University, Fuzhou, China.
⁵ Information Technology Research Center, Beijing Academy of Agriculture and Forestry Sciences, Beijing, China.

Abstract

Chlorophyll content is a vital indicator for evaluating vegetation health and estimating productivity. This study addresses the issue of Global Ecosystem Dynamics Investigation (GEDI) data discreteness and explores its potential in estimating chlorophyll content. This study used the empirical Bayesian Kriging regression prediction (EBKRP) method to obtain the continuous distribution of GEDI spot parameters in an unknown space. Initially, 52 measured sample data were employed to screen the modeling parameters with the Pearson and RF methods. Next, the Bayesian optimization (BO) algorithm was applied to optimize the KNN regression model, RFR model, and Gradient Boosting Regression Tree (GBRT) model. These steps were taken to establish the most effective RS estimation model for chlorophyll content in Dendrocalamus giganteus (D. giganteus). The results showed that: (1) The R ² of the EBKRP method was 0.34~0.99, RMSE was 0.012~3,134.005, rRMSE was 0.011~0.854, and CRPS was 965.492~1,626.887. (2) The Pearson method selects five parameters (cover, pai, fhd_normal, rv, and rx_energy_a3) with a correlation greater than 0.37. The RF method opts for five parameters (cover, fhd_normal, sensitivity, rh100, and modis_nonvegetated) with a contribution threshold greater than 5.5%. (3) The BO-GBRT model in the RF method was used as the best estimation model (R ² = 0.86, RMSE = 0.219 g/m², rRMSE = 0.167 g/m², p = 84.13%) to estimate and map the chlorophyll content of D. giganteus in the study area. The distribution range is 0.20~2.50 g/m². The findings aligned with the distribution of D. giganteus in the experimental area, indicating the reliability of estimating forest biochemical parameters using GEDI data.

Keywords: Bayesian optimization algorithm; EBKRP method; chlorophyll content; estimation; modeling factor selection; remote sensing.

Grants and funding

The author(s) declare that financial support was received for the research, authorship, and/or publication of this article. The work was supported by the National Key Research and Development Program of China (Nos. 2023YFD2201205), the Joint Agricultural Project of Yunnan Province (Nos. 202301BD070001-002), and the National Natural Science Foundation of China (Nos. 31860205 and 31460194).