Interannual variability and trends of gross primary production and transpiration in savannas and grasslands from 2000 to 2021

Abstract

Carbon and water fluxes of savannas and grasslands have large seasonal dynamics and inter-annual variation. In this study, we selected five savanna and grassland sites, each of them having 10+ years (11−21 years) of eddy covariance (EC) data, and a total of 85 site-years at these five sites which offers a unique opportunity for data analyses and model evaluation. We ran a long-term simulation (2000−2021) of the vegetation photosynthesis model (VPM, v3.0) and vegetation transpiration model (VTM, v2.0) to investigate the seasonal dynamics, interannual variation, and decadal trends of modeled gross primary production (GPP_VPM) and transpiration (T_VTM) at these sites. The seasonal dynamics of daily GPP_VPM and T_VTM track well with the seasonal dynamics of EC-based GPP (GPP_EC, R²: 0.76−0.93) and evapotranspiration (ET_EC, R²: 0.69−0.92). The inter-annual variation of annual GPP_VPM tracked well that of annual GPP_EC, with the linear regression slopes for GPP_EC versus GPP_VPM-EC ranging from 0.89 to 1.11. The simulation results of GPP_VPM and T_VTM using two different climate data sets (in situ climate data and European Center for Medium-Range Weather Forecasts Reanalysis v5 data set (ERA5)) were similar, suggesting that ERA5 data can be used for VPM/VTM simulations at large spatial scales. From 2000 to 2021, annual GPP_VPM and T_VTM had no significant inter-annual trends at one savanna and three grassland sites but increased significantly at one savanna site. The results demonstrate the potential of using VPM (v3.0) and VTM (v2.0) to predict the seasonal dynamics and inter-annual variation of GPP and T in savannas and grasslands.

Graphical abstract

Keywords

vegetation photosynthesis model / vegetation transpiration model / ERA5 / MODIS / carbon fluxes

RESEARCH ARTICLE

1 Introduction

Savannas and grasslands together account for nearly 50% of the earth’s land surface and are critical for terrestrial carbon cycles and ecosystem services (O’Mara, 2012; Kemp et al., 2013). They contribute the most to the interannual variation of terrestrial carbon sinks or sources, accounting for 51% of the variability (Biederman et al., 2017). Carbon cycles in grasslands and savannas are sensitive to air temperature and precipitation (Peng et al., 2013; Wagle et al., 2015a). The large variability in both timing and amount of precipitation in these areas leads to large inter-annual variability in gross primary production (GPP), making it difficult to estimate GPP in these areas. For example, one of the most widely used global GPP products (MOD17A2) underestimated GPP by 34% in 15 grassland sites with multi-year observations (> 5 years) around the world (Zhu et al., 2018). Another study found that some satellite-based models captured only 40% to 53% interannual variability of GPP at a woody savanna site (Scott et al., 2023). The under-estimation from these models can be attributed to underestimated values of the maximum light use efficiency parameter, uncertainty in estimating the fraction of absorbed photosynthetically active radiation (FPAR), and the uncertainty in estimating the effect of water stress on GPP (Sjöström et al., 2013; Zhu et al., 2018). Evapotranspiration (ET) comprises both soil and vegetation evaporation (E) and vegetation transpiration (T), playing a vital role in linking terrestrial water, carbon, and surface energy exchanges (Gentine et al., 2019). The projected changes in precipitation from future climate change scenarios may result in changes in plant water availability and GPP, making it more difficult to estimate spatial-temporal patterns of carbon and water fluxes (Swain et al., 2018; Konapala et al., 2020). Therefore, accurate observation and modeling of carbon and water cycles in these regions are needed.

The eddy covariance (EC) technique provides a continuous and reliable measurement of carbon dioxide (CO₂) and water vapor (H₂O) fluxes (Baldocchi, 2014). The GPP derived from EC sites (GPP_EC) has been widely used to support the development of process-based models and satellite-based remote sensing models (Jiang and Ryu, 2016; Zhang et al., 2017; Zheng et al., 2020). However, partitioning ET measured by EC instrumentation (ET_EC) into evaporation (E) and transpiration (T) is less common, limiting the development of process-based models to estimate T (Stoy et al., 2019). Several EC data-based techniques have been developed to partition ET, including the partition models that used the relationship between GPP and ET (Zhou et al., 2016; Scott and Biederman, 2017; Nelson et al., 2020), and the models that estimated canopy and soil conductance (Li et al., 2019). However, these models are dependent on the GPP_EC over the EC tower sites with varying sizes and shapes of footprints ranging from several hundred meters to kilometers.

Satellite remote sensing provides another option for estimating CO₂ and H₂O fluxes. Several light-use efficiency (LUE)-based production efficiency models (PEMs) have been developed to estimate the GPP of terrestrial ecosystems, driven by vegetation indices derived from optical images and climate data (Goetz et al., 1999; Turner et al., 2003; Zhao et al., 2005; Zhang et al., 2017; Zheng et al., 2020). As one of the LUE-based PEMs, the Vegetation Photosynthesis Model (VPM) uses the concept of light absorption by chlorophyll (APAR_chl) and considers the effect of air temperature (T_scalar) and water (W_scalar) stresses on LUE (Xiao et al., 2004a, 2024b). The biome-specific optimum air temperature parameter (TA_opt-biome) values were used in VPM v2.0 for savannas and grasslands (Zhang et al., 2017). A recent study introduced a site-specific TA_opt (TA_opt-site) parameter and assessed its effect on GPP estimation over 11 grassland sites (Chang et al., 2020). In addition, the vegetation transpiration model (VTM, v1.0) was developed to estimate T based on the intrinsic interactions between CO₂ (photosynthesis) and water (transpiration) processes at the leaf level and was evaluated with one-year data from 10 grassland sites in the US Southern Great Plains, as part of the 2002 International H₂O project (Alfieri et al., 2009). VTM estimates transpiration (T_VTM) as a function of GPP and transpiration ratio that is an inverse of water use efficiency (WUE). The GPP data used in VTM can come from either eddy flux tower sites (GPP_EC) or the VPM model (GPP_VPM).

In this study, we selected five savanna and grassland sites, each of which has 10+ (11−21 years) years of continuous EC measurements and ran VPM and VTM at these five sites with MODIS images and ERA5 climate data from 2000 to 2021. The objectives were to: i) understand the seasonal dynamics, inter-annual variation, and decadal change of climate, vegetation indices (Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Land Surface Water Index (LSWI)), and carbon and water fluxes at these sites; ii) assess the performance of VPM v3.0 and VTM v2.0 to predict seasonal dynamics and interannual variation of GPP and T; iii) investigate inter-annual variation and decadal change (trend) of atmospheric CO₂ concentration, climate, vegetation indices (NDVI, EVI, LSWI), GPP, and T from 2000 to 2021. The two research questions we asked are: i) can the VPM v3.0 and VTM v2.0 well simulate the seasonal dynamics and interannual variation of GPP and T in these grassland and savanna sites over the last two decades; ii) what are the inter-annual variation and trend of atmospheric CO₂ concentration, climate, vegetation indices (NDVI, EVI, LSWI), GPP, and T in these grassland and savanna sites over the last two decades. This site-level study with 10+ years of in situ data and VPM and VTM model simulations over the two decades will improve our understanding of carbon and water fluxes in savannas and grasslands and provide a case study that would help us to carry out regional and global simulations in the near future.

