Background

Terrestrial ecosystems in the Arctic-Boreal region play a critical role in the global carbon cycle and climate change mitigation as a significant net carbon sink through vegetation photosynthesis [1, 2]. Based on the Coupled Model Intercomparison Project Phase 6 (CMIP6; [3]), a tipping point of the carbon cycle is projected with reduced net sink in the Arctic-Boreal region by the end of the 21st century and potentially to turn the region into a net carbon source in the next century [4, 5]. This weaker net carbon sink can be attributed to unique consequences of rapid warming in this region [6], which have divergent impact on photosynthesis. For example, (1) thawing permafrost enhances microbial decomposition, while it may favor photosynthetic carbon uptake through enriched soil nutrients [7,8,9]; (2) more frequent and widespread fires [10,11,12] rapidly clear vegetation coverage and release large amounts of carbon into the atmosphere reversing decades of carbon uptake [13, 14]; and (3) shifts of plant compositions [15,16,17], plant productivity [18, 19], and plant phenology [20,21,22] lead to heterogeneous changes in carbon uptake [23,24,25]. Due to complex and understudied ecosystem-climate feedbacks [26,27,28,29,30,31,32,33], the carbon uptake by photosynthesis at the ecosystem level, which is also known as Gross Primary Productivity (GPP), is highly uncertain in this region. Compared to the 24% uncertainty of global GPP (\(3.3 \pm 0.8\) \({\text {gt C yr}^{-1}}\); [34]), the uncertainty of Arctic-Boreal GPP is 227% of its typical value (\(0.22 \pm 0.50\) \({\text {kg C m}^{-2}~yr^{-1}}\)), becoming the second largest uncertainty source for the carbon cycle [35]. Therefore, to better understand the ecosystem-climate feedbacks in the Arctic-Boreal region and constrain the uncertainties on the global carbon cycle [36, 37], there is a pressing need to closely monitor GPP [5, 38, 39].

GPP can be evaluated both directly and remotely. Direct measurements of GPP rely on chamber-based measurements [40, 41] and tower-based Eddy Covariance techniques (EC) [42, 43]. Chamber-based measurements are advantageous at evaluating GPP from different vegetation types within an ecosystem. However, existing chambers in the Arctic-Boreal region are mostly manual and thus laborious [44]. Contrarily, EC averages out fine-scale heterogeneity but provides convenient long-term monitoring [43]. While the state-of-the-art flux product in this region (i.e., ABCflux [39]) synthesizes available chamber-based and EC observations, the representativeness of this product is still limited due to the spatial scarcity and temporal sporadicity of both methods [35, 39, 45, 46], especially compared to growing long-term EC networks in other regions [47,48,49,50,51,52,53,54,55,56,57,58,59],

Remote sensing techniques, on the other hand, can infer GPP continuously with extensive spatial coverage [60,61,62,63,64] in spite of some shortfalls unique to the Arctic-Boreal region, such as seasonal gaps of observations and complications from non-vegetation [65, 66]. A conceptual model of remotely inferred GPP can be expressed as:

$$\begin{aligned} \textrm{GPP} = \textrm{fPAR} \times \textrm{PAR} \times \textrm{LUE}, \end{aligned}$$
(1)

where PAR is Photosynthetic Activate Radiation, and fPAR is the fraction of PAR being absorbed by chlorophyll. The product of fPAR and PAR, that is the Absorbed PAR (APAR), will be primarily partitioned between photochemical quenching for photosynthesis and dissipation of excess energy as heat by photoprotective pigments. A small amount of APAR will be re-emitted in red and far-red wavelengths as Solar-Induced chlorophyll Fluorescence (SIF) [67]. Hence, Light Use Efficiency (LUE) quantifies the fraction of APAR utilized by photosynthesis.

Conventionally, remotely inferred GPP is based on the canopy color measured by optical reflectance because the canopy greenness is a proxy of APAR [64, 68, 69]. The vegetation indices of greenness measurements such as Normalized Difference Vegetation Index (NDVI) [68] and Enhanced Vegetation Index (EVI) [70] are commonly used for this purpose. Nevertheless, canopy greenness alone provides no information about LUE [71, 72]. This limitation of canopy greenness challenges the GPP estimation for land cover types with sustained canopy greenness and APAR, such as evergreen forests which is one of the dominant land cover types in the Arctic-Boreal region [73, 74].

More recently, advancements in remote sensing have revealed that canopy-scale far-red SIF, i.e., far-red SIF escaped from the canopy and detected by remote instruments, serves as a better proxy for GPP than conventional greenness measurements [75, 76]. A conceptual model for instantaneous canopy-scale SIF measurements [77, 78] can be written as:

$$\begin{aligned} \textrm{SIF} = \textrm{fPAR} \times \textrm{PAR} \times \Phi _{F} \times f_\textrm{esc} , \end{aligned}$$
(2)

where \(\Phi _{F}\) is the quantum yield of fluorescence, and \(f_\textrm{esc}\) is the probability of fluorescence escaping the canopy and reaching the remote instruments [79, 80]. To account for the dependence of SIF on the instantaneous PAR at the time of measurement, the instantaneous SIF measurements are often normalized to daily mean SIF (SIF\(_\textrm{dc}\)) by solar zenith angle [65, 81,82,83,84].

