Satellite-based GPP model

WY Wenping Yuan YZ Yi Zheng SP Shilong Piao PC Philippe Ciais DL Danica Lombardozzi YW Yingping Wang YR Youngryel Ryu GC Guixing Chen WD Wenjie Dong ZH Zhongming Hu AJ Atul K. Jain CJ Chongya Jiang EK Etsushi Kato SL Shihua Li SL Sebastian Lienert SL Shuguang Liu JN Julia E.M.S. Nabel ZQ Zhangcai Qin TQ Timothy Quine SS Stephen Sitch WS William K. Smith FW Fan Wang CW Chaoyang Wu ZX Zhiqiang Xiao SY Song Yang

This protocol is extracted from research article:

Increased atmospheric vapor pressure deficit reduces global vegetation growth

**
Sci Adv**,
Aug 14, 2019;
DOI:
10.1126/sciadv.aax1396

Increased atmospheric vapor pressure deficit reduces global vegetation growth

Procedure

We used two satellite-based GPP models to investigate the impacts of VPD on vegetation GPP. The first model is the MODIS-GPP model, and this study used long-term global MODIS GPP dataset driven by GIMMS fPAR data (*27*).

The second model is the revised EC-LUE model (*26*), derived by (i) integrating the impact of atmospheric CO_{2} concentration on GPP and (ii) adding the limit of VPD to GPP. The revised EC-LUE model simulates terrestrial ecosystem GPP as$$\text{GPP}=\text{PAR}\times \text{fPAR}\times {\mathrm{\epsilon}}_{\text{max}}\times {C}_{\mathrm{s}}\times \text{min}({T}_{\mathrm{s}},{W}_{\mathrm{s}})$$(10)where PAR is the incident photosynthetically active radiation (MJ m^{−2}) per time period (e.g., day); fPAR is the fraction of PAR absorbed by the vegetation canopy calculated by the GIMMS3g NDVI dataset; ε_{max} is the maximum LUE; *C*_{s}, *T*_{s}, and *W*_{s} represent the downward-regulation scalars for the respective effects of atmospheric CO_{2} concentration ([CO_{2}]), temperature (*T*_{a}), and atmospheric water demand (VPD) on LUE; and min denotes the minimum value of *T*_{s} and *W*_{s}.

The effect of atmospheric CO_{2} concentration on GPP was calculated according to Farquhar *et al*. (*44*) and Collatz *et al*. (*45*)$${C}_{\mathrm{s}}=\frac{{C}_{\mathrm{i}}-\mathrm{\theta}}{{C}_{\mathrm{i}}+2\mathrm{\theta}}$$(11)$${C}_{\mathrm{i}}={C}_{\mathrm{a}}\times \mathrm{\chi}$$(12)where θ is the CO_{2} compensation point in the absence of dark respiration (ppm) and *C*_{i} is the CO_{2} concentration in the intercellular air spaces of the leaf (ppm), which is the product of atmospheric CO_{2} concentration (*C*_{a}) and the ratio of leaf internal to ambient CO_{2} (χ). χ is estimated (*46*–*48*) by$$\mathrm{\chi}=\frac{\mathrm{\gamma}}{\mathrm{\gamma}+\sqrt{\text{VPD}}}$$(13)$$\mathrm{\gamma}=\sqrt{\frac{356.51K}{1.6\mathrm{\eta}*}}$$(14)$$K={K}_{\mathrm{c}}(1+\frac{{P}_{0}}{{K}_{0}})$$(15)$${K}_{\mathrm{c}}=39.97\times {\mathrm{e}}^{\frac{79.43\times ({T}_{\mathrm{a}}-298.15)}{298.15R{T}_{\mathrm{a}}}}$$(16)$${K}_{\mathrm{o}}=27480\times {\mathrm{e}}^{\frac{36.38\times ({T}_{\mathrm{a}}-298.15)}{298.15R{T}_{\mathrm{a}}}}$$(17)where *K*_{c} and *K*_{o} are the Michaelis-Menten coefficient of Rubisco for carboxylation and oxygenation, respectively, expressed in partial pressure units, and *P*_{o} is the partial pressure of O_{2} (ppm). *R* is the molar gas constant (8.314 J mol^{−1} K^{−1}), and η* is the viscosity of water as a function of air temperature (*49*).

*T*_{s} and *W*_{s} were calculated using the following equations$${T}_{\mathrm{s}}=\frac{({T}_{\mathrm{a}}-{T}_{\text{min}})\times ({T}_{\mathrm{a}}-{T}_{\text{max}})}{({T}_{\mathrm{a}}-{T}_{\text{min}})\times ({T}_{\mathrm{a}}-{T}_{\text{max}})-({T}_{\mathrm{a}}-{T}_{\text{opt}})\times ({T}_{\mathrm{a}}-{T}_{\text{opt}})}$$(18)$${W}_{\mathrm{s}}=\frac{{\text{VPD}}_{0}}{\text{VPD}+{\text{VPD}}_{0}}$$(19)where *T*_{min}, *T*_{max}, and *T*_{opt} are the minimum, maximum, and optimum air temperature (°C) for photosynthetic activity, which were set to 0°, 40°, and 20.33°C, respectively (*50*). VPD_{0} is an empirical coefficient of the VPD constraint equation.