2 Materials and methods

2.1 Study sites

In this study, we selected two savanna sites (US-TON and US-SRM) and three grassland sites (US-VAR, US-SRG, and US-WKG). Among the grassland and savanna sites in the AmeriFlux network, these five sites have 10+ years of continuous measurements and the data that are available to the public range from 11 to 21 years (Tab.1). The geo-locations of the five sites are shown in Fig.1(a). Landscape features of the five sites are shown in Fig.1. A summary of vegetation composition and climate characteristics of the five sites are listed in Tab.1. Detailed descriptions of the five sites can be found on the AmeriFlux website.

2.2 Data sources

2.2.1 In situ data at the eddy flux tower sites

All CO₂, H₂O, energy, and meteorological data were downloaded from the AmeriFlux (US-VAR, US-TON, US-WKG, and US-SRG) and FLUXNET2015 (US-SRM) website. The data sets consist of data at five temporal resolutions: half-hourly, daily, weekly, monthly, and yearly. Data were processed using a consistent and uniform processing pipeline (Pastorello et al., 2020). Half-hourly flux and meteorological data, including GPP_EC, latent heat flux (LE), photosynthetically active radiation (PAR_EC), air temperature, atmospheric CO₂ concentration, precipitation (P_EC), and soil water content (SWC_EC) were used in this study. The LE was used to calculate ET_EC. We used a threshold of incoming photosynthetic photon flux density greater than 0.5 μmol Photon·m⁻²·s⁻¹ to define the daytime period. Daily daytime mean air temperature (TA_DT-EC) was computed to drive VPM. Daily data were further averaged over an 8-day interval to match the temporal resolution of the MODIS 8-day composite images.

2.2.2 MODIS product during 2000−2021

Time-series MOD09A1 data (500-m spatial resolution and 8-day temporal resolution) for each site were extracted from the Google Earth Engine (GEE) based on the geo-location information (latitude and longitude) of individual sites. The quality detection method of QA bands was used to identify and discard bad-quality data (i.e., cloud and cloud shadow observations). The resultant missing (gap) observations in the time series data were gap-filled by the linear interpolation gap-fill method over a 3- or 5- observation window (t−1, t, t + 1, or t−2, t−1, t, t + 1, t + 2) to generate continuous good-quality surface reflectance MOD09A1 data. For each MODIS 8-day surface reflectance observation, blue (459−479 nm), red (620−670 nm), near-infrared (NIR1, 841−876 nm), and shortwave infrared (SWIR1, 1628−1652 nm) bands were used to calculate three vegetation indices: NDVI (Tucker, 1979), EVI (Huete et al., 1997; Huete et al., 2002), and LSWI (Xiao et al., 2004a, b).

2.2.3 ECMWF reanalysis v5 climate data (ERA5)

The ERA5 data set provides hourly data for many atmospheric and surface variables with ~9 km spatial resolution. Hourly resolutions of the air temperature, downward shortwave radiation, and precipitation were used to compute daily daytime mean air temperature (TA_DT-ERA5), daily downward shortwave radiation flux, and precipitation (P_ERA5), respectively. In this study, we used a constant ratio of 0.45 to convert incident shortwave radiation into photosynthetically active radiation (PAR_ERA5) (Ryu et al., 2018) and 4.56 μmol·J⁻¹ as the energy-quanta conversion factor (Dye, 2004). Daily data were averaged over an 8-day interval, following the temporal resolution of MODIS 8-day composite images.

The comparison between TA_DT and PAR data from ERA5 and EC sites was shown in Fig. S1. Two California sites (US-VAR and US-TON) showed small differences (1%−4%) in TA_DT and PAR between ERA5 and EC data sets, with TA_DT RMSE ranging from 0.93°C to 1.29°C and R² > 0.99, and PAR RMSE ranging from 2.77 to 3.45 and R² > 0.99. At the US-WKG site, TA_DT-ERA5 and PAR_ERA5 were higher than TA_DT-EC and PAR_EC (3% and 8% difference between 1:1 line). At the US-SRG site, TA_DT-ERA5 was 6% lower than TA_DT-EC and PAR_ERA5 was 11% higher than PAR_EC. At the US-SRM site, TA_DT-ERA5 was 7% lower than TA_DT-EC and PAR_ERA5 was 14% higher than PAR_EC. Thus, these results indicate the reanalysis data can be used with high confidence at the California sites. However, the reanalysis data should be used with caution for the sites in Arizona.

2.3 Models description and simulation

2.3.1 The Vegetation Photosynthesis Model (VPM v3.0)

The VPM was developed to estimate daily GPP (g C·m⁻²·day⁻¹) as a product of light absorption by chlorophyll (APAR_chl) and light use efficiency (LUE, {\epsilon }_{\mathrm{g}}) (Xiao et al., 2004a, 2004; Xin et al., 2017; Chang et al., 2020). As shown in Eq. (1), The VPM estimates daily GPP as

(1)\text{G}\text{P}\text{P}={\text{APAR}}_{\text{chl}}\times {\epsilon }_{\text{g}},

where {\epsilon }_{\mathrm{g}} is the light use efficiency (μmol CO₂/μmol PPFD) and APAR_chl is the amount of PAR absorbed by chlorophyll. APAR_chl in Eq. (1) can be estimated as

(2){\mathrm{A}\mathrm{P}\mathrm{A}\mathrm{R}}_{\mathrm{c}\mathrm{h}\mathrm{l}}= {\text{FPAR}}_{\text{chl}} \times \mathrm{P}\mathrm{A}\mathrm{R},

where PAR is the photosynthetically active radiation (μmol photosynthetic photon flux density, PPFD) and FPAR_chl is the fraction of PAR absorbed by chlorophyll in the canopy.

The FPAR_chl is estimated as a linear function of EVI (Pan et al., 2024):

(3){\mathrm{F}\mathrm{P}\mathrm{A}\mathrm{R}}_{\mathrm{c}\mathrm{h}\mathrm{l}}=\text{E} \text{V}\text{I}-\mathrm{0.1.}

The coefficient of 0.1 is used to adjust for sparsely vegetated or barren land, which has been validated by solar-induced chlorophyll fluorescence data (Zhang et al., 2017).

The light use efficiency ({\epsilon }_{\mathrm{g}}) can be calculated as

(4){\epsilon }_{\mathrm{g}}= {\epsilon }_0\times T_{\mathrm{s}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{r}}\times W_{\mathrm{s}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{r}},

where {\epsilon }_0 is the maximum light use efficiency (g C/mol APAR), T_scalar and W_scalar are the downward regulation of temperature and water effects on maximum light use efficiency. The {\epsilon }_0 is a fixed value acquired from literature and varies in C3 and C4 plants, with 0.53 g C·mol⁻¹ APAR for C₃ plants, 0.79 g C·mol⁻¹ APAR for C₄ plants, respectively (Xiao, 2006). In this study, we made educated guesses on C3/C4 plant ratios in each site based on plant cover estimates through communication with principal investigators at these sites (Tab.1). For the two savanna sites (US-VAR and US-TON), there are no C4 grass plants, and the C3/C4 plant coverage ratio used in the model was 100%C3F and 0%C4F at the US-VAR site, 100%C3F and 0%C4F at the US-TON site. For the other three sites, the C3/C4 plant coverage ratios used in the model were 10%C3F and 90%C4F at the US-WKG site, 20%C3F and 80%C4F at the US-SRG site, 40%C3F and 60%C4F at the US-SRM site, respectively. So, in a pixel with different C₃ and C₄ plant fractions (C3F and C4F), {\epsilon }_0 can be calculated as

(5){\epsilon }_0=\mathrm{C}3\mathrm{F}\times {\epsilon }_{0-\mathrm{C}3}+\mathrm{C}4\mathrm{F}\times {\epsilon }_{0-\mathrm{C}4}.

