Emerging approaches to measure photosynthesis from the leaf to the ecosystem

The terrestrial biosphere consists of an assemblage of diverse ecosystems. Its complexity is illustrated with a diversity of plants with distinct canopy structures subject to changing environmental conditions. Life on earth relies on the energy captured by these ecosystems through photosynthesis, which accounts for the single largest flux associated with the global carbon cycle [1]. Photosynthesis varies among plant functional types (e.g. C3 vs. C4) and over a wide range of spatial and temporal scales associated with changes in light, temperature, water and nutrients [2,3]. Global climate change driven by anthropogenic activities is having profound impacts on terrestrial ecosystems, with global temperatures rising faster than worst-case predictions [4]. Increasing agricultural demands associated with a growing population requires a doubling of crop yields by 2050 to keep up with demands [5], yet current rates of improvement fall short of this goal [6,7], which is likely to suffer with continued global warming [8–11].

Photosynthesis is a highly complex and relatively inefficient process, yet it is a critical component of the biosphere. Understanding photosynthetic responses over a range of spatial and temporal scales is needed to understand current and to predict future global carbon cycling. This understanding will also lead to improving photosynthesis, which can lead to higher productivity to meet growing agricultural demands [12,13]. These goals can only be achieved through the ability to measure photosynthesis over time and space, yet photosynthesis is difficult to measure directly. This is due to the multiple processes that are represented by the exchange of CO₂ between plants/ecosystems and the surrounding air. For example, at the leaf scale CO₂ is removed from the air by photosynthesis but this is partially countered by photorespiration and respiration, both of which release CO₂ [14,15]. The combined fluxes of these three processes represents net carbon assimilation (A) and partitioning this net flux into the component fluxes is challenging [16]. Scaling beyond the leaf only presents additional challenges. At canopy or ecosystem scales, respiration from non-photosynthetic tissues and heterotrophic organisms also release CO₂, which combined with A provide measures of Net Ecosystem Exchange (NEE; Table 1). In this review, we outline the current and emerging approaches to measure photosynthesis at multiple scales and address the challenges and opportunities at each scale (Figure 1). We begin with a focus on the well-established and widely used gas exchange techniques and follow with more recent approaches made available through recent technological advances.

The fundamentals of gas exchange at any scale are relatively similar and require the ability to measure gas concentrations in air surrounding and the flow rate in which the air interacts with photosynthetic tissue. In addition to these measurements, numerous assumptions, corrections, and parameterizations are required to fully exploit the power of this technique [16,17]. Gas exchange methods have been applied at scales ranging from the organelle (e.g. [18,19]) to the whole ecosystem/region [20] to provide a basic understanding of how leaves, plants, and ecosystems function and respond to their environment (Figure 1). Historically, gas exchange measurements were limited to enclosed sampling chambers, ranging from sections of leaves to whole plant canopies, where the rate of CO₂ exchange was measured over time. With the advent of micrometeorological techniques, gas exchange measurements at large scales (e.g. whole ecosystems) were developed that removed the need for enclosures (Table 2). Despite errors, uncertainties and challenges associated with gas exchange, the various techniques are the current ‘gold standard’ by which emerging techniques are compared. This section provides an overview of gas exchange measurements at the leaf to ecosystem scales as a baseline in the understanding of emerging techniques.

Knowledge of leaf photosynthetic physiology stems from the development and application of leaf-level gas exchange systems [16]. Gas exchange technology has matured to the point where commercial gas exchange systems are widely available from many vendors. In addition to providing the key variables necessary to assess leaf scale carbon assimilation, these systems now provide the opportunity to precisely control the environmental conditions surrounding the photosynthetic tissue and to measure more than just carbon assimilation, including but not limited to transpiration, intercellular CO₂ concentrations, and stomatal conductance. Gas exchange techniques have been used for decades and most recent advancements have focused on improvements in accuracy, precision, usability, environmental control, and reduction in time to stable measurements. Despite the ease with which leaf level gas exchange can be measured, the importance of understanding gas exchange theory to ensure proper measurement and analysis cannot be overstated.

