Estimation Of Gross Primary Production Using Sun-induced Chlorophyll Fluorescence

Carbon dioxide fixed through photosynthesis by terrestrial vegetation is known as gross primary production(GPP)at the ecosystem level.Terrestrial GPP is the largest global carbon flux,and it drives several ecosystem functions.Hence,accurately estimating the spatio-temporal patterns of GPP is not only essential for quantifying the feedbacks between atmosphere and biosphere,but also significant for better understanding carbon cycle and coping with climate change.As a by-product of photosynthesis,sun-induced chlorophyll fluorescence(SIF)is a better indicator of photosynthetic activity than traditional vegetation indices(VIs).Recent studies have found a strong linear relationship between satellite SIF and GPP for specific ecosystem types,and demonstrated that SIF can provide a direct measure of ecosystem GPP.However,the correlation between GPP and SIF at the canopy scale is highly affected by environmental stress,canopy structure and plant functional type.The mechanistic link between GPP and remotely-sensed SIF still remains unclear.Therefore,the objectives of this study are to analyze the sensitivity of GPP-SIF relationship to plant photosynthetic physiology,to compare the ability of red SIF and far-red SIF in estimating GPP,and to model and evaluate the global terrestrial SIFbased GPP product.The major research contens and results are as follows:1)We firstly evaluated the potential of SIF760 to track the diurnal variations in GPP and investigated whether the GPP-SIF760 relationship was dependent on the type of photosynthesis.Based on synchronous observations of GPP and SIF760,we found that the GPP and SIF760 exhibited similar diurnal patterns to those of PAR in that lower values occurred in the early morning and late afternoon and higher values were observed around midday and SIF760 successfully captured the diurnal dynamics of plant photosynthetic processes.Our results also demonstrated that SIF760 showed a good linear correlation with GPP for both C3 and C4 crops,which confirms the capability of remotely sensed SIF signals to act as a direct proxy for GPP.However,it was also shown that the slope of the GPP-SIF760 relationship for C4 crops was 1.16 times higher than C3 crops,due to the similar ? but large difference in ? for C3 and C4 crops(if the ‘midday depression' point pairs were removed,the C4 slope was 0.82 times higher than C3 crops).Therefore,C3 and C4 species should be considered separately when using SIF to model GPP by applying the linear scaling approach(GPP=?_P/?_F×SIF).2)In this paper,we also compared the ability of red SIF and farred SIF in estimating GPP based on simulations and observations.Many theoretical studies and experiments have found the difference of red SIF and farred SIF in monitoring the physiological state and their different properties of absorption and scattering by canopies.However,there still no agreement on which band is a better proxy for GPP.We used SCOPE model simulations to produce diurnal variation datasets of GPP,SIF760 and SIF688 with ranging leaf chlorophyll content(Cab)and leaf area index(LAI).The results showed that the relationship between GPP and SIF760 can still be expressed by a unified linear regression model;however,the relationship between GPP and SIF688 was sensitive to Cab and LAI,and presented like a “wing” shape.As a good measure of canopy architechture,the functions of VIs(especially NDVIrededge?CIrededge and NDVI×WDRVIrededge)provided rational approximations of fesc(688),which accounted for the fraction of SIF photons escaping the canopy,and lowered the sensitivity to Cab and LAI.Therefore,the GPP-SIF688×f(VI)linear regression model performed as well as GPP-SIF linear model for the winter wheat canopies.Considering the greater challenge involved in the retrieval and reabsorption of red SIF,far-red SIF may currently be a better method to assess terrestrial photosynthesis.3)Previous studies have demonstrated that remote sensing of SIF can provide a direct estimation of GPP.However,it remains very weak in global GPP modeling based on satellite SIF.Based on the GPPEC and GOME-2 SIF740 retrievals,we established the linear GPPEC-SIF740 regression model for each vegetation type,and obtained the global terrestrial GPP product spanning the period 2007-2014.We compared the performance of GPPSIF with GPP estimates from data-driven model and(MODIS)MOD17 GPP.Overall,GPPSIF data showed a good agreement with MOD17 GPP in spatiotemporal patterns.Moreover,the algorithm of GPPSIF is simple and not dependency on climate input and satellite-based vegetation indices.However,the accuracy of GPPSIF is up to data quality and the availability of flux data,and the accuracy of input data determines the accuracy of GPPMOD.There remains a degree of uncertainty in savannas,shrublands and evergreen broadleaf forests for both products.In addition,we compared the GPPEC-SIF740 relationship between C3 and C4 crops at the landscape scale.The linear slope for C4 crops was 1.81 times the value of that for C3 crops,which was consistent with the results derived from diurnal varation experiments.Therefore,the GPP-SIF relationship is also sensitive to photosynthetic pathway type at the landscape scale.It has further confirmed that the spatio-temporal dynamics of C3 and C4 plants are of great importance in accurately quantifying the global carbon uptakes.The innovation of the paper can be listed below:1)This paper has illustrated the different characteristics in GPP-SIF relationship between C3 and C4 crops based on the theoretical derivation and observations.In order to quantitatively analyze the potential of SIF in modeling GPP for C3 and C4 crops,we conducted a series of diurnal variation experiments and investigated the relationship between GOME-2 SIF740 and EC-based GPP.We found that GPP values of C4 crops were twice the value of C3 crops when they had the same SIF values.Our results are consistent with the theoretical derivation based on GPP-ETR model.2)Combined with the SCOPE simulations and in-situ observations,we have quantitatively compared the ability of red SIF and farred SIF in estimating GPP.Since red SIF is difficult to achieve,most remote sensing applications of SIF focus on farred SIF.Based on SCOPE simulations and observations by using custommade QE Pro spectrometer(SR=0.3 nm),we found that farred SIF was less sensitive to canopy structure and may currently be a better method to assess photosynthetic carbon uptakes.3)Based on the EC-based GPP and GOME-2 SIF740 retrievals,we have established the linear GPP-SIF regression model for each vegetation type,and obtained the global terrestrial GPPSIF product.Although remote sensing of SIF has been paid more attentions,there remains no global terrestrial GPP products based on satellite SIF.GPPSIF dataset showed a good agreement with MOD17 GPP in spatiotemporal patterns,and the algorithm of GPPSIF is simple and not dependency on climate input and satellite-based vegetation indices.Our study will provide experimental and theoretical basis for a better mechanistic understanding of the relationship between GPP and SIF at different spatial and temporal scales.Additionally,the SIF-based GPP product can be further used in a wide variety of applications,especially in those studies related to the evaluation of ecological service value,model comparison,and the carbon cycle under global climate change.