This conceptual model underscores that SIF is primarily driven by APAR as a bi-product of photosynthesis ([76]). Meanwhile, SIF also contains LUE information [75] because \(\Phi _{F}\) is under the regulation of photoprotective pigments [85]. These links with APAR and LUE make SIF a critical tool for mechanistically tracking GPP in land cover types with and without persistent canopy greenness. Besides, SIF retrieval is less sensitive to common background noise in the Arctic-Boreal region, such as water and snow [66, 69, 86, 87]. Given these strengths, SIF is an effective tool in the Arctic-Boreal region. Several studies have presented this advantage of SIF over canopy greenness empirically [66, 86, 88,89,90,91]. However, the full potential of SIF in evaluating vegetation dynamics and constraining uncertainties on carbon fluxes with spatial details is still underexplored.

There have been a few review articles on SIF [92,93,94] that extensively discuss observational platforms, retrieval techniques, the physiological link between SIF and photosynthesis at the molecular level and global scale. However, limited by their global scopes, these reviews have not assessed the unique challenges of SIF in the Arctic-Boreal region, such as limited observations and underrepresented land cover types. To advance the application of SIF in the Arctic-Boreal region, this review focuses on recent progress on Arctic-Boreal SIF observations and assessing the prospects of SIF research across the Arctic-Boreal region for better monitoring vegetation dynamics and carbon fluxes in response to climate change.

Cross-scale SIF Measurements

This section aims to comprehensively overview cross-scale instruments used for measuring SIF in the Arctic-Boreal region. Currently, there are only a few studies [66, 95,96,97] quantitatively analyze the linear proximity of SIF and GPP from different observational platforms (Fig. 1). By examining the advantages and limitations of spaceborne, airborne, and tower-based instruments, their applications can be optimized for capturing the spatial distribution and temporal dynamics of SIF in the Arctic-Boreal region [66, 91, 98].

Fig. 1
figure 1

A graphic summary of published r\(^{2}\) in the Arctic-Boreal region between GPP and SIF from different observational platforms and spatial resolutions, including an airborne platform with spatial resolutions of 10 m, 20 m, 40 m, and 80 m; tower-based platforms without uniform spatial resolution (Table 2); and a spaceborne platform at 10km resolution. Each scatter represents a study site/spatial resolution. The airborne data points are from the modeled instantaneous GPP and airborne snapshot of SIF [95], while the tower-based and spaceborne data points are from daily mean Eddy Covariance (EC) GPP and daily mean tower/satellite SIF [66, 96, 97]. Cheng et al., [66] reported r\(^{2}\) based on the climatology, which may not be derived from synced timeseries of GPP and SIF

Full size image

Spaceborne SIF

Several generations of spaceborne instruments have been deployed to measure SIF across the globe, but not all of them are optimal for the Arctic-Boreal region. Notable satellite missions/instruments overpassing the Arctic-Boreal region (north of \(50^{\circ }\)N) include the Global Ozone Monitoring Experiment-2 (GOME-2) [99], Greenhouse gases Observing SATellite (GOSAT) [100], SCanning Imaging Absorption spectroMeter for Atmospheric CartograpHY (SCIAMACHY) [101], Orbiting Carbon Observatory-2 (OCO-2) [102], Carbon Dioxide Observation Satellite Mission (TanSat) [103], and TROPOspheric Monitoring Instrument (TROPOMI) [82]. Each satellite mission/instrument has different scanning patterns and satellite orbits, resulting in diverse spatial and temporal characteristics of SIF measurements (Table 1, Fig. 2). Therefore, it is important to consider these differences when comparing their SIF measurements across satellite missions/instruments and inferring GPP from their SIF measurements [98]. To address this issue in the context of Arctic-Boreal ecosystems, common methods of processing SIF measurements will be discussed in this section.

Spatial Coverage and Resolution

The spatial coverage of satellite missions/instruments determines the geographic range of SIF measurements. Most existing satellites, such as SCIAMACHY, GOME/GOME-2, GOSAT/GOSAT-2, TanSAT, OCO-2, and TROPOMI (Table 1, Fig. 2a), have polar or near-polar orbits that allow for SIF measurements in the Arctic-Boreal region as they pass over the entire latitudinal range of the region. However, certain satellites, namely Tropospheric Emissions: Monitoring Pollution (TEMPO) and Sentinel-4, have limited coverage and primarily focus on lower latitudes of the region, up to \(58 ^{\circ }\) N and \(65 ^{\circ }\) N, respectively (refer to Table 1 and Fig. 2a). As a result, these satellite missions/instruments will not provide sufficient SIF measurements for the entire Arctic-Boreal region.

