Remote sensing of canopy light use efficiency in temperate and boreal forests of North America using MODIS imagery
Highlights
► The potential of EVI as a proxy of LUE depends on the standard deviation in EVI. ► A new LUE algorithm is proposed for temperate and boreal forests in North America. ► The RMSE of the new model for ten independent sites is 0.055 g C mol− 1 PAR. ► A 28% improvement is observed in GPP using this LUE algorithm compared to MODIS GPP.
Introduction
Forest ecosystems play an important role in global carbon sequestration (Beer et al., 2010, Zhao and Running, 2010). Annual global carbon uptake by vegetation, also referred to as gross primary production (GPP), is around 123 ± 8 petagrams of carbon per year (Beer et al., 2010). However, substantial variations in GPP are observed among different models and ecoregions, both at plant and stand levels, and these discrepancies illustrate the limits to our full understanding of the global carbon cycle.
Light use efficiency (LUE), defined as the amount of carbon fixed in photosynthesis per unit of absorbed solar radiation, is an important variable for the estimation of GPP from satellite inputs when using the Monteith equation (Monteith, 1972):where APAR is the absorbed photosynthetically active radiation calculated as the product of an absorbed fraction (fAPAR) and the amount of incident photosynthetically active radiation (PAR).
Typical methods for the simulation of LUE require prior specification of a maximum LUE (ε0) for a specific land cover type and additional input of climate variables (e.g., temperature, water stress) representing canopy stresses that modulate this maximum LUE (Running et al., 2004). This method has been useful in productivity models, such as the Moderate Resolution Imaging Spectroradiometer (MODIS) GPP product (Running et al., 2004) and the Vegetation Photosynthesis Model (Xiao et al., 2004). However, the dependence on input climate variables and biome-scale maximum LUE can cause significant deviation from observed GPP (Heinsch et al., 2006, Mu et al., 2011, Zhao et al., 2006).
With ongoing improvements in our ability to remotely sense the land surface, there has been an increase in efforts to directly infer LUE using these observations. For example, the photochemical reflectance index (PRI), defined as a normalized difference index using reflectance at 531 and 570 nm, is suggested to have potential in tracking LUE both at the leaf scale based on ground spectral measurements (Filella et al., 2009, Gamon et al., 1997) and satellite observations (Drolet et al., 2005, Drolet et al., 2008, Goerner et al., 2009, Hall et al., 2008, Hall et al., 2011, Hilker et al., 2009). However, PRI shows high sensitivity to various extraneous effects such as canopy structure and the view observer geometry, which prevents its use at landscape and global scales and requires the appropriate upscaling algorithms to account for structural differences in vegetation (Hilker et al., 2010). Hilker et al. (2008) shows that while isotropic PRI scattering is correlated to LUE variation, geometric scattering can be attributed to canopy level shading. Therefore, remote sensing of forest LUE from space would be achieved by measuring PRI as a function of shadow fraction using multi-angle observations (Hall et al., 2008), which is further confirmed by the relationship between spaceborne PRI and canopy shadow fractions (Hilker et al., 2009). A theoretical concept using a canopy reflectance model proposed by Hall et al. (2011) recently further validates that using PRI alone to predict canopy LUE is confounded by the shadow fraction viewed by the sensor.
A thorough analysis of PRI is shown in Garbulsky et al. (2011), indicating that calibration of the PRI-LUE relation across biomes and a careful attention to potentially confounding factors are both needed for future improvement. Apart from PRI, Inoue et al. (2008) demonstrate that a number of bands centered at the red edge and near-infrared ranges also have potential in deriving indicators of LUE for a wheat canopy. Correlations are reported between LUE and a number of vegetation indices, including the normalized difference vegetation index (NDVI, Rouse et al., 1974) and the enhance vegetation index (EVI, Huete et al., 2002) in crop (Wu et al., 2010, Wu et al., 2009) and peatland ecosystems (Schubert et al., 2010). However, the potential of a certain vegetation index as a proxy of LUE across biomes and the underlying mechanism are still unknown (Huemmrich et al., 2010, Zhang et al., 2009), probably due to a number of influencing factors, including temperature, soil moisture, vapor pressure deficit (VPD) and light quality (Coops et al., 2010).