The T_scalar is calculated by the following equation:

(6)T_{\mathrm{s}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{r}}=\frac{\left(\mathrm{T}\mathrm{A}-{\mathrm{T}\mathrm{A}}_{\mathrm{m}\mathrm{a}\mathrm{x}}\right)\times \left(\mathrm{T}\mathrm{A}-{\mathrm{T}\mathrm{A}}_{\mathrm{m}\mathrm{i}\mathrm{n}}\right)}{\left(\mathrm{T}\mathrm{A}-{\mathrm{T}\mathrm{A}}_{\mathrm{m}\mathrm{a}\mathrm{x}}\right)\times \left(\mathrm{T}\mathrm{A}-{\mathrm{T}\mathrm{A}}_{\mathrm{m}\mathrm{i}\mathrm{n}}\right)-{\left(\mathrm{T}\mathrm{A}-{\mathrm{T}\mathrm{A}}_{\mathrm{o}\mathrm{p}\mathrm{t}}\right)}^2},

where TA, TA_max, TA_min, and TA_opt are daily daytime mean air temperature, maximum, minimum, and optimum temperature for photosynthesis, respectively. When the air temperature falls below TA_min, TA_scalar is set to zero. TA_max and TA_min were set to 48°C and − 1°C for woody savanna, and 48°C and 0°C for grassland, respectively (Zhang et al., 2017).

In comparison to VPM v1.0 and VPM v.2.0 which used the TA_opt-biome parameter, VPM v3.0 used the TA_opt-site parameter. The methods to estimate TA_opt-site with GPP_EC and MODIS images at individual EC tower sites were reported in previous publications (Chang et al., 2020; Chang et al., 2021; Pan et al., 2024). The TA_opt-site value is calculated based on the response curves of GPP_EC versus TA_DT-ERA5 and EVI versus TA_DT-ERA5. Both GPP_EC and EVI increased with an increase in TA_DT-ERA5, reached a peak, and then declined as an increase in TA_DT-ERA5 (Fig.2). For the TA_opt-site value calculation, first, we found the maximum GPP_EC and EVI values for each site, respectively. Then, TA_opt-site was calculated as the mean TA_DT-ERA5 value during those GPP_EC or EVI observations equal to or higher than 95% maximum GPP_EC and EVI (Chang et al., 2020; Chang et al., 2021). Tab.2 lists the differences between the TA_opt-site calculated by GPP_EC and EVI. The TA_opt-site values varied among the five sites and showed some differences with TA_opt-biome. Considering EVI data are available globally and GPP_EC data are only available in EC tower sites, we selected TA_opt-site-EVI as the input for the VPM model simulations, which could help us to carry out global simulations of the VPM model in the future.

The W_scalar was calculated as

(7)W_{\text{scalar}} \text{=}\frac{\text{1+LSWI}}{\text{1+}{\text{LSWI}}_{\text{max}}},

where LSWI_max is the largest LSWI value during the growing season (April to October at US-WKG, US-SRG, and US-SRM sites, and the whole year at US-VAR and US-TON sites) within one year. Equation (7) was reported to have good performance in semi-humid and humid areas (Xiao et al., 2004a; Yan et al., 2009) but would underestimate the degree (severity) of water stress on photosynthesis under the very dry conditions in arid and semi-arid areas (He et al., 2014). As shown in Fig.3, long-term mean LSWI values were lower than −0.1 during the very dry period at these sites. Thus, we applied the modified W_scalar equation under very dry conditions, using the same procedure reported in our previous publications (He et al., 2014; Wagle et al., 2015b):

(8){\mathit{\text{W}}}_{\text{scalar}}=0.5 +\mathrm{L}\mathrm{S}\mathrm{W}\mathrm{I}.

When LSWI value was lower than −0.1, which represents a dry condition, we applied Eq. (8) in the VPM simulation. This modification improves the accuracy of water stress estimation by better reflecting the severity of water limitation on photosynthesis. For example, on June 17, 2008, the US-WKG site had an LSWI value of −0.155, and W_scalar calculated by Eq. (7) was 0.83 but W_scalar calculated by Eq. (8) was 0.35, which more accurately accounted for the degree of water stress on GPP.

2.3.2 The Vegetation Transpiration Model (VTM v2.0)

The VTM was developed to estimate daily transpiration (mm H₂O /day), based on the intrinsic coupling between photosynthesis and transpiration processes at the leaf scale, i.e., water use efficiency (WUE) (Eq. (9)), and it combines the inverse of WUE and GPP to estimate transpiration (T) (Alfieri et al. 2009; Pan et al. 2024). The equation can be described as follows:

(9)\mathrm{W}\mathrm{U}\mathrm{E}=\mathrm{G} \mathrm{P}\mathrm{P}/\mathrm{T}\mathrm{r}\mathrm{a}\mathrm{n} \mathrm{s}\mathrm{p}\mathrm{i}\mathrm{r}\mathrm{a} \mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n},

(10)\mathrm{T}=\frac{1}{\mathrm{W}\mathrm{U}\mathrm{E}}\times \mathrm{G} \mathrm{P}{\mathrm{P}}_{\mathrm{V}\mathrm{P}\mathrm{M}},

(11)\frac{1}{\mathrm{W}\mathrm{U}\mathrm{E}}=\mathrm{C}3\mathrm{F}\times \frac{1}{{\mathrm{W}\mathrm{U}\mathrm{E}}_{\mathrm{C}3}}+\mathrm{C}4\mathrm{F} \times \frac{1}{{\mathrm{W}\mathrm{U}\mathrm{E}}_{\mathrm{C}4}}.

At the leaf level, 1/WUE = 0.27 mm H₂O/g C for the C3 plant and 0.135 mm H₂O/g C for the C4 plant, respectively (Nobel, 2020). More details about the 1/WUE can be found in supplementary materials for Method 1. The GPP data from the simulations of VPM (GPP_VPM) in an 8-day temporal interval was used to estimate T (mm·day⁻¹).

2.3.3 Simulation of VPM and VTM models

In this study, two meteorological data sets: in situ meteorological data from the EC tower and reanalysis meteorological data from the ERA5 data set were used to drive VPM and VTM simulations. These two meteorological data sets were averaged over an 8-day interval, matching the temporal resolution of MODIS composite images using the equations in Sections 2.3.1 and 2.3.2, and were used to run VPM and VTM. For the in situ meteorological data set, the model simulation was dependent on the data availability. For the ERA5 reanalysis data set, we simulated VPM and VTM from 2000 to 2021 (22 years). We defined active growing season as GPP > 1 g C·m⁻²·day⁻¹ for model evaluation (Wagle et al., 2014).

2.4 Statistical data analysis

Three statistical metrics: regression coefficient (α), the coefficient of determination (R²), and root mean square error (RMSE) were calculated to assess the accuracies and uncertainties of the model simulations. The R² value denotes the amount of variation in the measurements that could be predicted by the estimates, and the RMSE expresses the differences between the estimates and the measurements.

Mean annual environmental factors data, vegetation indices, GPP, and T were calculated as the averaged value of each variable over their temporal coverage. The interannual trends of vegetation indices, GPP, T, and climate in this study were analyzed by using the Mann-Kendall test and Sen’s method. The Mann-Kendall test was used to evaluate the reliability of the results (Mann, 1945; Kendall, 1948), while Sen’s slope method was used to quantify the trends (Sen, 1968).