Table 1 Existing and upcoming satellite missions/instruments measuring SIF in the Arctic-Boreal region (north of \(50^{\circ }\)N)
Full size table

The spatial resolution of SIF measurements determines their spatial representativeness relative to the heterogeneous surface in the Arctic-Boreal. Fine spatial resolution is important for the Arctic-Boreal region because the land cover in this region is highly heterogeneous, with the co-existence of different land cover types, such as vegetation, water, and snow [104]. In addition, complex topography poses challenges in resolving these diverse land cover types within large pixels [66, 105, 106]. Coarse spatial resolutions may yield a wide range of correlations between SIF and GPP (Fig. 1). Therefore, SIF measurements with finer spatial resolution have greater potential to represent different land cover types more accurately and distinguish their contributions to the carbon cycle in the Arctic-Boreal region [96, 107, 108]. Also, measurements with lower spatial resolutions (larger pixel sizes) are more susceptible to contaminations from clouds and aerosols, such as SCIAMACHY, GOME/GOME-2, GOSAT/GOSAT-2 [99, 109]. Currently, state-of-the-art satellite missions/instruments like TROPOMI and OCO-2 offer SIF measurements with finer spatial resolutions. Upcoming missions, particularly the European Space Agency’s FLuorescence EXplorer (FLEX) [110, 111], will significantly enhance the spatial resolution, with a footprint size of 300 x 300 m\(^2\), which will be the smallest among current SIF measurements (Fig. 2b). It is worth noting that SIF measurements from a smaller footprint are more likely to be influenced by topography and thus need more rigorous corrections when instantaneous SIF measurements (Eq. 2) are normalized to daily mean values [65].

To provide more surface details at smaller spatial scales than the existing spaceborne measurements, there are several hybrid products that downscale SIF measurements to scales as fine as hundreds of meters using optical reflectance measurements and environmental data [112,113,114,115,116,117]. However, the accuracy of these downscaled global products (such as spatially contiguous SIF (CSIF) [113] and Global ’OCO-2’ SIF (GOSIF) [114]) and their performance of tracking GPP have not been assessed across the Arctic-Boreal region at the finer spatial scale. Madani et al. [118] only validated the temporal variations of CSIF but not its spatial representativeness for a few Arctic-Boreal EC towers. Wen et al. [119] cross-validated the global spatial patterns of downscaled and satellite SIF products and found that downscaled SIF products can yield large biases in the Arctic-Boreal region because of the poor performance of universal downscaling models in this region. The higher noise in optical reflectance measurements due to background soil, water, and snow [66, 86, 113] can potentially affect the accuracy and reliability of the downscaled SIF products as well. To better train the downscaling models, more satellite observations with finer spatial resolutions in the Arctic-Boreal region are needed. Airborne instruments with much higher spatial resolutions (e.g., 30 m; Section 2.2) can help validate the spatial representativeness of downscaled products. However, such validation has only been done outside the Arctic-Boreal region [119]. Therefore, further validation and assessments of these downscaled products are needed for the Arctic-Boreal region to ensure their performance in this unique region.

Fig. 2
figure 2

A graphic summary of existing and upcoming satellite missions measuring SIF in the Arctic-Boreal region (north of \(50^{\circ }\)N): (a) The spatial coverage of satellites that do not cover the entire Arctic-Boreal region; (b) The comparison of satellite footprints; And (c) the mission span and temporal resolution of SIF measurements. In (c), arrows indicate ongoing or future missions, and dashes suggest the temporal resolution of SIF measurements. A densely dashed line means a higher temporal frequency. The exact value of temporal frequency can be referred to Table 1

Full size image

Sampling Frequency and Overpass Time

The revisit time and swath width of satellites together determine the time required to complete one cycle of global SIF measurements and, thus, the sampling frequency at a specific location. These factors together determine the temporal coverage and resolution of SIF measurements in the Arctic-Boreal region. Existing and upcoming satellite missions have a wide range of sampling frequencies (Table 1, Fig. 2c), allowing for tracking the dynamics of SIF across temporal scales.

Fig. 3
figure 3

The zonal mean of minimum solar zenith angle over a year, representing the availability of solar radiation at the local solar noon across different latitudes. The minimum solar zenith angle greater than \(90^{\circ }\)  means polar night, when solar radiation is zero throughout the day. The white contours represent the two common thresholds for filtering out SIF measurements at low solar radiation conditions (\(80^{\circ }\)   in [109] and \(70^{\circ }\)   in [82, 99])). These thresholds indicate the maximum solar zenith angle (minimum solar radiation) acceptable for valid SIF measurements. SIF measurements taken at solar zenith angles higher than the thresholds are excluded due to the low signal-to-noise ratio