Gas exchange systems are the most commonly utilized technique for leaf scale photosynthetic measurements. While systems provide measures of A, various techniques can be applied to separate fluxes of photosynthesis, photorespiration, and respiration. However, many challenges exist with gas exchange that limit the wide application of the technique. These include cost, usability, data processing requirements, and time needed for ensuring quality measurements. Off-the-shelf gas exchange systems cost tens of thousands of dollars and require frequent maintenance that challenges their widespread use. Most gas exchange systems limit the area of measurement to, at most, several cm², which presents issues related to scaling photosynthesis beyond a small section of one leaf. Typical measurements of in situ gas exchange require a minimum of 2–3 min to allow for both the system and the leaf to stabilize. Using these systems to measure beyond a simple survey of gas exchange, for example to measure light response or CO₂ response curves of A, requires substantially more time for each leaf. Recent techniques that exploit improved instrument precision can reduce the time for some measurements but generally at the expense of accuracy, and often require more advanced post-processing [21].

Scaling gas exchange measurements to the canopy or whole ecosystem presents significantly more challenges than at the leaf level, yet there are also more options (Table 2). Canopy chambers work in much the same way as leaf chambers, although at a larger scale. The general principle follows that of leaf-level measurements, although chambers are required to be much larger to encompass multiple plants and the potential is greater for errors associated with leaks or pressure fluctuations [17]. Canopy chambers have been extensively used to measure CO₂ fluxes for a wide range of vegetation types and their strengths lie in their ability to address small-scale spatial variability (Table 2). Furthermore, canopy chambers have been used both as a measurement and treatment system in global change studies to impose treatments as open-top chambers and acting as sample chambers when enclosed (e.g. [22]). Canopy chambers, however, can be limited in sampling frequency and spatial integration (Table 2) while also having a profound impact on the canopy microclimate.

Micrometeorological approaches to gas exchange lack the need for chambers but require large spatial areas (>4 Ha) and a sensor suite that can measure the upward/downward movement of air coupled with the gas concentrations in the air [20]. The dominant micrometeorological technique, eddy covariance (EC), provides near-continuous measurements of NEE integrated over large spatial areas, called the flux footprint, with minimal disturbance (Table 2) [20]. Air flow over a canopy consists of numerous rotating eddies. Measuring the speed and CO₂ concentrations of the eddies moving air upward and downward, provides the basic data needed to calculate fluxes of the footprint, which varies with wind speed and direction [23]. EC requires several important considerations to ensure the NEE data are robust and reliable [24], including ensuring sufficient atmospheric turbulence [23], applying corrections to exclude data fluxes extending beyond the area of measurements [25,26], and ensuring all measured fluxes follow the laws of thermodynamics [27,28] (Table 3). Because of inevitable gaps in data collection associated with field instrumentation, gap filling strategies are used to complete the time-series of flux data (Table 4). In addition to NEE, EC can apply to any measurable component of the atmosphere provided high temporal resolution sensors (≥10 Hz) exist (e.g. water vapor, methane, etc.). A global EC flux network, called FLUXNET, provides data from over 900 sites globally, allowing for a link between ecosystem and global NEE. This network provides unprecedented insights into environmental and biological drivers of ecosystem NEE [3,20,29–33]. Among other purposes, the long-term measurements of NEE from this network have improved understanding of ecosystem responses to climate and land-use change [34], and the data are essential to validate remote sensing and modeling products that scale to regions and the globe [35,36].

Whether using chamber-based or micrometeorological approaches, measured NEE provides an opportunity to explore changes in ecosystem-scale gas exchange at high temporal frequency. Photosynthesis at the ecosystem scale is generally defined as gross primary productivity (GPP), which is only one component of NEE. GPP is derived as the difference between measured NEE and modeled ecosystem respiration (ER; Table 5). Obtaining GPP from NEE involves modeling ER using temperature and light response functions; a process typically referred to as flux partitioning [24,32,37,38]. Flux partitioning allows for the investigation over time of GPP and ER in response to a variety of conditions [39–41]. A challenge with flux partitioning is introduced by the inhibitory effect of light on leaf respiration rates, known as the Kok effect [42]. In the light, autotrophic respiration can be significantly lower than at night resulting in GPP estimation errors when ignored [43].