Two specific aspects of known variations in LUE need to be considered for use in operational algorithms related to the use of vegetation indices. First, environmental effects of temperature and water stress considerably modify LUE and cannot be interpreted by a single vegetation index. These environmental controls likely limit the competency of productivity models that incorporate vegetation indices under drought conditions (Mu et al., 2011, Sims et al., 2008, Sims et al., 2006). A second uncertainty is the application of only a single vegetation index in evergreen biomes that show low dynamic ranges in greenness but potentially large variations in LUE (Garbulsky et al., 2008, Nakaji et al., 2008). Hence, environmental variables are potentially helpful in resolving these present limitations in remotely sensed LUE. A possible candidate among those climate variables is air temperature (Ta) because of its importance in influencing the magnitude and timing of plant growth (Chen et al., 2003), ecosystem respiration (Tang et al., 2008), LUE (Schwalm et al., 2006), and its correlation with other environmental variables, such as vapor pressure deficit (VPD) and PAR (Sims et al., 2008). This potential has been demonstrated in previous studies, such as the MODIS GPP product which uses temperature and VPD to reduce LUE under unfavorable conditions (Running et al., 2004, Zhao et al., 2006). However, the dependence on the requirements of meteorological inputs at desired temporal and spatial scales generally limits its global application (Mu et al., 2011).
With the availability of global carbon flux data that can be used to calculate canopy level LUE for multiple ecosystems (Baldocchi, 2008), it is possible to validate and compare the candidate LUE models and their impacts on GPP for a broad array of forest types. Here we present a methodology for estimating monthly LUE using MODIS observations and compare simulated values against LUE derived from flux measurements obtained from multiple temperate and boreal forest sites within North America. MODIS-derived EVI and land surface temperature (LST) are examined for their potential in estimating monthly LUE across biomes. The objectives of this study are: (1) to analyze the potential of EVI in evaluating monthly LUE for both deciduous forests and evergreen forests, (2) to derive a new model that can provide better estimates of monthly LUE using the MODIS EVI and LST observations, and (3) to show the usefulness of the new LUE algorithm in the estimation of forest GPP. This effort could result in an improved algorithm for remote sensing of GPP and aid in understanding of terrestrial carbon cycle-climate feedbacks in forested ecosystems.
Section snippets
Study sites
We focused on twenty forest sites in the North American flux networks, including twelve deciduous forests (DF) and eight evergreen forests (EF) (Fig. 1). Half of these sites, composed of five DF and five EF, were used to derive the LUE model and the remaining sites were used for validation. Detailed descriptions of these sites are shown in Table 1.
Flux and meteorological measurements
Flux data for the nine Canadian sites were downloaded from the Fluxnet-Canada Data Information System (http://www.fluxnet-canada.ca) while data for
Relationship between LUE and EVI
The relationship between all tower-derived monthly LUE and remotely sensed EVI reveals a relationship with a coefficient of determination (R2) equal to 0.46 (Fig. 2a) with evident differences among forest functional types. An R2 of 0.62 was found for deciduous forest (DF) sites, suggesting that EVI can better simulate the dynamics of LUE for ecosystems with wider dynamical ranges in EVI. For example, EVI values for DF sites generally fluctuated between 0.15 and 0.75 with a mean standard
Application of this LUE algorithm to estimate GPP
When the new model was applied to estimate GPP, we also observed a better performance than the standard MODIS GPP products. An R2 of 0.73 (p < 0.001) and an overall RMSE of 52.7 g C m− 2 month− 1 are obtained between the flux-measured GPP and the MODIS GPP products for all sites (Fig. 9a). Similar to our findings, MODIS GPP also shows better suitability for DF sites (R2 = 0.71, RMSE = 60.7 g C m− 2 month− 1) than for EF sites (R2 = 0.67, RMSE = 35.7 g C m− 2 month− 1). For all observations, there is a clear pattern for
Conclusions
Light use efficiency is an important variable characterizing plant eco-physiological function and has been widely used to estimate forest productivity from remotely sensed data. By incorporation of both canopy greenness and temperature, here we propose a new algorithm of EVI × Tm that shows reasonably good estimates of monthly LUE in various forest ecosystems in North America. We also find that a simple metric based on the average of EVI and Tmin allows for calibration of model coefficients.