Full size image

At seasonal and longer scales, polar-orbiting satellites have difficulties getting the complete and accurate seasonal cycle of SIF in the Arctic-Boreal region because of the lack of valid sampling during shoulder seasons and winters. Because SIF is driven by solar radiation (Eq. 2), to keep the measurement noises low, SIF measurements are often filtered out if they are taken at low solar radiation conditions, such as when the solar zenith angle exceeds certain thresholds. This will remove most of the observations during the Arctic-Boreal winters and potentially miss the onset and cessation of growing seasons depending on the threshold and sampling frequency (Fig. 3). A stricter threshold, such as \(70^{\circ }\)   used in [82, 99], will exclude more SIF measurements during shoulder seasons compared to a looser threshold, such as \(80^{\circ }\)   used in [109] (Fig. 3). This artificial cutoff of growing season varies by instruments with different sampling frequencies leading to misshaped growth seasonality, particularly for instruments with low sampling frequency [120, 121]. Missing the onset and cessation of GPP can cause large uncertainties in the net carbon flux during the shoulder seasons, when the majority of net carbon emission happens in the Arctic-Boreal region [122]. Such artificial cutoffs of growing season also fail to precisely track the temporal shifts of growth onset and cessation with climate change [33, 123, 124].

To track the subdaily and diurnal variations of photosynthesis, satellite missions/instruments like TROPOMI with different overpass times at certain locations can be useful [125, 126]. Additionally, geostationary satellite missions/instruments like TEMPO, Sentinel-4, and canceled Geostationary Carbon Cycle Observatory (GeoCarb) [127] can offer multiple SIF measurements each day. Unfortunately, none of the existing or upcoming geostationary satellite missions have complete coverage in space for the Arctic-Boreal region (Table 1, Fig. 2a). Therefore, relying on satellite SIF measurements alone is challenging to track the subdaily and diurnal variations in photosynthesis in the entire Arctic-Boreal region. Instead, tower-based instruments can be more feasible for tracking the dynamics of SIF at a subdaily scale and over longer terms (Section 2.3).

Table 2 Existing and new SIF observation systems
Full size table

Airborne SIF

As an analog of spaceborne instruments, airborne instruments measure SIF at enhanced spatial resolutions benefiting from a closer distance between the aircraft and the ground. Among a few airborne SIF instruments (summarized in [137]), NASA Jet Propulsion Laboratory’s Chlorophyll Fluorescence Imaging Spectrometer (CFIS) [138] is the only instrument has extensively flown in the Arctic-Boreal region. During NASA’s Arctic-Boreal Vulnerability Experiment (ABoVE) airborne campaign in 2017 [139], CFIS SIF was retrieved at the finest resolution of 30 m, with a focus on more than 20 EC tower sites across Alaska and northwest Canada [140]. Unfortunately, there has not been reported SIF-GPP relationship using CFIS SIF. Tagliabue et al. [95] reported r\(^{2}\) of 0.5 between the snapshots of GPP and SIF from a European airborne instrument (HyPlant). The optimal airborne SIF-GPP relationship can be achieved by spatially aggregating SIF observations (Fig. 1). Therefore, the small pixels of CFIS SIF, together with optical reflectance measurements, can better resolve the spatial distribution of different land covers [91, 102, 141] and align with the footprints of EC GPP [142, 143]. The existing collections of snapshots can also help validate the spatial representativeness of downscaled SIF products discussed in Section 2.1.1. However, such validation with a few snapshots does not hold over time. For example, CFIS sampled SIF in the Arctic-Boreal region for only two days during the ABoVE airborne campaign in 2017. Without additional repeated sampling, it is challenging to assess the rapid and heterogeneous changes of Arctic-Boreal vegetation under climate change.

Tower-Based SIF

Existing tower-based instruments (summarized in [98]) are competent for continuously measuring day-to-day and subdaily variations in SIF at specific locations. Tower-based SIF and EC GPP measurements often have overlapping footprints [57], making them ideal for deriving SIF-GPP relationships at diurnal and seasonal scales [66, 144].

However, similar to the spatially limited EC GPP measurements, tower-based SIF measurements are also scarce in the Arctic-Boreal region. Currently, PhotoSpec (in Delta Junction, Alaska and in Saskatchewan, Canada) [96] and FluoSpec2 (in Toolik, Alaska) [145] are the only tower-based instruments actively measuring SIF in the Arctic-Boreal region, which hinders extensive examination of SIF-GPP relationship across the Arctic-Boreal region. Only two Boreal forests have reported tower-based SIF-GPP relationship (Fig. 1; [66, 96, 97]). Nevertheless, novel tower-based SIF observation systems are progressing quickly with the potential to be deployed in the Arctic-Boreal region, e.g., Tower Spectrometer on Wheels for Investigating Frequent Timeseries (TSWIFT; [146]), the Fluorescence Box (JB Hyperspectral; Dusseldorf, Germany), Automatically long-term SIF observation system (AutoSIF; [83]), the Fluorescence Auto-Measurement Equipment (FAME; [147]), and a NASA/Goddard Space Flight Center Prototype for Field Spectroscopy (FUSION). The specifications of these tower-based SIF observation systems are summarized in Table 2. It is worth noting that these systems have various spectral characteristics, field of view, and mobility (Table 2), which are critical for the representativeness of SIF retrievals and comparison retrievals across different systems. For example, although tower-based SIF is often retrieved at around 760 nm, the various spectral ranges and spectral resolutions (Full Width at Half Maximum) lead to uncertainties in the retrieved SIF values [99, 148]. The accuracy of different retrieval algorithms is summarized by Mohammed et al. [92].