Recent micrometeorological approaches have attempted to measure GPP using a sulfur-containing analog of CO₂, carbonyl sulfide (COS) that acts as natural ‘tracer’ molecule for GPP. This molecule enters a leaf in the same manner as CO₂ and is broken down by the enzyme carbonic anhydrase. Because of this, COS ‘uptake’ should scale with GPP, removing the need for partitioning NEE into the GPP and respiration components [44]. Studies using this method are showing promising insights with GPP estimated using CO₂ vs. COS measurements agreeing within 15% in forests and crops [45]. Another study that investigated variability in COS uptake and release in forests found agreement to within 3.5% between the two methods when GPP was high [46]. These results suggest an opportunity to use indirect methods for assessing GPP at larger scales, although recent work also suggests that photosynthetic tissues are not the only sink for COS [46–48].

Obtaining photosynthetic carbon uptake measurements using gas exchange systems is laborious resulting in efforts to replace this technique with other high-throughput methods. There exists a rapid growth in plant phenotyping greenhouses with the goal of automated measurement capabilities [49] at scales ranging from leaf to globe (Figure 1). Even with the most modern technologies, direct monitoring of leaf or plant level gas exchange would require substantial effort and resources. Thus, there are emerging technologies that provide means to infer plant responses to their growth environments that overcome the limitation of gas exchange [50–53]. Commercial sensors are available that provide information about plant canopy architecture and volume, which is important to infer growth over time [54], yet disentangling the underlying factors that lead to this growth requires physiological understanding. In the field, plot-level estimations of photosynthetic traits have been successfully estimated using a variety of platforms [55–57]. However, there needs to be improvements to the precision, accuracy, repeatability, and data pipeline before we can use these methods to estimate photosynthesis. Nonetheless, these new methods have a large potential impact on leaf to canopy understanding of plant physiology, ecosystem functioning and improving breeding efforts to maximize crop yields. In this section, we will discuss emerging technologies to monitor photosynthesis using spectral reflectance or fluorescence techniques. We will first outline the tools used for these approaches followed by a description of how these tools are being used.

Hyperspectral analysis is a non-destructive means of analysis that uses light reflected from vegetation to infer leaf, plant, canopy, or ecosystem performance. At the leaf and single-plant level, spectral sensors funnel light reflected from vegetation through a holographic diffraction grating, which separates light by wavelength across the electromagnetic spectrum [58]. Hyperspectral imaging data is in three ‘cubed’ dimensions with spectral wavelength (z) across spatial co-ordinates (x,y). Depending on the size of a single-pixel hyperspectral cameras can image vegetation from the whole plant to ecosystem scale [58].

Reflected light has become a powerful tool to characterize plant traits, including photosynthesis, given the varying response of light to leaf structure and pigment content at different wavelengths. In the near infrared (770–1300 nm), differences in chlorophyll and plant nitrogen content indicate a variety of vegetation stressors such as nutrient deficiency [59,60], plant disease status [61,62], and ozone damage [63], while the short wave infrared (SWIR1; 1300–2500 nm) indicates plant water status based traits [64,65]. In the past, discrete spectral reflectance indices were used as proxies for crop status [66]. However, computational and technological advances make it possible to derive photosynthetic capacities (maximum rate of carboxylation for C3 and C4 plants, V_cmax and V_max, respectively; and maximum rate of electron transport, J_max) and make predictions about photosynthetic performance scaling from the leaf [67–71] to the plot [72,73] and ecosystem scales [74].

One significant advance is the commercial availability of high-resolution fiber optic leaf clip-attachments. Hyperspectral radiometers typically contain a radiometrically calibrated light source and standardized white and dark reference panels for calibration. Leaf-level reflective intensity is compared with the reference material. Computer models (discussed later) are then used to correlate portions of the leaf's reflective spectrum with traditional measurements of gas exchange. Hyperspectral data can provide significant information about leaf photosynthesis at a fraction of the time compared with gas exchange [67–71,75]. These measurements can offer insight for upscaling to the plot level using field push carts [76] or drones mounted with hyperspectral cameras for breeding and research trials.