3 Results

3.1 Seasonal dynamics of environmental factors, vegetation indices, and carbon and water fluxes

Fig.3 shows the averaged seasonal dynamics of PAR_EC, TA_DT-EC, P_EC, SWC_EC, vegetation indices, GPP_EC, and ET_EC at the five sites. With the Mediterranean climate in California, USA, the grassland site (US-VAR) and woody savanna site (US-TON) had similar seasonal dynamics of PAR_EC and TA_DT-EC. They started to increase in spring and reached their peaks in summer. The seasonal dynamics of precipitation were characterized by a dry season from late spring to early fall and a wet season from late fall to spring. SWC_EC at the woody savanna site was higher than the grassland site, with low SWC_EC in the summer (< 10%) and high SWC_EC in winter and spring (up to 45%). Vegetation indices (VIs, NDVI, EVI, and LSWI) had similar seasonal dynamics, characterized by a peak in spring. Both GPP_EC and ET_EC had similar seasonal dynamics, characterized by a peak in spring. The seasonal dynamics of GPP_EC and ET_EC agreed well with the seasonal dynamics of VIs. The period of GPP_EC ≥ 1 g C·m⁻²·day⁻¹ (early March – July at the woody savanna site, February–May at the grassland site) corresponded well with the period with increased vegetation indices (i.e., an indication of growing seasons).

Under the monsoon climate in southern Arizona, USA, the grassland sites (US-WKG (Fig.3(c)), US-SRG (Fig.3(d)) and woody savanna site (US-SRM, Fig.3(e)) had similar seasonal dynamics of environmental factors. PAR_EC and TA_DT-EC reached their peak values in June and decreased gradually in summer. The seasonal dynamics of precipitation were characterized by dry spring and fall seasons, a moderately wet winter (November–February), and a wetter monsoon (July–October). SWC_EC was relatively high from July to February and low from March to June. Three VIs (NDVI, EVI, and LSWI) had similar seasonal dynamics, characterized by a large peak period in August−September. GPP_EC and ET_EC had similar seasonal dynamics, with moderate values in the spring growing season and larger values in the summer growing season.

3.2 Comparisons of GPP_EC and GPP_VPM from the VPM simulation

A comparison between GPP_EC and GPP_VPM is shown in Fig.4. In general, both GPP_VPM-EC and GPP_VPM-ERA5 tracked the seasonal dynamics of GPP_EC reasonably well at the five sites. Specifically, for the two sites under the Mediterranean climate (Fig.4(a1) and Fig.4(b1)), both GPP_VPM-EC and GPP_VPM-ERA5 precisely tracked the dynamics of GPP_EC from February to June but showed moderate differences from July to October. For the three monsoon sites (Fig.4(c1) and Fig.4(e1)), both GPP_VPM-EC and GPP_VPM-ERA5 agreed reasonably well with GPP_EC during the summer growing season (July to October), GPP_VPM-EC showed slight differences in May–June with GPP_EC, while the GPP_VPM-ERA5 showed noticeable difference. Linear regression showed a high correlation of GPP_VPM-EC and GPP_VPM-ERA5 with GPP_EC, with R² values ranging from 0.76 to 0.93 and RMSE values ranging from 0.8 to 1.7 g C·m⁻²·day⁻¹. The linear regression slope values between GPP_VPM-EC and GPP_EC differ among the sites: −10% for US-TON, +2% for US-VAR, +5% for US-SRG, and +9% for both US-WKG and US-SRM. The VPM driven by ERA5 data showed slightly larger differences: −14% for US-TON, +5% for US-VAR, +15% for US-WKG, and +25% for both US-SRG and US-SRM.

3.3 Temporal consistency of evapotranspiration and transpiration at temporal scales of 8-day period

The seasonal dynamics of T_VTM-EC and T_VTM-ERA5 matched reasonably well with the seasonal dynamics of ET_EC (Fig.5). However, there were large differences in their magnitudes. Linear regression showed a strong correlation of T_VTM-EC and T_VTM-ERA5 with ET_EC, with R² values ranging from 0.69 to 0.92, and RMSE ranging from 0.21 to 0.37 mm·day⁻¹. The T_VTM-EC / ET_EC ratio varies across the sites, ranging from 0.71 for US-VAR and 0.61 for US-TON under Mediterranean climate, to 0.29 for US-SRM, 0.27 for US-SRG and 0.23 for US-WKG, which is in the monsoon climate area. The T_VTM-ERA5 / ET_EC ratios showed similar differences with T_VTM-EC / ET_EC, ranging from 0.27 to 0.73 at the five sites.

3.4 Comparison between annual GPP_EC vs. GPP_VPM and annual ET_EC vs. T_VTM

A comparison between the annual GPP_EC and GPP_VPM is shown in Fig.6. The difference between the regression slopes for annual GPP_EC versus GPP_VPM-EC was within 11% of the 1-1 line at the five sites. Results were similar to GPP_VPM-ERA5. At the US-SRG and US-SRM sites, the differences between annual GPP_EC and GPP_VPM-ERA5 (26% and 25% difference at the US-SRG and US-SRM sites, respectively) were noticeably higher than those between GPP_EC and GPP_VPM-EC (6% and 11% difference at the US-SRG and US-SRM site, respectively).

At the annual scale, the ratio of annual T_VTM-EC and annual ET_EC varied among the five sites (0.68 at the US-VAR, 0.65 at the US-TON, 0.30 at the US-SRM, 0.26 at the US-SRG, and 0.22 at the US-WKG). Results were similar for ET_EC and T_VTM-ERA5. The ratio of annual T_VTM-ERA5 and annual ET_EC was 0.70 at the US-VAR site, 0.63 at the US-TON site, 0.34 at the US-SRM site, 0.31 at the US-SRG site, and 0.23 at the US-WKG site.

3.5 Interannual variability of atmospheric CO₂, environmental factors, VIs, gross primary production, and transpiration during 2000–2021

The interannual variability of atmospheric CO₂ concentration, environmental factors (PAR_ERA5, TA_DT-ERA5, and P_ERA5), vegetation indices (NDVI, EVI, and LSWI), GPP_VPM, and T_VTM during 2000−2021 at the five sites are shown in Fig.7. Atmospheric CO₂ con-centration showed a significant increasing trend. The Sen’s slope ranged from 2.16 ppm·yr⁻¹ to 2.48 ppm·yr⁻¹ among the five sites. Daytime air temperature showed a significant increasing trend (Sen’s slope: 0.05). PAR_ERA5 and P_ERA5 showed no significant trend. EVI at the US-SRM site showed significant increasing trends.

The GPP_VPM-ERA5 and T_VTM-ERA5 exhibited no significant trend during 2000−2021 at one savanna and three grassland sites, but they increased substantially at the US-SRM savanna site. At the US-SRM, GPP_VPM-ERA5, and T_VTM-ERA5 showed an increasing trend of 10.59 g C·m⁻²·yr⁻¹ and 2.0 mm H₂O·yr⁻¹, respectively. The Mediterranean climate sites had the highest GPP and T followed by the US-SRG grassland site and the US-SRM savanna site. The US-WKG grassland had the smallest GPP and T.