A small field of view, such as in PhotoSpec, makes it easy to target individual trees, filter out background noises (e.g., snow), and resolve SIF signals from different parts of the canopy by scanning individual trees [149]. Pierrat et al. [96] remotely detected diverse growing onsets across species in a mixed-species Boreal forest by taking advantage of the small field of view in PhotoSpec.

The directional effect between the solar incidence angle and the viewing angle is not negligible for tower-based instruments because of large variations in the solar incidence angle in the Arctic-Boreal region [150]. The directional effect can be reduced by fusing observations from different view geometries [146, 151]. Therefore, the systems with scanning telescopes, e.g., PhotoSpec, TSWIFT, and FUSION, are advantageous, although harsh winter weather can be challenging for the parts of rotating mechanics.

Enhancing Cross-scale Observational Network

Airborne and tower-based SIF measurements complement the coarse spatiotemporal resolutions of spaceborne SIF in the Arctic-Boreal region. However, the current availability of airborne and tower-based instruments is constrained to a few sites and a brief sampling duration in the Arctic-Boreal region compared to lower latitudes. Additionally, regional studies on the Arctic-Boreal vegetation dynamics are imbalanced across continents, with a greater focus on North America than Eurasia as a result of imbalanced data availability, while Eurasia is also a significant contributor to the global carbon cycle [124]. Therefore, enhancing the international network of airborne and tower-based SIF measurements is the key to validating spaceborne measurements and investigating scale-dependent vegetation dynamics throughout the entire Arctic-Boreal region [91, 107].

As sensing technology advances, flying lower-cost and lighter airborne instruments on Unmanned Aerial Vehicles (UAV) becomes convenient to provide frequent SIF measurements at fine spatial resolution [152, 153]. More encouragingly, collaborations among agencies are exploring the opportunities to fly airborne instruments more frequently and regularly, such as NASA ABoVE and National Ecological Observatory Network (NEON) airborne observation platform [139].

The success of volunteer-based EC networks, such as Fluxnet, is an excellent example of strengthening carbon cycle studies through global collaboration [57]. In the spectral imaging community, a similar network of optical reflectance called SpecNet [154] has made fruitful findings to boost the connections between remote sensing and carbon uptake [155]. There are only a few tower-based SIF measurements in the Arctic-Boreal region and mostly in North America. Going forward, fostering a similar tower-based SIF network is strongly favored for understanding the biochemical and biophysical processes in the Arctic-Boreal region. The two PhotoSpec instruments [96, 144] and one FluoSpec2 instrument [145] are prototypes of such a network, showing promise in this regard. Expanding the SIF network in the Arctic-Boreal region is viable with lower-cost and easier-maintained sensors [156, 157].

Quantitative Estimation of GPP in the Arctic-Boreal Region

Empirical SIF-GPP Relationship

GPP is often quantitatively evaluated by satellite SIF using empirical models at the seasonal scale [66, 158]. Taking Eqs. (1) and (2) together, inferring seasonal GPP from SIF\(_\textrm{dc}\) leads to solving the GPP/SIF\(_\textrm{dc}\) ratio (k), which contains the information of LUE, \(\Phi _{F}\), and \(f_\textrm{esc}\):

$$\begin{aligned} \textrm{GPP}&= \textrm{k} \times \textrm{SIF}_\textrm{dc}, \end{aligned}$$
(3)
$$\begin{aligned} \textrm{k}&= f(LUE,\Phi _{F}, f_\textrm{esc}). \end{aligned}$$
(4)

At the canopy level, this SIF-GPP relationship (Eq. 3) is approximately linear such that k is relatively constant across different months [159, 160]. Empirical studies [66, 161,162,163] solve the parameter k by linearly regressing GPP and SIF. Then, the k values are categorized based on generic plant functional type to integrally represent different plant physiology (e.g., LUE and \(\Phi _{F}\)) and canopy structures (e.g., \(f_\textrm{esc}\)) (Eq. 4). The first study [66] to derive k values for unique Arctic-Boreal land cover types shows that high-stature land cover types (e.g., evergreen and deciduous forests) have higher k than lower-stature land cover types (e.g., low shrubs and tundra). However, the reported k values still have large uncertainties inherited from SIF and GPP datasets due to large footprints, background noise, and lack of validation [90, 164]. Validating the satellite SIF and GPP measurements with more evenly distributed tower-based SIF and GPP can significantly reduce extrapolation [165] and improve the confidence of k values and GPP predictions thanks to relatively small and homogeneous footprints of tower-based instruments [66, 166, 167]. The spatial representativeness of k values within the footprint of tower-based instruments can be further validated with airborne snapshots [91] once more repeated airborne missions are available.

