Assessing a soil-removed semi-empirical model for estimating leaf chlorophyll content

Leaf chlorophyll content (LCC) is an important indicator of foliar nitrogen status and photosynthetic capacity. Accurate and timely mapping of LCC will benefit agronomists to guide fertilizer applications and ecologists to improve carbon flux estimation. In recent decades, remote sensing techniques have been widely used to estimate LCC with empirical and physical models.    Physical models are developed based on radiative transfer models that relate LCC to observed reflectance. The common practice of empirical modeling is to establish statistical relationships between LCC and vegetation indices (VIs) based on ground measurements. A recent study developed the LAI-insensitive chlorophyll index (LICI) and established a semi-empirical LICI-based LCC quantification model, which inherits both the robustness of physical models and the simplicity of empirical models. However, it is unclear whether such a simple model is as accurate and generic as physical models.

Based on this, a group of Chinese scientists used two experimental datasets: the first one from crop fields at Rugao (120°45′ E, 32°16′ N) and Xinghua (119°53′ E, 33°05′ N), Jiangsu Province, China (denoted as crop datasets) was to obtain the ground data (LCC, canopy reflectance spectra (ASD FieldSpec 4 spectrometer)), UAV canopy spectra data and Satellite data (for sampling LCC) and then to evaluate the LCC and LAI effects on LICI with a novel disentangling approach. The second from multiple vegetation types and locations in the USA (denoted as the multi-ecosystem dataset) was to evaluate the soil effects on the estimation of LCC due to its wide range of soil background. The objectives of this study were (1) to evaluate the effects of soil background on the LCC estimation using the LICI-based semiempirical model; (2) to propose a simple but efficient algorithm to remove soil effects on the LICI-based model; and (3) to compare the performance of a soil-removed LICI-based model with a MatrixVI-based physical model on LCC estimation in terms of their accuracy and generality.

Results:

Disentangling LAI and LCC effects on LICI based on ground measurements. (A) Average LICI in each percentile bin of LAI and LCC. (B) Relationships between LICI and LCC at eight bins of LAI. (C) Relationships between LICI and LAI at eight bins of LCC. (D) Boxplots of slopes for LICI~LCC and LICI~LAI at eight bins of LAI or LCC. (E-H) Similar to (A-D) but for MTCI.

Scatter plots of measured LCC and LCC (μg/cm²) estimated from contributed reflectance of vegetation (CRv) using (A-C) the LICI-based model, (D–F) the MTCI-based model, and (G-I) the MatrixVI-based model for crop datasets at three observation platforms: ground (the first column), UAV (the second column), and satellite (the third column). RMSEs for using both canopy reflectance (Rc) and CRv are presented in text for comparison.

Scatter plots of measured LCC and LCC (μg/cm²) estimated from contributed reflectance of vegetation (CRv) using (A-H) the LICI-based model, (I–P) the MTCI-based model, and (Q-X) the MatrixVI-based model for the multi-species dataset under different levels of equivalent wet soil fraction (fs: 0–0.1, 0.1–0.2, 0.2–0.3, 0.3–0.4, 0.4–0.5, 0.5–0.6, 0.6–0.7, and 0.7–0.8).

Conclusions: 

This study assessed the performance of the LICI-based semi-empirical model on estimating leaf chlorophyll content (LCC). By disentangling the LAI and LCC effects on LICI and MTCI based on both field measurements and model simulations, the authors revealed that LICI had higher sensitivity to LCC and remarkably lower sensitivity to LAI than MTCI. Moreover, this study proposed a 3SV algorithm to reduce the soil effects on the LICI-based model and this algorithm did not require any prior information about the soil background. Compared with the MTCI-based semi-empirical model and the MatrixVI-based physical model, the soil removed LICI-based semi-empirical model yielded the best estimation accuracy for the crop datasets (RMSE = 6.22–6.87 μg/cm²) and multi-ecosystem dataset (RMSE = 10.68 μg/cm²). Due to its practicability, generality, and accuracy, the LICI-based semi-empirical model is recommended for estimating LCC after removing the soil effects using the 3SV algorithm.