Ensemble models enhanced estimates of leaf chlorophyll from fractional order derivatives transformed spectra

Leaf chlorophyll serves as a crucial indicator of ecosystem functioning, offering valuable insights into plant growth performance and overall status. However, accurate estimation of leaf chlorophyll content through remote sensing remains challenging due to spectral baseline drift and the lack of gen...

Full description

Saved in:
Bibliographic Details
Main Authors: Guangman Song, Yi Gan, Lingfeng Mao, Zhiwei Ge, Jiangshan Lai, Quan Wang
Format: Article
Language:English
Published: Elsevier 2025-12-01
Series:Smart Agricultural Technology
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2772375525004745
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Leaf chlorophyll serves as a crucial indicator of ecosystem functioning, offering valuable insights into plant growth performance and overall status. However, accurate estimation of leaf chlorophyll content through remote sensing remains challenging due to spectral baseline drift and the lack of generalized methods applicable over large spatial scales and diverse environmental conditions. To address these challenges, we applied the fractional order derivatives (FOD) transformed spectra into ensemble models, consisting of two popular data-oriented approaches, to refine the spectral response characteristics of leaf chlorophyll using a composite dataset (two public datasets: LOPEX and ANGERS; two collected datasets: Naeba and Nakakawane) containing different species samples. The results indicated that vegetation indices constructed from the FOD-transformed spectra performed better in estimating leaf chlorophyll a (Chla), chlorophyll b (Chlb), total chlorophyll (Chl), and chlorophyll a/b (Chla/Chlb) than published chlorophyll-related indices, demonstrating the advantage of FOD in reducing noises. Much enhanced estimates of chlorophyll parameters were achieved using the stepwise partial least squares regression (PLSR) based on FOD spectra. Further improvements were observed in the estimation of Chl (R2 = 0.80, NRMSE = 20.39 %) and Chla/Chlb (R2 = 0.73, NRMSE = 15.48 %) using the ensemble models. These results provide a thorough evaluation of FOD-transformed spectra and advanced ensemble techniques for leaf chlorophyll retrieval, providing the necessary information tailored to pigment absolute levels and functional adaptability, and advancing remote sensing applications in vegetation monitoring.
ISSN:2772-3755