As the record of SIF measurements becomes longer, mapping the changes in GPP over the long term is possible. In the Arctic-Boreal region, the empirically derived k should be calibrated regularly to keep up to date with the changing plant composition and canopy structure due to warming, such as shrub expansion, forest-tundra ecotone shifts, and wildfires. Within a satellite footprint, these changes can be reflected in the mixing of vegetation types and changes in k. However, there are not sufficient observations to quantify such changes [168] including SIF.

There have been discussions on the goodness of assuming linear SIF-GPP relationship [162, 167]. Interestingly, the reported SIF-GPP relationship in the Arctic-Boreal region is debatable. At monthly and daily scales, the Arctic-Boreal k value of the linear SIF-GPP relationship shifts with season (Fig. 4a) according to both regional-scale [66, 118, 167] and tower-based [144] studies. As the temporal scale refines to half-hourly, the SIF-GPP relationship becomes nonlinear (Fig. 4b) due to small light response of SIF in winter and seasonal variations in light use efficiency [144]. There have not been mechanistic analyses exploring the linearity/nonlinearity of SIF-GPP relationship over the entire Arctic-Boreal region. Existing global-scale analyses attribute the nonlinearity to the nonlinear response to environmental stresses [166, 169] and aggregated directional effect [80, 170, 171]. Considering environmental stresses caused by climate change, large seasonal variations of solar incidence angle, and heterogeneous canopy structure in the Arctic-Boreal region, more work should be done to mechanistically explain the linearity/nonlinearity of the SIF-GPP relationship across this region. Future works resolving such complex roles of climate drivers and radiative factors on the SIF-GPP relationship will also benefit from evaluating the climate change impacts on carbon uptake and tracking physiological and structural changes of Arctic-Boreal vegetation at different temporal scales.

Fig. 4
figure 4

A graphic demonstration of the seasonal variations in (a) linear and (b) nonlinear SIF-GPP relationships reported by Pierrat et al. [144] and Chen et al. [167]. The linear relationship is mostly found in monthly or daily scale analyses. The nonlinear relationship is mostly found in half-hourly scale analyses. The linear relationship follows Eq. 3. The nonlinear relationship follows the mathematical form of \(y = a*x/(b+x)\) [144]. Dormancy has a smaller range of SIF due to the low productivity in the Arctic-Boreal region. Outside the dormancy, there is not enough evidence to determine the relative magnitudes of the nonlinear relationship curve from season to season

Full size image

Machine Learning Models

Without the need to solve the complex mechanisms, machine learning approaches [166, 172, 173] conveniently offer a potential solution to simulate the complex SIF-GPP relationship and predict GPP using satellite SIF, optical reflectance, environmental data, and land cover information. FluxSat and FluxSat V2.0 are examples of such datasets [172, 174, 175]. In FluxSat, the prediction of GPP in high productivity areas directly benefits from incorporating SIF [174]. In other regions, including the Arctic-Boreal region, the significance of SIF data for the machine learning-based GPP prediction needs further investigations [172]. Meanwhile, these machine learning models also inherent the compounded uncertainties from SIF and other optical remote sensing observations in the Arctic-Boreal region, such as spatial heterogeneity (Section 2.1) and snow contamination [66]. More tower-based observations of both GPP and SIF across the Arctic-Boreal region can help reduce the extrapolation uncertainties [165] and improve the accuracy of machine learning models in the Arctic-Boreal region.

Advances in Carbon Cycle Modeling with SIF

Intrinsically, process-based models (Table 3) simulate GPP using parameterized photosynthetic traits, e.g., the maximum rate of carboxylation and photosynthesis yield, [176]. These parameters are estimated based on tower-based measurements and then extrapolated globally [177].

Dynamic vegetation models and [123, 178, 179] and other process-based terrestrial ecosystem models [180,181,182] simulate the mechanics of fluorescence and radiative transfer, where SIF measurements can be assimilated into these models and optimize the photosynthetic parameters. On the other hand, conventional land surface models, which do not simulate SIF directly, SIF-driven GPP can be used to optimize the parameterized photosynthetic traits and constrain the errors in simulating global carbon fluxes in land surface models [183,184,185,186]. In the Arctic-Boreal region, GPP simulated by the optimized models has improved the spatial distribution and temporal patterns, which reveal a strong reduction of GPP in the Arctic-Boreal region [123, 178, 187].

Recently, there has been significant progress in simulating radiative transfer and SIF within land surface models [188]. This advancement provides opportunities for physically simulating the SIF-GPP relationship. Future studies can benchmark the model simulation with remote sensing in order to better map photosynthetic traits and investigate the complex climate drivers of carbon fluxes, especially in the Arctic-Boreal region.