Acknowledgements
We like to thank Dr. Andrew Richardson for the suggestions to the initial manuscript. Though comments from both reviewers are also appreciated. This work used data from flux sites from both the AmeriFlux and Fluxnet-Canada and we appreciate the flux PIs providing these valuable data and helpful explanations. This work was funded by an NSERC Strategic Grant (381474-09), the National Natural Science Foundation (Grant No. 41001210), the Knowledge Innovation Program of CAS (KZCX2-EW-QN302), and the
References (69)
- et al.
Carbon, energy and water fluxes at mature and disturbed forest sites, Saskatchewan, Canada
Agricultural and Forest Meteorology
(2006) - et al.
Soil-plant nitrogen cycling modulated carbon exchanges in a western temperate conifer forest in Canada
Agricultural and Forest Meteorology
(2006) - et al.
Interannual variability in the leaf area index of a boreal aspen-hazelnut forest in relation to net ecosystem production
Agricultural and Forest Meteorology
(2004) - et al.
Assessing eddy-covariance flux tower location bias across the Fluxnet-Canada Research Network based on remote sensing and footprint modelling
Agricultural and Forest Meteorology
(2011) - et al.
Utilizing in situ directional hyperspectral measurements to validate bio-indicator simulations for a corn crop canopy
Ecological Informatics
(2010) - et al.
Using LiDAR and quickbird data to model plant production and quantify uncertainties associated with wetland detection and land cover generalizations
Remote Sensing of Environment
(2009) - et al.
Carbon exchange and venting anomalies in an upland deciduous forest in northern Wisconsin, USA
Agricultural and Forest Meteorology
(2004) - et al.
Biometric and eddy-covariance based estimates of annual carbon storage in five eastern North American deciduous forests
Agricultural and Forest Meteorology
(2002) - et al.
Comparing net ecosystem exchange of carbon dioxide between an old-growth and mature forest in the upper Midwest, USA
Agricultural and Forest Meteorology
(2005) - et al.
Cross site evaluation of eddy covariance GPP and RE decomposition techniques
Agricultural and Forest Meteorology
(2008)
A MODIS-derived photochemical reflectance index to detect inter-annual variations in the photosynthetic light-use efficiency of a boreal deciduous forest
Remote Sensing of Environment
Regional mapping of gross light-use efficiency using MODIS spectral indices
Remote Sensing of Environment
Evaluation of MODIS LAI, fAPAR and the relation between fAPAR and NDVI in a semi-arid environment using in situ measurements
Remote Sensing of Environment
The photochemical reflectance index (PRI) and the remote sensing of leaf, canopy and ecosystem radiation use efficiencies. A review and meta-analysis
Remote Sensing of Environment
Ecosystem-level CO2 fluxes from a boreal cutover in eastern Canada before and after scarification
Agricultural and Forest Meteorology
Tracking seasonal drought effects on ecosystem light use efficiency with satellite-based PRI in a Mediterranean forest
Remote Sensing of Environment
Advances in satellite remote sensing of environmental variables for epidemiological applications
Advances in Parasitology
PHOTOSYNSAT, photosynthesis from space: Theoretical foundations of a satellite concept and validation from tower and spaceborne data
Remote Sensing of Environment
Multi-angle remote sensing of forest light use efficiency by observing PRI variation with shadow fraction
Remote Sensing of Environment
Separating physiologically and directionally induced changes in PRI using BRDF models
Remote Sensing of Environment
Remote sensing of photosynthetic light-use efficiency across two forested biomes: Spatial scaling
Remote Sensing of Environment
An assessment of photosynthetic light use efficiency from space: Modeling the atmospheric and directional impacts on PRI reflectance
Remote Sensing of Environment
Remote sensing of tundra gross ecosystem productivity and light use efficiency under varying temperature and moisture conditions
Remote Sensing of Environment
Overview of the radiometric and biophysical performance of the MODIS vegetation indices
Remote Sensing of Environment
