Ground far-red sun-induced chlorophyll fluorescence and vegetation indices in the US Midwestern agroecosystems

Abstract

Sun-induced chlorophyll fluorescence (SIF) provides an opportunity to study terrestrial ecosystem photosynthesis dynamics. However, the current coarse spatiotemporal satellite SIF products are challenging for mechanistic interpretations of SIF signals. Long-term ground SIF and vegetation indices (VIs) are important for satellite SIF validation and mechanistic understanding of the relationship between SIF and photosynthesis when combined with leaf- and canopy-level auxiliary measurements. In this study, we present and analyze a total of 15 site-years of ground far-red SIF (SIF at 760 nm, SIF₇₆₀) and VIs datasets from soybean, corn, and miscanthus grown in the U.S. Corn Belt from 2016 to 2021. We introduce a comprehensive data processing protocol, including different retrieval methods, calibration coefficient adjustment, and nadir SIF footprint upscaling to match the eddy covariance footprint. This long-term ground far-red SIF and VIs dataset provides important and first-hand data for far-red SIF interpretation and understanding the mechanistic relationship between far-red SIF and canopy photosynthesis across various crop species and environmental conditions.

Fig. 1

Schematic layout and deployment of FluoSpe2. (a) Schematic diagram of a FluoSpe2 system; (b) Conceptual field deployment of a FluoSpe2 system. FOV: field of view.

Fig. 2

Field sites of our long-term ground measurements and some examples of field setups of FluoSpec2 systems.

Fig. 3

Flowchart of data processing at each site-year. sFLD: standard Fraunhofer line depth; 3FLD: three-band Fraunhofer line depth; iFLD: improved Fraunhofer line depth; SFM-nonlinear: spectral fitting method with the assumption of non-linear variation of fluorescence and reflectance over the absorption band; SFM-linear: spectral fitting method with the assumption of linear variation of fluorescence and reflectance over the absorption band; f_{cal-corr-QEPRO}: the calibration adjustment factor for SIF; EC: eddy covariance; $Rati{o}_{ECfootprint,SIFpixel}$: the ratio between EC footprint weighted VI and SIF tower located pixel VI.

Fig. 4

An example showing the calculation of the calibration adjustment factor for SIF₇₆₀ (f_{cal-corr-QEPRO}) at US-Ne2 2017 corn. (a) the relationship between PAR calculated from HR2000 + spectrometer and measured PAR from LiCor quantum sensor; (b) the relationship between near-infrared irradiance integrated from 730 nm calculated from QEPRO spectrometer and that from HR2000 + . Red lines are fitted linear regression lines without intercept.

Fig. 5

The variation of the calibration adjustment factor for SIF₇₆₀ (f_{cal-corr-QEPRO}) from 2016 to 2021. The first calibrated light source is used for irradiance calibration from 2016 to 2019, and the second one is used from 2020 to 2021.

Fig. 6

The diurnal variations of retrieved SIF₇₆₀ from five methods (colored lines) and enclosure temperature (black lines) at eight representative days. The upper panel represents days when enclosure temperatures are well controlled, while the bottom panel represents days when enclosure temperatures fluctuate substantially except for US-UiC 2018 when enclosure temperature is well controlled across the whole data period.

Fig. 7

The relationship between different method retrieved SIF₇₆₀ under different enclosure temperatures. The relationship between iFLD SIF₇₆₀ and sFLD SIF₇₆₀ (first row), between iFLD SIF₇₆₀ and 3FLD SIF₇₆₀ (second row), between iFLD SIF₇₆₀ and SFM-nonlinear SIF₇₆₀ (third row), and between iFLD SIF₇₆₀ and SFM-linear SIF₇₆₀ (fourth row) at US-UiC 2017 corn (first column), US-UiC 2018 corn (second column), US-Ne3 2018 soy (third column) and US-UiB 2019 Mis (fourth column). Colormap represents enclosure temperature. Black lines are 1:1 line.

Fig. 8

Seasonal variation of daytime average SIF₇₆₀ from local time 8 am to 6 pm at each site-year. Grey, blue, and red circles represent raw iFLD SIF₇₆₀, calibration corrected iFLD SIF_760, and calibration + footprint corrected iFLD SIF₇₆₀.

Fig. 9

Histogram and Gaussian kernel estimate (KDE) density of peak season half-hourly raw iFLD SIF₇₆₀ (grey), calibration corrected iFLD SIF₇₆₀ (blue) and calibration + footprint corrected SIF₇₆₀ (red) in (a) corn, (b) soybean, and (c) miscanthus.

Fig. 10

Seasonal variations of daytime average VIs from 8 am to 6 pm at each site-year. Different VIs are represented by different colours, with NDVI by grey circles, EVI by blue circles, NIRv by yellow circles, CI_rededge divided by 10 by green circles, CI_green divided by 10 by cyan circles, and PRI by red circles. CI_rededge and CI_green were divided by 10 to match the magnitude of the other VIs.

Fig. 11

Boxplot of peak season half-hourly NDVI, EVI, NIRv, CI_rededge divided by 10, CI_green divided by 10, and PRI in corn (orange), soybean (yellow), and miscanthus (green).

Fig. 12

Relationship between calibration corrected iFLD SIF₇₆₀, VI and the product of VI and PAR (VI × PAR) in corn, soybean, and miscanthus. All data available for the same species are combined for this analysis.

Fig. 13

Relationship between peak-season half-hourly APAR and calibration corrected iFLD SIF₇₆₀ in (a) corn, (b) soybean, and (c) miscanthus. APAR is calculated from VI (Rededge NDVI) in corn and soybean (APAR_VI), while APAR is measured in miscanthus (APAR_Meas).

Fig. 14

The relative importance of PAR, fPAR, f_esc, and Φ_{F, canopy} to peak season calibration-corrected iFLD SIF₇₆₀ for corn, soybean, and miscanthus calculated from the LMG method.