4 Discussion

4.1 Improvement of the VPM and GPP_VPM estimation

The VPM v2.0 was applied to generate a global GPP data set over the years 2000−2016, which was released to the public and used by the research community (Zhang et al., 2017). Both VPM v2.0 and the MOD17 GPP data products (Running and Zhao, 2019) use biome-specific TA_opt-biome parameters for photosynthesis. Previous studies revealed the limitation of biome-specific TA_opt-biome in the LUE-based GPP models and showed the need and potential of using site-specific TA_opt-site to improve GPP estimates in VPM (Chang et al., 2020, 2021). In this study, we used the same method from the previous publications (Chang et al., 2020, 2021) to estimate site-specific TA_opt-site. In comparison to the previous publications that used a few years of data, this study used longer time series data (11 to 21 years) and confirmed that TA_opt-site values calculated from the relationship between GPP_EC and TA_DT-ERA5 were consistent with TA_opt-site values from the relationship between EVI and TA_DT-ERA5. The VPM simulation results demonstrated the potential of using TA_opt-site to replace TA_opt-biome for improving GPP estimation by the VPM. The use of EVI in estimating TA_opt-site has great potential for improving global VPM modeling as time series satellite-derived EVI data are available globally.

The discrepancies between GPP_VPM and GPP_EC in some periods can be attributed to several sources of errors and uncertainties. First, climate data (e.g., PAR) affect model simulations (He et al., 2014). PAR is one of the most sensitive input variables of the VPM model (He et al., 2014). In this study, PAR_ERA5 exhibited larger differences with PAR_EC at the US-SRG (11% difference) and US-SRM (14% difference) sites than the other three sites. These differences propagate to GPP_VPM driven by ERA5 climate data, resulting in much larger discrepancies with GPP_EC than did GPP_VPM driven by EC climate data at these two sites. It highlights the need for careful quality control of PAR data and suggests that an empirical or semi-empirical correction may be needed when using the ERA5 as model input data. Second, the time series of vegetation index data often have missing (bad quality) data due to clouds, aerosol, and other noise, and how to gap-fill the missing data for a complete time series of remote sensing data is still under debate (Shen et al., 2015). The time lag of EVI when drought occurs also introduces biases to VPM simulation (Dong et al., 2015). For example, at the US-SRG site, there was moderate disagreement between GPP_EC and GPP_VPM during the spring dry period (April – June). In this site, there is a prominent dry period from April to June. When drought occurs, even at similar light conditions from morning to afternoon hours, photosynthesis could decrease considerably due to stomatal closure (Wagle and Kakani, 2014). However, significant alterations in pigment concentration and canopy structure, such as leaf shedding and reductions in leaf area index, which influence NDVI and EVI, typically require several days to several weeks to become noticeable. Due to the time lag of EVI, the simulated GPP peaked slightly later than the peak GPP_EC during the spring dry period.

Third, various proportions of trees and grasses occurred in three of the five sites in this study, and various proportions of C3 and C4 grasses also occurred at the study sites. C3 and C4 plants have different photosynthetic pathways, which leads to different LUE values. In the VPM, the proportions of C3 and C4 plants in a grassland or savanna affect the maximum LUE parameter (LUE₀ = C3F × LUE_0-C3 + C4F × LUE_0-C4). However, it is difficult to measure or estimate accurately species composition, which changes over time. Two approaches have been used to estimate C3/C4 ratios in grasslands and savannas: 1) make an educated guess of C3/C4 mixing ratios based on site-specific knowledge or 2) use known C3/C4 ratios from other data sources (in situ observations or model-based estimation). As the US-WKG, US-SRG, and US-SRM sites do not have sufficient information on varying mixing ratios of C3 and C4 grasses, we estimated C3/C4 plant ratios based on available plant cover estimates, which may contribute to the moderate discrepancies between GPP_EC and GPP_VPM. Additionally, fixed proportions of C3 and C4 species do not account for seasonal changes in species composition. For example, C4 plants might dominate in hotter, drier conditions typically found in summer, while C3 plants might be more prevalent in cooler, wetter conditions. Fixed C3/C4 ratios fail to reflect these shifts, potentially leading to over- or underestimation of GPP_VPM during specific periods. To date, the capacity to model and predict spatial distribution and temporal (seasonal and inter-annual) abundance of C4 plants across the scales from local to global is still poor (Still et al., 2003; Winslow et al., 2003; Luo et al., 2024). The result of this study further highlights the need to model and generate accurate local regional and global C4 plant distribution maps in the future. The fourth uncertainty could be attributed to the estimation error of GPP_EC. For example, during the summer, EVI at the US-VAR site was close to 0.2, which indicated the presence of green vegetation. Many studies have shown EVI > 0.1 can be an indicator of green vegetation (Xiao et al., 2009; Wang et al., 2020). However, GPP_EC at the US-VAR site during this period was almost zero. This causes a moderate disagreement between GPP_EC and GPP_VPM at this site during summer.

4.2 The range of T/ET ratios in grasslands and savanna

Due to the lack of in situ T measurement for model validation, we compared the T or T/ET ratio estimates across the five sites with published studies using several methods. A previous study reported T of 281 mm·yr⁻¹ and T/ET of 0.6 at the US-TON site, based on EC tower measurements of above and below the tree canopy for ET partition (Ma et al., 2020). Based on modeling results, Ma et al. (2020) reported that at the US-TON site, T/ET ratio varies from 0.74 (with the assumption of t = 0 if GPP = 0) to 0.39 (with the assumption of the linear relationship between GPP and ET). Our results were within these ranges, with multi-year averages of T of 243 mm·yr⁻¹ and T/ET of 0.69, respectively. Similarly, at the US-VAR grassland site, the T/ET ratio ranged from 0.66 (with the assumption of t = 0 if GPP = 0) to 0.47 (with the assumption of the linear relationship between GPP and ET) (Ma et al., 2020). Our analysis yielded a T/ET_EC ratio of 0.70 at the US-VAR Mediterranean grassland site. By using the x-axis intercepts for ET vs GPP as an estimate of evaporation (E), and by subtracting E from ET measured by the EC tower, the annual mean T/ET ratio was 0.62 at the US-SRM woody savanna site, 0.58 at the US-SRG grassland site, and 0.55 at the US-WKG grassland site, respectively (Scott et al., 2015). Scott and Biederman (2017) applied this method to monthly growing-season ET partitioning and predicted that T/ET ratio ranged from 0.45 to 0.69 (mean of 0.62, July–October) at the US-SRM site, from 0.51 to 0.62 (mean of 0.55, July–October) at the US-SRG site, and from 0.26 to 0.58 (mean of 0.46, July–October) at the US-WKG site, respectively. Our T/ET_EC ratio estimates at the US-SRM, US-SRG, and US-WKG sites (0.23 to 0.29) were substantially lower than these results.

In this study, we conducted preliminary work on using VTM to estimate T. This simple method, which assumes a constant WUE, performed reasonably at the US-VAR and US-TON sites but was less effective at the US-WKG, US-SRG, and US-SRM sites. The WUE may be affected by different factors such as climate, soil moisture, plant functional types, etc. (Cooley et al. 2022). These factors might restrict the model’s effectiveness and precision when applied to various ecosystems and climatic areas. The limitation suggests that it is a challenging task to accurately estimate T using such a simple approach. It is clear that the magnitude of T estimates needs to be further evaluated, and water use efficiency or transpiration ratio parameters need to be carefully evaluated, too, especially in sparse vegetation ecosystems under semiarid and arid environments.