Normalized difference spectral indices for estimating photosynthetic efficiency and capacity at a canopy scale derived from hyperspectral and CO2 flux measurements in rice
Remote Sensing of Environment
Refining light-use efficiency calculations for a deciduous forest canopy using simultaneous tower-based carbon flux and radiometric measurements
Agricultural and Forest Meteorology
Improvements to a MODIS global terrestrial evapotranspiration algorithm
Remote Sensing of Environment
Utility of spectral vegetation indices for estimation of light conversion efficiency in coniferous forests in Japan
Agricultural and Forest Meteorology
Impact of nutrients on peatland GPP estimations using MODIS time series data
Remote Sensing of Environment
Photosynthetic light use efficiency of three biomes across an east–west continental-scale transect in Canada
Agricultural and Forest Meteorology
A new model of gross primary productivity for North American ecosystems based solely on the enhanced vegetation index and land surface temperature from MODIS
Remote Sensing of Environment
Exploring the potential of MODIS EVI for modeling gross primary production across African ecosystems
Remote sensing of environment
Ecosystem respiration and its components in an old-growth forest in the Great Lakes region of the United States
Agricultural and Forest Meteorology
Eight years of carbon dioxide exchange above a mixed forest at Borden, Ontario
Agricultural and Forest Meteorology
Cited by (51)
How well do light-use efficiency models capture large-scale drought impacts on vegetation productivity compared with data-driven estimates?
2023, Ecological IndicatorsCitation Excerpt :The main LUE models include VPM, CASA, MOD17, TL-LUE, etc. The main differences among LUE models are the different ways of characterizing the moisture and temperature factors (Wu et al., 2012). Although LUE models are widely applied to estimating regional to global GPP, they suffer from various uncertainties, mainly including the accuracy of the model input data, the variation of the maximum light use efficiency, and the inability to accurately simulate the moisture factor (Anav et al., 2015).
Evolution of light use efficiency models: Improvement, uncertainties, and implications
2022, Agricultural and Forest MeteorologyUsing remote sensing to identify the peak of the growing season at globally-distributed flux sites: A comparison of models, sensors, and biomes
2021, Agricultural and Forest MeteorologyCitation Excerpt :Data for the flux site locations were downloaded from the MODIS product, MOD13A2 Version 6, which has a 16-day interval and 1 km spatial resolution (https://earthdata.nasa.gov/), and from the SPOT-VGT product, S10, which has a temporal resolution of 10 days and a spatial resolution of 1 km (https://www.vito-eodata.be/PDF/portal/Application.html#Home). We followed the rules in Wu et al. (2012) to obtain the NDVI for each flux site. Both products were systematically corrected for topographic and atmospheric effects.
Comparison of satellite-based models for estimating gross primary productivity in agroecosystems
2021, Agricultural and Forest MeteorologyAssessments of gross primary productivity estimations with satellite data-driven models using eddy covariance observation sites over the northern hemisphere
2020, Agricultural and Forest MeteorologyCitation Excerpt :Regarding aforementioned models, remotely sensed data provide spatially continuous estimates of photosynthesizing vegetation cover (Boisvenue and White, 2019; Ryu et al., 2019). LUE models use remotely sensed data to estimate the absorbed photosynthetically active radiation (Running et al., 2000; Wu et al., 2012), while VI-based models usually use remotely sensed data to describe the seasonal dynamics of leaf phenology (Sims et al., 2008, 2006). In process-based models, remotely sensed data are the key model inputs used to describe vegetation structure (Chen et al., 1999; Liu et al., 1997) and atmospheric radiative transfer processes (Pinker and Laszlo, 1992; Yan et al., 2018; Yin et al., 2018).
Evaluating and comparing remote sensing terrestrial GPP models for their response to climate variability and CO <inf>2</inf> trends
2019, Science of the Total Environment