In addition to the hyperspectral methods mentioned above, recently handheld multispectral tools (e.g. FluroPen, Photo Systems Instruments, Drásov, Czech Republic; MultispeQ, PHOTOSYNQ INC. East Lancing MI, U.S.A.; and LI-600, LiCOR Biosciences Lincoln NE, U.S.A.) are used to monitor fluorescence and other parameters associated with leaves. Compared with hyperspectral leaf clips or fluorescence chambers sold with gas exchange units, these leaf tools can be used to more quickly and inexpensively screen for the vitality of photosynthetic systems under biotic and abiotic stresses (e.g. [77]). Furthermore, these tools provide opportunity, in some cases, to specify wavebands of interest for specific phenotypes that can extend beyond photosynthetic measurements.

Inspired by the successful leaf-level estimations of photosynthetic capacities, hyperspectral imaging (HSI) techniques are increasingly applied to canopy-scale measurements [73,78]. Imaging hyperspectral spectrometers provide more spatial information than a leaf-clip portable radiometer. Because of this, these sensors are being utilized to reveal variability in photosynthetic traits of interest across leaves, plants, and/or over large geographic areas [72,74]. These HSI sensors can scan individual plants in a few seconds [79] or provide analysis spanning several km² if mounted on aircraft or Earth-orbiting satellites [80,81]. Compared with point-based portable radiometers, these HSI sensors result in the accumulation of large amounts of data that need to be processed in an innovative way.

To link reflectance spectra to photosynthetic physiological parameters, data processing pipelines must be tailored to specific sensing platforms. These data pipelines are critical to applications such as field phenotyping in a high-throughput manner. For leaf-level estimations of photosynthetic variables using reflectance spectra, great efforts have been made to select statistical techniques that can provide the best predictive power [75]. Partial Least Square Regression (PLSR) [82] is currently the most common technique used to relate reflectance spectra to photosynthesis associated parameters [68,71] due to its ability to reduce tens to hundreds of spectral bands to just a few orthogonal principle components (also known as latent variables). There are also other machine learning algorithms such as Artificial Neural Network (ANN)-based regression and Least Absolute Shrinkage and Selection Operator (LASSO) that have been used to estimate photosynthesis [83]. The availability of these machine learning and empirical algorithms also poses a dilemma regarding the most effective approach. Collectively harnessing the strengths of individual empirical or machine learning algorithms through regression stacking shows promise [72] although further studies are needed to test its effectiveness across more plant species. For estimations of photosynthesis using reflectance spectra at the plot and ecosystem levels, further data processing steps are necessary to account for spurious variations in reflectance caused by sun-target-sensor geometry, canopy structure, leaf scattering, atmospheric contaminations, and background soil [75]. These steps are required to ensure that only reflectance data associated with photosynthesis are used for estimations. Although Radiative Transfer Models (RTMs) such as PROSAIL [84] are developed to remove those spurious variations, few of them can be directly used in the proximal sensing setting [85]. However, these RTMs provide an alternative way to reduce hyperspectral data into several meaningful leaf traits, such as chlorophyll concentration, that can serve as a proxy for photosynthesis. For example, RTMs-inverted traits were shown to explain up to 60% of variation in photosynthetic physiology in a crop species [72].

Remote-sensing products that measure GPP are traditionally based on the Light-Use Efficiency (LUE) concept of ecosystem modeling [86] and empirical models that rely on the relationships between remote sensing-derived variables and GPP [87–90]. These methods provide reasonable estimates of GPP compared with measured EC fluxes, however, new emerging spectral sensing technologies including Solar-Induced chlorophyll Fluorescence (SIF) are providing potential for estimating GPP at the ecosystem scale [91–93]. A fraction of solar radiation absorbed by chlorophyll is emitted as fluorescence, hence SIF is more physiologically based than other traditional remote sensing products [94] as it is a direct product of the photosynthetic process [95–97]. While pulse amplitude modulated chlorophyll fluorescence has long been used to measure photochemical efficiencies and heat dissipation in individual leaves [98], this should not be confused with SIF, which relies on measuring of the radiance chlorophyll fluorescence from an ecosystem.