4.3 Interannual variability of GPP and T

In grassland and savanna ecosystems, GPP is impacted by environmental factors such as atmospheric CO₂ con-centration, air temperature, precipitation, and solar radiation. Increasing atmospheric CO₂ concentration could increase GPP but reduce the T of grassland and savanna (CO₂ fertilization effect). A previous study found that no interannual trend in ET or oak canopy T was observed using the EC method at the US-TON woody savanna and the US-VAR grassland site from 2001 through 2019 despite an increasing trend of atmospheric CO₂ concentration and an increasing trend of air temperature and large variability of annual precipitation (ranging from 133 to 890 mm·yr⁻¹) (Baldocchi et al., 2021). We also found no interannual trend of GPP derived from the EC tower from 2001 to 2021 at these two sites (Fig. S3). The US-SRG grassland sites also showed no interannual trend of GPP during 14 years of in situ observations. Consistent with the observation results, our model simulation showed no significant trends of GPP and T over the last 22 years (2000−2021) at the US-TON woody savanna, US-VAR grassland, and US-SRG grassland sites. At the US-WKG grassland site, GPP_EC showed a significant increasing trend from 2004 to 2021, but VPM simulation results showed no significant trend from 2000 to 2021. Our VPM simulation results showed a significant increase in GPP from 2000 to 2021 at the US-SRM savanna site, which aligns with the trend of GPP_EC rising from 2004 to 2022 (Scott et al., 2023). This trend can be attributed to the increasing overstory woody plant cover and understory grass cover (Scott et al., 2023). The observed trends and interannual variability in GPP and T highlight the importance of continuous observation and modeling, which can guide effective land management strategies, ensuring the sustainability of these vital ecosystems.

5 Conclusions

In this study, we employed VPM and VTM to estimate GPP and T, respectively. We explored the seasonal dynamics, interannual variability, and decadal trends of atmospheric CO₂ concentration, environmental factors, vegetation indices, GPP, and T across five savanna and grassland sites. VPM demonstrated good performance in capturing the seasonal dynamics and interannual variability of GPP_EC. Using EVI as an alternative to EC-derived GPP to estimate the TA_opt for photosynthesis has improved the VPM simulation. Employing ERA5 reanalysis data as inputs for VPM simulations aligned well with site EC-observed results, supporting the potential of VPM simulations at regional and global scales. The VTM effectively tracked the seasonal dynamics of ET in savanna and grassland, though additional parameterization is needed for this simple leaf-level WUE approach to accurately estimate T in different climate zones. This study combines climate and satellite data to simulate long-term carbon and water fluxes, the results align well with eddy covariance observation, which plays a vital role in understanding the carbon and water cycles in response to atmospheric CO₂ concentration increase and climate change. By scaling up this methodology, the VPM model could become a valuable tool for estimating GPP in grasslands and savannas on regional to global scales, thereby supporting grassland management and conservation efforts under climate change scenarios.

References

[1] Alfieri J G, Xiao X, Niyogi D, Pielke R A Sr, Chen F, LeMone M A (2009). Satellite-based modeling of transpiration from the grasslands in the Southern Great Plains, USA.Global Planet Change, 67(1−2): 78–86

[2] Baldocchi D (2014). Measuring fluxes of trace gases and energy between ecosystems and the atmosphere – The state and future of the eddy covariance method.Glob Change Biol, 20(12): 3600–3609

[3] Baldocchi D, Ma S, Verfaillie J (2021). On the inter- and intra-annual variability of ecosystem evapotranspiration and water use efficiency of an oak savanna and annual grassland subjected to booms and busts in rainfall.Glob Change Biol, 27(2): 359–375

[4] Biederman J A, Scott R L, Bell T W, Bowling D R, Dore S, Garatuza-Payan J, Kolb T E, Krishnan P, Krofcheck D J, Litvak M E, Maurer G E, Meyers T P, Oechel W C, Papuga S A, Ponce-Campos G E, Rodriguez J C, Smith W K, Vargas R, Watts C J, Yepez E A, Goulden M L (2017). CO₂ exchange and evapotranspiration across dryland ecosystems of southwestern North America.Glob Change Biol, 23(10): 4204–4221

[5] Chang Q, Xiao X, Doughty R, Wu X, Jiao W, Qin Y (2021). Assessing variability of optimum air temperature for photosynthesis across site-years, sites and biomes and their effects on photosynthesis estimation.Agric For Meteorol, 298–299: 108277

[6] Chang Q, Xiao X, Wu X, Doughty R, Jiao W, Bajgain R, Qin Y, Wang J (2020). Estimating site-specific optimum air temperature and assessing its effect on the photosynthesis of grasslands in mid- to high-latitudes.Environ Res Lett, 15(3): 034064

[7] Cooley S S, Fisher J B, Goldsmith G R (2022). Convergence in water use efficiency within plant functional types across contrasting climates.Nat Plants, 8(4): 341–345

[8] Dong J, Xiao X, Wagle P, Zhang G, Zhou Y, Jin C, Torn M S, Meyers T P, Suyker A E, Wang J, Yan H, Biradar C, Moore B III (2015). Comparison of four EVI-based models for estimating gross primary production of maize and soybean croplands and tallgrass prairie under severe drought.Remote Sens Environ, 162: 154–168

[9] Dye D G (2004). Spectral composition and quanta-to-energy ratio of diffuse photosynthetically active radiation under diverse cloud conditions.J Geophys Res, 109(D10): 2003JD004251

[10] Gentine P, Green J K, Guérin M, Humphrey V, Seneviratne S I, Zhang Y, Zhou S (2019). Coupling between the terrestrial carbon and water cycles—a review.Environ Res Lett, 14(8): 083003

[11] Goetz S J, Prince S D, Goward S N, Thawley M M, Small J (1999). Satellite remote sensing of primary production: an improved production efficiency modeling approach.Ecol Modell, 122(3): 239–255

[12] He H, Liu M, Xiao X, Ren X, Zhang L, Sun X, Yang Y, Li Y, Zhao L, Shi P, Du M, Ma Y, Ma M, Zhang Y, Yu G (2014). Large-scale estimation and uncertainty analysis of gross primary production in Tibetan alpine grasslands.J Geophys Res Biogeosci, 119(3): 466–486

[13] Huete A R, Liu H Q, Batchily K, van Leeuwen W (1997). A comparison of vegetation indices over a global set of TM images for EOS-MODIS.Remote Sens Environ, 59(3): 440–451

[14] Huete A, Didan K, Miura T, Rodriguez E P, Gao X, Ferreira L G (2002). Overview of the radiometric and biophysical performance of the MODIS vegetation indices.Remote Sens Environ, 83(1−2): 195–213

[15] Jiang C, Ryu Y (2016). Multi-scale evaluation of global gross primary productivity and evapotranspiration products derived from Breathing Earth System Simulator (BESS).Remote Sens Environ, 186: 528–547

[16] Kemp D R, Guodong H, Xiangyang H, Michalk D L, Fujiang H, Jianping W, Yingjun Z (2013). Innovative grassland management systems for environmental and livelihood benefits.Proc Natl Acad Sci USA, 110(21): 8369–8374

[17] Kendall M G (1948). Rank Correlation Methods.Griffin, Oxford: England

[18] Konapala G, Mishra A K, Wada Y, Mann M E (2020). Climate change will affect global water availability through compounding changes in seasonal precipitation and evaporation.Nat Commun, 11(1): 3044

[19] Li X, Gentine P, Lin C, Zhou S, Sun Z, Zheng Y, Liu J, Zheng C (2019). A simple and objective method to partition evapotranspiration into transpiration and evaporation at eddy-covariance sites.Agric For Meteorol, 265: 171–182

[20] Luo X, Zhou H, Satriawan T W, Tian J, Zhao R, Keenan T F, Griffith D M, Sitch S, Smith N G, Still C J (2024). Mapping the global distribution of C4 vegetation using observations and optimality theory.Nat Commun, 15(1): 1219

