Reply on RC1

<strong class="journal-contentHeaderColor">Abstract.</strong> We develop high resolution (1 km) estimates of Gross Primary Productivity (GPP), Ecosystem Respiration (ER) and Net Ecosystem Exchange (NEE) over the Australian continent for the period January 2003 to June 2022 by empirical upscaling of flux tower measurements. We compare our estimates with nine other products that cover the three broad categories that define current methods for estimating the terrestrial carbon cycle and assess if consiliences between datasets can point to the correct dynamics of Australia&rsquo;s carbon cycle. Our results indicate that regional empirical upscaling greatly improves upon the existing global empirical upscaling efforts, outperforms process-based models, and agrees much better with the dynamics of CO₂ flux over Australia as estimated by two regional atmospheric inversions. Our nearly 20-year estimates of terrestrial carbon fluxes revealed Australia is a strong net carbon sink of -0.44 (IQR=0.42) PgC/year on-average, with an inter-annual variability of 0.18 PgC/year and an average seasonal amplitude of 0.85 PgC/yr. Annual mean carbon uptake estimated from other methods ranged considerably, while carbon flux anomalies showed much better agreement between methods. NEE anomalies were predominately driven by cumulative rainfall deficits and surpluses, resulting in larger anomalous responses from GPP over ER. In contrast, we show that the long-term average seasonal cycle is dictated more by the variability in ER than GPP, resulting in peak carbon uptake typically occurring during the cooler, drier Austral autumn, and winter months. This new estimate of Australia&rsquo;s terrestrial carbon cycle provides a benchmark for assessment against Land Surface Model simulations, and a means for monitoring of Australia&rsquo;s terrestrial carbon cycle at an unprecedented high-resolution. We call this new estimate of Australia&rsquo;s terrestrial carbon cycle, &ldquo;AusEFlux&rdquo; (Australian Empirical Fluxes).

Inaccurate parameter estimation is a significant source of uncertainty in complex terrestrial biosphere models. Model parameters may have large spatial variability, even within a vegetation type. Model uncertainty from parameter uncertainty can be significantly reduced by model-data fusion (MDF), which, however, is difficult to implement over a large region with traditional methods due to the high computational cost. This study proposed a hybrid modeling approach that couples a terrestrial biosphere model with a data-driven machine learning method, which uses satellite information and considers the physical mechanisms. We developed a two-step framework to estimate the essential parameters of the revised Integrated Biosphere Simulator (IBIS) pixel by pixel using the satellite-derived leaf area index (LAI) and gross primary productivity (GPP) products as “true values.” The first step was to estimate the optimal parameters for each sample using a modified adaptive surrogate modeling algorithm (MASM). We applied the Gaussian Process Regression algorithm (GPR) as a surrogate model to learn the relationship between model parameters and errors. In our second step, we built an eXtreme Gradient Boosting (XGBoost) model between the optimized parameters and local environmental variables. The trained XGBoost model was then used to predict optimal parameters spatially across the deciduous forests in the eastern United States. The results showed that the parameters were highly variable spatially and quite different from the default values over forests, and the simulation errors of the GPP and LAI could be markedly reduced with the optimized parameters. The effectiveness of the optimized model in estimating GPP, ecosystem respiration (ER) and net ecosystem exchange (NEE) were also tested through site validation. The optimized model reduced the root-mean-squared-error (RMSE) from 7.03 to 6.22 gC m-2 8 d-1 for GPP, 2.65 to 2.11 gC m-2 8 d-1 for ER and 4.45 to 4.38 gC m-2 8 d-1 for NEE. The mean annual GPP, ER and NEE of the region from 2000 to 2019 were 5.79, 4.60 and -1.19 Pg year-1, respectively. The strategy used in this study requires only a few hundred model runs to calibrate regional parameters and is readily applicable to other complex terrestrial biosphere models with different spatial resolutions. Our study also emphasizes the necessity of pixel-level parameter calibration and the value of remote sensing products for per-pixel parameter optimization.

