Plant Traits Help Explain the Tight Relationship between Vegetation Indices and Gross Primary Production

César Hinojo-Hinojo,Michael L Goulden

doi:10.3390/rs12091405

César Hinojo-Hinojo, Michael L Goulden

Open Access

https://doi.org/10.3390/rs12091405

Copy DOI

Journal: Remote Sensing	Publication Date: Apr 29, 2020
Citations: 27	License type: CC BY 4.0

Affiliation: University of California, Irvine

Abstract

Remotely-sensed Vegetation Indices (VIs) are often tightly correlated with terrestrial ecosystem CO2 uptake (Gross Primary Production or GPP). These correlations have been exploited to infer GPP at local to global scales and over half-hour to decadal periods, though the underlying mechanisms remain incompletely understood. We used satellite remote sensing and eddy covariance observations at 10 sites across a California climate gradient to explore the relationships between GPP, the Enhanced Vegetation Index (EVI), the Normalized Difference Vegetation Index (NDVI), and the Near InfraRed Vegetation (NIRv) index. EVI and NIRv were linearly correlated with GPP across both space and time, whereas the relationship between NDVI and GPP was less general. We explored these interactions using radiative transfer and GPP models forced with in-situ plant trait and soil reflectance observations. GPP ultimately reflects the product of Leaf Area Index (LAI) and leaf level CO2 uptake (Aleaf); a VI that is sensitive mainly to LAI will lack generality across ecosystems that differ in Aleaf. EVI and NIRv showed a strong, multiplicative sensitivity to LAI and Leaf Mass per Area (LMA). LMA was correlated with Aleaf, and EVI and NIRv consequently mimic GPP’s multiplicative sensitivity to LAI and Aleaf, as mediated by LMA. NDVI was most sensitive to LAI, and was relatively insensitive to leaf properties over realistic conditions; NDVI lacked EVI and NIRv’s sensitivity to both LAI and Aleaf. These findings carry implications for understanding the limitations of current VIs for predicting GPP, and also for devising strategies to improve predictions of GPP.

Highlights

Accurate estimates of Gross Primary Production (GPP, i.e., an ecosystem’s ground-area based photosynthetic rate) are needed to better assess ecosystem function and stress, as well as the role of terrestrial ecosystems in the global carbon cycle [1]
We explored the relationships and underlying linkages between Gross Primary Production (GPP) and three commonly used Vegetation Indices: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Near InfraRed Vegetation (NIRv)
We analyzed data from 10 eddy covariance sites located in the California analogues of six major terrestrial biomes, and found that GPP, EVI and NIRv have strong and convergent within- and between-site relationships, whereas NDVI and GPP showed weaker and more site-specific relationships

Summary

Introduction

Accurate estimates of Gross Primary Production (GPP, i.e., an ecosystem’s ground-area based photosynthetic rate) are needed to better assess ecosystem function and stress, as well as the role of terrestrial ecosystems in the global carbon cycle [1]. Remote sensing provides a cost-effective strategy to extrapolate the GPP observations from eddy covariance towers to larger spatial and longer temporal scales. Recent developments using Solar-Induced Fluorescence (SIF) provide a promising approach for extrapolating GPP [4], but the satellite records required for SIF retrieval are brief and/or at coarse spatial resolution (a few years at 10 km or larger; longer at coarser resolution). Vegetation Indices (VIs) provide an alternative, simple approach to extrapolate GPP, which often performs as well as comparatively complex strategies [5,6,7,8,9]. Time series of satellite-based VI observations extend back to at least the mid-1980s, and are available at high spatial (30 m) and temporal resolution (bi-weekly to daily), providing a useful tool for exploring GPP

Methods

Results

Discussion

Conclusion