[21] Ma S, Baldocchi D, Wolf S, Verfaillie J (2016). Slow ecosystem responses conditionally regulate annual carbon balance over 15 years in Californian oak-grass savanna.Agric For Meteorol, 228–229: 252–264

[22] Ma S, Eichelmann E, Wolf S, Rey-Sanchez C, Baldocchi D D (2020). Transpiration and evaporation in a Californian oak-grass savanna: field measurements and partitioning model results.Agric For Meteorol, 295: 108204

[23] Mann H B (1945). Nonparametric tests against trend.Econometrica, 13(3): 245–259

[24] Nelson J A, Pérez-Priego O, Zhou S, Poyatos R, Zhang Y, Blanken P D, Gimeno T E, Wohlfahrt G, Desai A R, Gioli B, Limousin J M, Bonal D, Paul-Limoges E, Scott R L, Varlagin A, Fuchs K, Montagnani L, Wolf S, Delpierre N, Berveiller D, Gharun M, Belelli Marchesini L, Gianelle D, Šigut L, Mammarella I, Siebicke L, Andrew Black T, Knohl A, Hörtnagl L, Magliulo V, Besnard S, Weber U, Carvalhais N, Migliavacca M, Reichstein M, Jung M (2020). Ecosystem transpiration and evaporation: insights from three water flux partitioning methods across FLUXNET sites.Glob Change Biol, 26(12): 6916–6930

[25] NobelP S (2020) Physicochemical and Environmental Plant Physiology (5th Ed). New York: Academic Press

[26] O’Mara F P (2012). The role of grasslands in food security and climate change.Ann Bot (Lond), 110(6): 1263–1270

[27] Pan L, Xiao X, Pan B, Meng C, Staebler R M, Zhang C, Qin Y (2024). Interannual variations and trends of gross primary production and transpiration of four mature deciduous broadleaf forest sites during 2000–2020.Remote Sens Environ, 304: 114042

[28] Pastorello G, Trotta C, Canfora E, Chu H, Christianson D, Cheah Y W, Poindexter C, Chen J, Elbashandy A, Humphrey M, Isaac P, Polidori D, Reichstein M, Ribeca A, van Ingen C, Vuichard N, Zhang L, Amiro B, Ammann C, Arain M A, Ardö J, Arkebauer T, Arndt S K, Arriga N, Aubinet M, Aurela M, Baldocchi D, Barr A, Beamesderfer E, Marchesini L B, Bergeron O, Beringer J, Bernhofer C, Berveiller D, Billesbach D, Black T A, Blanken P D, Bohrer G, Boike J, Bolstad P V, Bonal D, Bonnefond J M, Bowling D R, Bracho R, Brodeur J, Brümmer C, Buchmann N, Burban B, Burns S P, Buysse P, Cale P, Cavagna M, Cellier P, Chen S, Chini I, Christensen T R, Cleverly J, Collalti A, Consalvo C, Cook B D, Cook D, Coursolle C, Cremonese E, Curtis P S, D’Andrea E, da Rocha H, Dai X, Davis K J, Cinti B D, Grandcourt A, Ligne A D, De Oliveira R C, Delpierre N, Desai A R, Di Bella C M, Tommasi P, Dolman H, Domingo F, Dong G, Dore S, Duce P, Dufrêne E, Dunn A, Dušek J, Eamus D, Eichelmann U, ElKhidir H A M, Eugster W, Ewenz C M, Ewers B, Famulari D, Fares S, Feigenwinter I, Feitz A, Fensholt R, Filippa G, Fischer M, Frank J, Galvagno M, Gharun M, Gianelle D, Gielen B, Gioli B, Gitelson A, Goded I, Goeckede M, Goldstein A H, Gough C M, Goulden M L, Graf A, Griebel A, Gruening C, Grünwald T, Hammerle A, Han S, Han X, Hansen B U, Hanson C, Hatakka J, He Y, Hehn M, Heinesch B, Hinko-Najera N, Hörtnagl L, Hutley L, Ibrom A, Ikawa H, Jackowicz-Korczynski M, Janouš D, Jans W, Jassal R, Jiang S, Kato T, Khomik M, Klatt J, Knohl A, Knox S, Kobayashi H, Koerber G, Kolle O, Kosugi Y, Kotani A, Kowalski A, Kruijt B, Kurbatova J, Kutsch W L, Kwon H, Launiainen S, Laurila T, Law B, Leuning R, Li Y, Liddell M, Limousin J M, Lion M, Liska A J, Lohila A, López-Ballesteros A, López-Blanco E, Loubet B, Loustau D, Lucas-Moffat A, Lüers J, Ma S, Macfarlane C, Magliulo V, Maier R, Mammarella I, Manca G, Marcolla B, Margolis H A, Marras S, Massman W, Mastepanov M, Matamala R, Matthes J H, Mazzenga F, McCaughey H, McHugh I, McMillan A M S, Merbold L, Meyer W, Meyers T, Miller S D, Minerbi S, Moderow U, Monson R K, Montagnani L, Moore C E, Moors E, Moreaux V, Moureaux C, Munger J W, Nakai T, Neirynck J, Nesic Z, Nicolini G, Noormets A, Northwood M, Nosetto M, Nouvellon Y, Novick K, Oechel W, Olesen J E, Ourcival J M, Papuga S A, Parmentier F J, Paul-Limoges E, Pavelka M, Peichl M, Pendall E, Phillips R P, Pilegaard K, Pirk N, Posse G, Powell T, Prasse H, Prober S M, Rambal S, Rannik Ü Raz-Yaseef N, Rebmann C, Reed D, Dios V R, Restrepo-Coupe N, Reverter B R, Roland M, Sabbatini S, Sachs T, Saleska S R, Sánchez-Cañete E P, Sanchez-Mejia Z M, Schmid H P, Schmidt M, Schneider K, Schrader F, Schroder I, Scott R L, Sedlák P, Serrano-Ortíz P, Shao C, Shi P, Shironya I, Siebicke L, Šigut L, Silberstein R, Sirca C, Spano D, Steinbrecher R, Stevens R M, Sturtevant C, Suyker A, Tagesson T, Takanashi S, Tang Y, Tapper N, Thom J, Tomassucci M, Tuovinen J P, Urbanski S, Valentini R, van der Molen M, van Gorsel E, van Huissteden K, Varlagin A, Verfaillie J, Vesala T, Vincke C, Vitale D, Vygodskaya N, Walker J P, Walter-Shea E, Wang H, Weber R, Westermann S, Wille C, Wofsy S, Wohlfahrt G, Wolf S, Woodgate W, Li Y, Zampedri R, Zhang J, Zhou G, Zona D, Agarwal D, Biraud S, Torn M, Papale D (2020). ., 7(1): 225

[29] Peng S, Piao S, Shen Z, Ciais P, Sun Z, Chen S, Bacour C, Peylin P, Chen A (2013). Precipitation amount, seasonality and frequency regulate carbon cycling of a semi-arid grassland ecosystem in Inner Mongolia, China: a modeling analysis.Agric For Meteorol, 178–179: 46–55

[30] Running S W, Zhao M (2019). Daily GPP and Annual NPP (MOD17A2H/A3H) and Year-end Gap- Filled (MOD17A2HGF/A3HGF) Products NASA Earth Observing System MODIS Land Algorithm (For Collection 6). In: The Numerical Terradynamic Simulation Group, Missoula, MT, USA

[31] Ryu Y, Jiang C, Kobayashi H, Detto M (2018). MODIS-derived global land products of shortwave radiation and diffuse and total photosynthetically active radiation at 5km resolution from 2000.Remote Sens Environ, 204: 812–825

