Quantifying Uncertainty in Spectral Leaf Chlorophyll Content Retrieval: A Propagation Analysis

Abstract

Reliable estimation of leaf chlorophyll content (LCC) from spectral measurements requires explicit quantification of measurement and model uncertainties. This study presents an end-to-end uncertainty propagation analysis for leaf-level LCC retrieval using the automated radiative transfer models operator. A controlled laboratory dataset of winter wheat leaves was acquired with an ASD FieldSpec 4 spectroradiometer and a CCM-300 chlorophyll meter. Measurement uncertainties from sensor noise, reclipping effects and reference-panel characterization were quantified and propagated using the Community Metrology toolkit. Two machine-learning regression algorithms - Gaussian process regression (GPR) and kernel ridge regression (KRR) - were trained using principal component analysis for dimensionality reduction, and epistemic uncertainties were estimated analytically (GPR) or by bootstrapping (KRR). Results demonstrate that epistemic uncertainty dominates total uncertainty, while the contribution of measurement uncertainty remains comparatively small (≈0.2% for reflectance; ≈5% for chlorophyll meter readings). GPR produced smooth and coherent uncertainty estimates, whereas KRR exhibited higher sensitivity to sampling effects. The workflow provides reproducible uncertainty budgets and correlation structures, and is transferable to canopy- or satellite-based retrievals, supporting metrologically consistent uncertainty propagation in forthcoming hyperspectral missions.