Parameters ε_{max}, θ, and VPD_{0} were calibrated using estimated GPP at EC towers (table S7). The nonlinear regression procedure (Proc NLIN) in the Statistical Analysis System (SAS; SAS Institute Inc., Cary, NC, USA) was applied to estimate the three parameters in the revised EC-LUE model. The revised EC-LUE model was calibrated at 50 EC towers and validated at 41 different towers (table S3). The results showed good model performance of the revised EC-LUE model for simulating biweekly GPP variations (fig. S13). To estimate global GPP, EC-LUE and MODIS models used the MERRA dataset (i.e., air temperature, VPD, PAR). Because of the different model algorithm, GIMMS3g NDVI and fPAR products were used to indicate vegetation conditions for EC-LUE and MODIS, respectively.

We performed two types of experimental simulation to evaluate the relative contribution of three main driving factors: CO_{2} fertilization, climate change, and satellite-based NDVI/fPAR changes. The first simulation experiment (S_{ALL}) was a normal model run, and all drivers were set to change over time to examine the responses of GPP to all environmental changes, including climate, atmospheric [CO_{2}], and NDVI/fPAR. The second type of simulation experiments (S_{CLI0}, S_{NDVI0}, and S_{CO20}) allowed two driving factors to change with time while holding the third constant at an initial baseline level. For example, the S_{CLI0} simulation experiment allowed NDVI and atmospheric [CO_{2}] to change with time, while climate variables were held constant at 1982 values. S_{NDVI0} and S_{CO20} simulation experiments kept NDVI and atmospheric [CO_{2}] constant at 1982 values and varied the other two variables.

We considered the differences between simulation results of the first type (S_{ALL}) and second type (S_{CO20} and S_{NDVI0}) of experiments to estimate the sensitivity of GPP to atmospheric [CO_{2}] (β_{CO2}) and NDVI/fPAR (β_{NDVI}). β_{CO2} and β_{NDVI} were calculated on the basis of the following equations$${\mathrm{\Delta}\text{GPP}}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{CO}20})\mathrm{i}}={\mathrm{\beta}}_{\text{CO}2}\times \mathrm{\Delta}\text{CO}{2}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{CO}20})\mathrm{i}}+\mathrm{\epsilon}$$(20)$${\mathrm{\Delta}\text{GPP}}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{NDVI}0})\mathrm{i}}={\mathrm{\beta}}_{\text{NDVI}}\times \mathrm{\Delta}{\text{NDVI}}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{NDVI}0})\mathrm{i}}+\mathrm{\epsilon}$$(21)where ΔGPP_{i}, ΔCO_{2i}, and ΔNDVI_{i} represent the differences of GPP simulations, atmospheric [CO_{2}], and NDVI between two model experiments from 1982 to 2015, and ε is the residual error term.

A multiple regression approach was used to estimate GPP sensitivities to three climate variables: air temperature (β_{Ta}), VPD (β_{VPD}), and PAR (β_{PAR})$${\mathrm{\Delta}\text{GPP}}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{CLI}0})\mathrm{i}}={\mathrm{\beta}}_{\text{Ta}}\times \mathrm{\Delta}{{T}_{\mathrm{a}}}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{CLI}0})\mathrm{i}}+{\mathrm{\beta}}_{\text{VPD}}\times {\mathrm{\Delta}\text{VPD}}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{CLI}0})\mathrm{i}}+{\mathrm{\beta}}_{\text{PAR}}\times {\mathrm{\Delta}\text{PAR}}_{({\mathrm{S}}_{\text{ALL}}-{\mathrm{S}}_{\text{CLI}0})\mathrm{i}}+\mathrm{\epsilon}$$(22)where ΔTa_{i}, ΔVPD_{i} and ΔPAR_{i} represent the differences of air temperature, VPD, and photosynthetically active radiation time series between two model experiments (S_{ALL} and S_{CLI0}), respectively. The regression coefficient β was estimated using maximum likelihood analysis.