One drawback of these models is their global universal parameterization, which often neglects or oversimplifies the complex vegetation distribution and land cover types in the Arctic-Boreal region, leading to large uncertainties in the simulated carbon fluxes and their spatial variations. Better resolving the SIF-GPP relationship for Arctic-Boreal land cover types [118] and justifying its spatial representativeness given the heterogeneous land cover within coarse footprints [91, 168] is critical for simulating spatially detailed changes in carbon fluxes.

Table 3 Process-based models that incorporate SIF
Full size table

Emergent Changes in Arctic-Boreal Region

As SIF measurements become more abundant in the Arctic-Boreal region, we are gaining more insights into mechanistically tracking GPP using SIF data obtained at various spatiotemporal scales [189]. This section focuses on reviewing the applications of SIF in monitoring crucial changes in the Arctic-Boreal region [10]. Meanwhile, outlooks on future research are provided for deepening our understanding of the resulting changes in the global carbon cycle.

Arctic “Greening”/“Browning” and Its Climate Drivers

Under the scenario of global warming, Arctic “greening” and “browning” have been signature indicators of changing vegetation dynamics, which are composed of changes in vegetation phenology and vegetation distribution [32]. SIF has been mostly used in evaluating the vegetation phenology over the Arctic-Boreal region, including the peak GPP during the growing season, the timing of growth onset and cessation, and the length of the growing season. Implementing SIF in detecting changes in vegetation distribution and land cover types is understudied.

Vegetation Phenology

Large-scale and long-term studies [32, 190,191,192] rely on long records of satellite greenness measurements, such as NDVI or EVI, to evaluate changes in vegetation phenology across the Arctic-Boreal region. However, these greenness measurements primarily reflect APAR and leaf area rather than actual GPP [71, 72]. This limitation restricts the accurate estimation of changing carbon fluxes and the identification of underlying mechanisms driving these changes [193,194,195], especially in needle-leaf forests where the responses of GPP and greenness measurements to climate diverge [86, 89, 196]. Since SIF varies with both APAR and LUE, it is a better proxy than the greenness measurements for tracking the seasonal GPP across various land cover types in the Arctic-Boreal region [85, 88, 197].

The spring onsets of SIF and GPP in the Arctic-Boreal region closely follow the rising temperature and landscape thawing [86, 194, 198, 199], which is consistent with findings from proximal measurements in a sub-alpine ecosystem [200]. Warming causes extended non-frozen periods, leading to a longer growing season [33, 201]. Both SIF and GPP indicate that an early onset caused by warming may induce reduced magnitude of photosynthesis in fall, as soil moisture depletes [202]. Meanwhile, exposure to cold temperatures can delay the spring onset of SIF and GPP [203]. It is worth noting that the onset of SIF may occur earlier than GPP due to the earlier activation of the photosystem compared to photosynthesis [204].

The cessation of SIF and GPP in the Arctic-Boreal region is regulated by temperature, soil moisture, and radiation [193, 205,206,207,208]. In ecosystems with weaker radiation limitation, GPP increases more with warming compared to the ecosystems strongly limited by radiation [209]. SIF and GPP have an earlier cessation than canopy greenness due to divergent responses to temperature in the fall [89]. As the Arctic-Boreal region warms, the cessation of growth is projected to be more water-limited [33, 194, 205].

In the long term, the existing satellite SIF measurements do not have sufficiently long records to robustly derive the trends in changing vegetation phenology (Fig. 2c) compared to conventional greenness measurements (e.g., more than 50-year records of Landsat reflectance [210]). Nevertheless, SIF records have been mostly used to validate the long-term trend observed from the vegetation indices of greenness measurements. SIF measurements in the Arctic-Boreal region agree with greenness measurements in showing that warming leads to an earlier and higher peak in GPP [118, 201, 211,212,213]. However, further warming will not continue promoting GPP once the optimal temperature is surpassed [214, 215]. Notably, the peaks of GPP and SIF occur earlier than the peak of greenness, indicating a decoupling of peak timing between photosynthetic rate and canopy greenness [90]. This mismatch between the peaks, which could be a result of delayed canopy development, increases with rising atmospheric CO\(^{2}\) and reduced maximum photosynthetic rate [216].

Across the Arctic-Boreal region, these temporal changes in vegetation phenology are not homogeneous [32]. Madani et al. [118] use CSIF, a downscaled SIF product, to study the spatial pattern of changes in Arctic-Boreal phenology and suggest that such heterogeneous climate sensitivity can be explained at the regional level and characterized by plant functional traits. Incorporating SIF with better-represented plant functional traits in land surface models can improve the estimation of carbon fluxes as vegetation phenology changes.