[32] Scott R L, Biederman J A (2017). Partitioning evapotranspiration using long-term carbon dioxide and water vapor fluxes.Geophys Res Lett, 44(13): 6833–6840

[33] Scott R L, Biederman J A, Hamerlynck E P, Barron-Gafford G A (2015). The carbon balance pivot point of southwestern U. S. semiarid ecosystems: insights from the 21st century drought.J Geophys Res Biogeosci, 120(12): 2612–2624

[34] Scott R L, Johnston M R, Knowles J F, MacBean N, Mahmud K, Roby M C, Dannenberg M P (2023). Interannual variability of spring and summer monsoon growing season carbon exchange at a semiarid savanna over nearly two decades.Agric For Meteorol, 339: 109584

[35] Sen P K (1968). Estimates of the regression coefficient based on Kendall’s Tau.J Am Stat Assoc, 63(324): 1379–1389

[36] Shen H, Li X, Cheng Q, Zeng C, Yang G, Li H, Zhang L (2015). Missing information reconstruction of remote sensing data: a technical review.IEEE Geosci Remote Sens Mag, 3(3): 61–85

[37] Sjöström M, Zhao M, Archibald S, Arneth A, Cappelaere B, Falk U, de Grandcourt A, Hanan N, Kergoat L, Kutsch W, Merbold L, Mougin E, Nickless A, Nouvellon Y, Scholes R J, Veenendaal E M, Ardö J (2013). Evaluation of MODIS gross primary productivity for Africa using eddy covariance data.Remote Sens Environ, 131: 275–286

[38] Still C J, Berry J A, Collatz G J, DeFries R S (2003). Global distribution of C3 and C4 vegetation: carbon cycle implications.Global Biogeochemical Cycles 17, 6–1–6–14

[39] Stoy P C, El-Madany T S, Fisher J B, Gentine P, Gerken T, Good S P, Klosterhalfen A, Liu S, Miralles D G, Perez-Priego O, Rigden A J, Skaggs T H, Wohlfahrt G, Anderson R G, Coenders-Gerrits A M J, Jung M, Maes W H, Mammarella I, Mauder M, Migliavacca M, Nelson J A, Poyatos R, Reichstein M, Scott R L, Wolf S (2019). Reviews and syntheses: turning the challenges of partitioning ecosystem evaporation and transpiration into opportunities.Biogeosciences, 16(19): 3747–3775

[40] Swain D L, Langenbrunner B, Neelin J D, Hall A (2018). Increasing precipitation volatility in twenty-first-century California.Nat Clim Chang, 8(5): 427–433

[41] Tucker C J (1979). Red and photographic infrared linear combinations for monitoring vegetation.Remote Sens Environ, 8(2): 127–150

[42] Turner D P, Ritts W D, Cohen W B, Gower S T, Zhao M, Running S W, Wofsy S C, Urbanski S, Dunn A L, Munger J W (2003). Scaling Gross Primary Production (GPP) over boreal and deciduous forest landscapes in support of MODIS GPP product validation.Remote Sens Environ, 88(3): 256–270

[43] Wagle P, Kakani V G (2014). Environmental control of daytime net ecosystem exchange of carbon dioxide in switchgrass.Agric Ecosyst Environ, 186: 170–177

[44] Wagle P, Xiao X, Scott R L, Kolb T E, Cook D R, Brunsell N, Baldocchi D D, Basara J, Matamala R, Zhou Y, Bajgain R (2015a). Biophysical controls on carbon and water vapor fluxes across a grassland climatic gradient in the United States.Agric For Meteorol, 214–215: 293–305

[45] Wagle P, Xiao X, Suyker A E (2015b). Estimation and analysis of gross primary production of soybean under various management practices and drought conditions.ISPRS J Photogramm Remote Sens, 99: 70–83

[46] Wagle P, Xiao X, Torn M S, Cook D R, Matamala R, Fischer M L, Jin C, Dong J, Biradar C (2014). Sensitivity of vegetation indices and gross primary production of tallgrass prairie to severe drought.Remote Sens Environ, 152: 1–14

[47] Wang X, Xiao X, Zou Z, Chen B, Ma J, Dong J, Doughty R B, Zhong Q, Qin Y, Dai S, Li X, Zhao B, Li B (2020). Tracking annual changes of coastal tidal flats in China during 1986–2016 through analyses of Landsat images with Google Earth Engine.Remote Sens Environ, 238: 110987

[48] Winslow J C, Hunt E R Jr, Piper S C (2003). The influence of seasonal water availability on global C3 versus C4 grassland biomass and its implications for climate change research.Ecol Modell, 163(1−2): 153–173

[49] Xiao X (2006). Light absorption by leaf chlorophyll and maximum light use efficiency.IEEE Trans Geosci Remote Sens, 44(7): 1933–1935

[50] Xiao X, Biradar C M, Czarnecki C, Alabi T, Keller M (2009). A simple algorithm for large-scale mapping of evergreen forests in tropical America, Africa and Asia.Remote Sens (Basel), 1(3): 355–374

[51] Xiao X, Hollinger D, Aber J, Goltz M, Davidson E A, Zhang Q, Moore B III (2004a). Satellite-based modeling of gross primary production in an evergreen needleleaf forest.Remote Sens Environ, 89(4): 519–534

[52] Xiao X, Zhang Q, Braswell B, Urbanski S, Boles S, Wofsy S, Moore B, Ojima D (2004b). Modeling gross primary production of temperate deciduous broadleaf forest using satellite images and climate data.Remote Sens Environ, 91(2): 256–270

[53] Xin F, Xiao X, Zhao B, Miyata A, Baldocchi D, Knox S, Kang M, Shim K, Min S, Chen B, Li X, Wang J, Dong J, Biradar C (2017). Modeling gross primary production of paddy rice cropland through analyses of data from CO₂ eddy flux tower sites and MODIS images.Remote Sens Environ, 190: 42–55

[54] Xu L, Baldocchi D D (2004). Seasonal variation in carbon dioxide exchange over a Mediterranean annual grassland in California.Agric For Meteorol, 123(1−2): 79–96

[55] Yan H, Fu Y, Xiao X, Huang H Q, He H, Ediger L (2009). Modeling gross primary productivity for winter wheat–maize double cropping system using MODIS time series and CO₂ eddy flux tower data.Agric Ecosyst Environ, 129(4): 391–400

[56] Zhang Y, Xiao X, Wu X, Zhou S, Zhang G, Qin Y, Dong J (2017). A global moderate resolution dataset of gross primary production of vegetation for 2000–2016.Sci Data, 4(1): 170165

[57] Zhao M, Heinsch F A, Nemani R R, Running S W (2005). Improvements of the MODIS terrestrial gross and net primary production global data set.Remote Sens Environ, 95(2): 164–176

[58] Zheng Y, Shen R, Wang Y, Li X, Liu S, Liang S, Chen J M, Ju W, Zhang L, Yuan W (2020). Improved estimate of global gross primary production for reproducing its long-term variation, 1982–2017.Earth Syst Sci Data, 12(4): 2725–2746

[59] Zhou S, Yu B, Zhang Y, Huang Y, Wang G (2016). Partitioning evapotranspiration based on the concept of underlying water use efficiency.Water Resour Res, 52(2): 1160–1175

[60] Zhu X, Pei Y, Zheng Z, Dong J, Zhang Y, Wang J, Chen L, Doughty R B, Zhang G, Xiao X (2018). Underestimates of grassland gross primary production in MODIS standard products.Remote Sens (Basel), 10(11): 1771

RIGHTS & PERMISSIONS

Higher Education Press