The EC-LUE model suggested a CO_{2} sensitivity (β_{CO2}) of 19.01 ± 4.01 Pg C 100 ppm^{−1} (Fig. 6B and table S2), which indicates a 15.7% increase of GPP with a rise of atmospheric [CO_{2}] of 100 ppm. Our estimate is close to CO_{2} sensitivity derived from ecosystem models (β_{CO2} = 21.92 ± 4.55 Pg C 100 ppm^{−1}; Fig. 6B) and is comparable to the observed response of NPP (net primary production) to the increased CO_{2} at the FACE experiment locations (13% per 100 ppm) and estimates of other ecosytem models (5 to 20% per 100 ppm) (*51*).

A machine learning method (i.e., RF) was used to model the effects of VPD on NDVI. RF combines tree predictors such that each tree depends on the values of a random vector that is sampled independently, with the same distribution for all trees in the forest. We constructed RF models for simulating annual growing season mean NDVI at each pixel driven by air temperature, precipitation, radiation, wind speed, atmospheric [CO_{2}] concentration, and VPD. The training data were the GIMMS3g NDVI dataset from 1982 to 2015. The R package “randomForest” used in the study was modified by A. Liaw and M. Wiener from the original Fortran by L. Breiman and A. Cutler (https://cran.r-project.org/web/packages/randomForest/).

The RF model was driven by all variables (climate and atmospheric [CO_{2}]) changing over time (RF_{ALL}), and two factorial simulations of NDVI (RF_{CO20} and RF_{CLI0}) were produced by holding one driving factor (climate or atmospheric [CO_{2}]) constant at its initial level (first year of data) while allowing the other driving to change with time. The RF_{CLI0} simulation experiment allowed atmospheric [CO_{2}] other than climate variables to vary since 1982. RF_{CO20} simulation experiments kept atmospheric [CO_{2}] constant at 1982 values and varied the climate variables. At each pixel, we selected 33 years of NDVI observations out of the total 34 years (1982–2015) to develop the RF model, and the remaining 1 year of NDVI observations was used for cross-validation. The model was run 34 times to ensure that the data of each year can be selected to do model validation. The simulated NDVI of three model experiments (i.e., RF_{ALL}, RF_{CO20}, and RF_{CLI0}) are mean values of all 34 times simulations. The simulated NDVI only from the validation year constitutes the RF_{VLI} dataset, which was used to examine the performance of random forest for reproducing NDVI. The simulated NDVI of RF_{VLI} matched the GIMMS3g NDVI very well (fig. S7), and the correlation coefficient (*R*^{2}) is larger than 0.90 at the 88% vegetated areas globally. The tropical forest showed the relative low *R*^{2}. The relative predictive errors range from −1.2 to 1.04% globally and imply that the RF model can accurately simulate interannual variability and magnitude of NDVI.

On the basis of the three model experiments, we used the same method above shown at Eqs. 20 and 22 to estimate the sensitivity of NDVI to atmospheric [CO_{2}] (δ_{CO2}) and five climate variables: air temperature (δ_{Ta}), VPD (δ_{VPD}), PAR (δ_{PAR}), precipitation (δ_{Prec}), and wind speed (δ_{WS})$$\mathrm{\Delta}{\text{NDVI}}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CO}20})\mathrm{i}}={\mathrm{\delta}}_{\text{CO}2}\times \mathrm{\Delta}\text{CO}{2}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CO}20})\mathrm{i}}+\mathrm{\epsilon}$$(23)$$\begin{array}{cc}\hfill \mathrm{\Delta}{\text{NDVI}}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CLI}0})\mathrm{i}}& ={\mathrm{\delta}}_{\text{Ta}}\times \mathrm{\Delta}{{T}_{\mathrm{a}}}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CLI}0})\mathrm{i}}+{\mathrm{\delta}}_{\text{VPD}}\hfill \\ & \times {\mathrm{\Delta}\text{VPD}}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CLI}0})\mathrm{i}}+{\mathrm{\delta}}_{\text{PAR}}\hfill \\ & \times {\mathrm{\Delta}\text{PAR}}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CLI}0})\mathrm{i}}+{\mathrm{\delta}}_{\text{Prec}}\hfill \\ & \times \mathrm{\Delta}{\text{Prec}}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CLI}0})\mathrm{i}}+{\mathrm{\delta}}_{\text{WS}}\hfill \\ & \times {\mathrm{\Delta}\text{WS}}_{({\text{RF}}_{\text{ALL}}-{\text{RF}}_{\text{CLI}0})\mathrm{i}}+\mathrm{\epsilon}\hfill \end{array}$$(24)where Δ represents the differences of NDVI simulations, atmospheric [CO_{2}], and climate variables between two model experiments from 1982 to 2015, and ε is the residual error term. The regression coefficient δ was estimated using maximum likelihood analysis. We quantified the contributions of atmospheric [CO_{2}] and five climate variables to NDVI changes during the two periods (1982–1998 and 1999–2015) by multiplying the magnitude of their changes and sensitivity of NDVI (δ).

Note: The content above has been extracted from a research article, so it may not display correctly.

Q&A

Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.