Detecting Land Cover Changes and Vegetation Compositions

So far, this paper has mainly focused on far-red SIF. However, it is important to note that SIF emissions include both red and far-red wavelengths. Remote sensing instruments often report canopy-scale SIF in far-red wavelengths because of the reabsorption of red SIF by chlorophyll [217]. The ratio between red and far-red SIF at the leaf level can provide additional information about biochemical traits, leaf morphology, and photosynthetic phenology [218, 219]. Therefore, the leaf-level red:far-red SIF ratio may help to identify plant functional type and monitor land cover changes [220]. However, this conclusion may not hold at the canopy level with remotely measured SIF due to the reabsorption of red SIF.

Droughts and Wildfires

Rapid warming in the Arctic-Boreal region is projected to increase the frequency of droughts and wildfires, which may interrupt the increasing trend of GPP and release CO\(^{2}\) from thawing permafrost into the atmosphere. SIF measurements have been used to monitor ecosystem response to these disturbances and recovery.

Droughts

A case study of the 2010 Russian drought [221] uses SIF and greenness measurements to show that the drought impact on GPP through both reduced fPAR and LUE (refer to Eqs. 1 and 2). However, the reduction of LUE dominates the decreasing GPP in forests, while the reduction of fPAR is the main reason for the decreasing GPP in grasslands. Furthermore, Li et al. [222] use SIF as a proxy for photosynthetic phenology and find that Arctic-Boreal forests recovery from droughts not only depends on the severity of droughts but also the relative timing of droughts and vegetation phenology.

More advanced, SIF can help understand the ecosystem response to droughts from an ecohydrology perspective as transpiration and photosynthesis are coupled processes. For example, Recent studies [206, 223,224,225] use SIF to mechanistically constrain the dynamics of transpiration and characterize the seasonal patterns of transpiration in the Arctic-Boreal region, suggesting SIF has the potential to investigate climate drivers of carbon and water cycles simultaneously.

The diurnal dynamics of photosynthesis are important for understanding the coupled carbon-water cycles and ecosystem-climate feedbacks [226]. In the Arctic-Boreal region, the diurnal variations of GPP characterized by tower-based measurements reveal a nonlinear relationship between SIF-GPP in Boreal forests due to light saturation [144], underlying significance of the diurnal SIF measurements for improving the modeling of the vegetation dynamics. Spaceborne instruments, such as TROPOMI, TEMPO, and Sentinel-4, provide great potential for measuring SIF at different times of the day, while TROPOMI is the only instrument that covers the entire Arctic-Boreal region. The diurnal pattern of TROPOMI SIF is consistent with GPP [126] indicating the potential of evaluating diurnal variations of carbon-water cycles from space [227].

To comprehensively characterize biochemical and biophysical processes, combining multi-modal remote sensors (including SIF instruments) and addressing different components of the carbon-water cycles are beneficial. For example, using visible-near infrared reflectance (including SIF), microwave, lidar, and thermal imaging to simultaneously monitor photosynthesis, ecohydrology, canopy structure, and water stress in respective. For future reference in the Arctic-Boreal region, applying this multi-modal concept is available from space [228,229,230], aircraft [231], and towers [232].

Wildfires

In the case of wildfires, applying SIF to monitoring the change in GPP can be challenging because the SIF-GPP relationship may change if wildfires alter the plant composition as discussed in Section 3.1. Madani et al. [213] compare burned and unburned areas with the same land cover types and show that SIF-driven GPP has shown a faster recovery after wildfires happened, although there is an instantaneous decrease in GPP during wildfires [213].

Conclusion and Future Directions

This paper comprehensively summarizes cross-scale SIF instruments and new insights gained regarding carbon fluxes by using SIF in the Arctic-Boreal region. Through a thorough review of existing and upcoming satellite missions/instru-ments, this paper highlights the complementary nature of spaceborne, airborne, and tower-based SIF measurements, which collectively enable wide-ranging spatiotemporal coverage and resolutions in the Arctic-Boreal region. These cross-scale SIF measurements provide new insights into long-term variations and spatial patterns of photosynthetic dynamics and the carbon uptake from subdaily to seasonal scales in the Arctic-Boreal region, which are key to mechanistically constraining the carbon fluxes in land surface models.

For future references, it is important to acknowledge that current SIF measurements in the Arctic-Boreal region are still limited and their spatial resolutions are too coarse for the heterogeneous land cover. Overcoming these limitations requires fostering an extensive SIF observational network to improve land surface modeling in the Arctic-Boreal region and validate existing quantitative models. Integration of multi-modal instruments that combine SIF and other (a)biotic variables holds promise to comprehensively represent the diverse plant functional traits and their climate sensitivity with climate change. Such data synergies and model improvements are becoming possible as both NASA ABoVE and the Department of Energy’s Next Generation Ecological Experiment in the Arctic (DOE NGEE-Arctic) are towards their ending phases, which focus on data synthesis and modeling. Taken together, SIF still has great potential for advancing our understanding of the ecosystem-climate feedbacks and projecting climate change impact across the Arctic-Boreal region.