Passive SIF measurements were first applied at the satellite scale (Table 6) [99] to assess regional and global scale patterns of SIF alongside GPP [91–93] and is now being implemented at flux towers across multiple ecosystem types to determine the physiological and structural relationship between SIF and photosynthesis at this scale [100–103]. Likewise, the near-infrared radiance of vegetation index (NIR_v) has shown promising accuracy at detecting photosynthetic variability at the hourly scale over crop and forest system [104,105]. Therefore, both SIF and NIR_v should enable real-time monitoring of productivity and stress.

The relationship between SIF and GPP is primarily dominated by absorbed photosynthetic active radiation (APAR) [106,107], implying that the correlation between SIF and GPP is the highest when photosynthesis is primarily light-limited [108,109]. However, GPP is also controlled by environmental factors other than light, and recent insights suggest that SIF responded to environmental stresses in a similar way as GPP, encouraging the application of SIF to estimate photosynthesis [94]. A relationship between SIF and GPP was similar among ecosystems although the relationship was stronger for grasslands than forests, savannas and croplands, and for C4 grasslands and crops than C3 ecosystems [94]. This quasi-universal relationship indicates that SIF could be a valuable tool for inferring GPP of the land surface. More collaborative studies between the EC and remote sensing communities are needed to evaluate why the relationship between SIF and GPP varies among ecosystems and under differing environmental conditions to improve the ability of SIF products to estimate ecosystem GPP robustly to scale regionally and globally.

Much progress has been made to understand the relationship between SIF and GPP but many challenges remain [109–111]. Higher spatial and temporal resolution SIF measurements are needed to coincide with the continuous GPP measurements [112]. Promising solutions to these challenges would be to develop remote sensing approaches that can cross-calibrate and blend multi-source SIF and reflectance measurements for a consistent record in both spatial and temporal domains. For example, combining satellite SIF with satellite reflectance was used to generate a spatially and temporally continuous SIF dataset [113]. Another solution is to improve SIF sensor designs to facilitate measurements at a much higher spatial and temporal resolutions. For example, the Fluorescence Imaging Spectrometer (FLORIS) onboard the Fluorescence EXplorer (FLEX) satellite can provide SIF at a better spatial resolution than its predecessors (Table 6) [114] and the newly launched Orbiting Carbon Observatory 3 instrument (OCO-3) allow for more coverage globally at higher definition [115].

Interestingly, much of the work on remote sensing has initiated with large-scale measurements, yet there is a tremendous need to increase throughput of measurements at leaf and plot scales, particularly for application in high throughput phenotyping facilities. Whether these techniques are fully scalable remains uncertain, yet the opportunity for multidisciplinary research has advanced the versatility of the tools outlined in this review beyond their original users. Moving forward, simplifying data collection through ‘turn-key’ sensors and standardizing data analysis pipelines for the variety of techniques outlined here are certain to advance understanding of plant function from molecular to global scale.

• Monitoring Photosynthesis at every scale, from leaf to ecosystem, is an important task given the challenges of climate change and growing human populations.

• In the past 5 years there have been significant improvements to the technology and computational tools used to measure photosynthesis at every scale. And new facilities and equipment are being used around the world to monitor photosynthesis.

• Hyperspectral imaging at the leaf, and canopy scale paired with improved computational modeling allows for rapid estimates of important biochemical parameters.

• Micrometeorological approaches to estimate Gross Primary Productivity have been improved by the uses of sulfur tracing elements.

• Monitoring Solar Induced Fluorescence is a promising satellite-based method that should enable real-time monitoring of global ecosystem productivity.

The authors declare that there are no competing interests associated with the manuscript.

M.H.S., N.G.-C., and C.J.B. conceived the outline, all authors contributed to the organization and writing of the manuscript, M.H.S. and N.G-C. Edited the manuscript, and C.J.B. supervised the project.

This work is supported by funding from Global Change and Photosynthesis Research Unit of the USDA Agricultural Research Service. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the views of the U.S. Department of Agriculture. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.

COS

carbonyl sulfide

EC

eddy covariance

ER

ecosystem respiration

FLEX

fluorescence explorer

FLORIS

fluorescence imaging spectrometer

GPP

gross primary productivity

HSI

hyperspectral imaging

NEE

net ecosystem exchange

RTMs

radiative transfer models

SIF

solar induced fluorescence