We investigated the relative importance of climatic versus biotic controls on gross primary production (GPP) and water vapor fluxes in seasonally drought‐affected ponderosa pine forests. The study was conducted in young (YS), mature (MS), and old stands (OS) over 4 years at the AmeriFlux Metolius sites. Model simulations showed that interannual variation of GPP did not follow the same trends as precipitation, and effects of climatic variation were smallest at the OS (&lt;10%), largest at the MS (&gt;50%), and intermediate at the YS (&lt;20%). In the young, developing stand, interannual variation in leaf area has larger effects on fluxes than climate, although leaf area is a function of climate in that climate can interact with age‐related shifts in carbon allocation and affect whole‐tree hydraulic conductance. Older forests, with well‐established root systems, appear to be better buffered from effects of seasonal drought and interannual climatic variation. Interannual variation of net ecosystem exchange (NEE) was also lowest at the OS, where NEE is controlled more by interannual variation of ecosystem respiration, 70% of which is from soil, than by the variation of GPP, whereas variation in GPP is the primary reason for interannual changes in NEE at the YS and MS. Across spatially heterogeneous landscapes with high frequency of younger stands resulting from natural and anthropogenic disturbances, interannual climatic variation and change in leaf area are likely to result in large interannual variation in GPP and NEE.

Algorithms that use remotely-sensed vegetation indices to estimate gross primary production (GPP), a key component of the global carbon cycle, have gained a lot of popularity in the past decade. Yet despite the amount of research on the topic, the most appropriate approach is still under debate. As an attempt to address this question, we compared the performance of different vegetation indices from the Moderate Resolution Imaging Spectroradiometer (MODIS) in capturing the seasonal and the annual variability of GPP estimates from an optimal network of 21 FLUXNET forest towers sites. The tested indices include the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Leaf Area Index (LAI), and Fraction of Photosynthetically Active Radiation absorbed by plant canopies (FPAR). Our results indicated that single vegetation indices captured 50–80% of the variability of tower-estimated GPP, but no one index performed universally well in all situations. In particular, EVI outperformed the other MODIS products in tracking seasonal variations in tower-estimated GPP, but annual mean MODIS LAI was the best estimator of the spatial distribution of annual flux-tower GPP (GPP = 615 × LAI − 376, where GPP is in g C/m2/year). This simple algorithm rehabilitated earlier approaches linking ground measurements of LAI to flux-tower estimates of GPP and produced annual GPP estimates comparable to the MODIS 17 GPP product. As such, remote sensing-based estimates of GPP continue to offer a useful alternative to estimates from biophysical models, and the choice of the most appropriate approach depends on whether the estimates are required at annual or sub-annual temporal resolution.

Abstract. The terrestrial carbon cycle plays a critical role in modulating the interactions of climate with the Earth system, but different models often make vastly different predictions of its behavior. Efforts to reduce model uncertainty have commonly focused on model structure, namely by introducing additional processes and increasing structural complexity. However, the extent to which increased structural complexity can directly improve predictive skill is unclear. While adding processes may improve realism, the resulting models are often encumbered by a greater number of poorly determined or over-generalized parameters. To guide efficient model development, here we map the theoretical relationship between model complexity and predictive skill. To do so, we developed 16 structurally distinct carbon cycle models spanning an axis of complexity and incorporated them into a model–data fusion system. We calibrated each model at six globally distributed eddy covariance sites with long observation time series and under 42 data scenarios that resulted in different degrees of parameter uncertainty. For each combination of site, data scenario, and model, we then predicted net ecosystem exchange (NEE) and leaf area index (LAI) for validation against independent local site data. Though the maximum model complexity we evaluated is lower than most traditional terrestrial biosphere models, the complexity range we explored provides universal insight into the inter-relationship between structural uncertainty, parametric uncertainty, and model forecast skill. Specifically, increased complexity only improves forecast skill if parameters are adequately informed (e.g., when NEE observations are used for calibration). Otherwise, increased complexity can degrade skill and an intermediate-complexity model is optimal. This finding remains consistent regardless of whether NEE or LAI is predicted. Our COMPLexity EXperiment (COMPLEX) highlights the importance of robust observation-based parameterization for land surface modeling and suggests that data characterizing net carbon fluxes will be key to improving decadal predictions of high-dimensional terrestrial biosphere models.

Post-fire co-stimulation of gross primary production and ecosystem respiration in a meadow grassland on the Tibetan Plateau

Abstract. Improving the accuracy of monitoring cropland CO2 exchange at heterogeneous spatial scales is of great importance for reducing spatial and temporal uncertainty in estimating terrestrial carbon (C) dynamics. In this study, an approach to estimate daily cropland C fluxes is developed and tested by combining time series of field-scale eddy covariance (EC) CO2 flux data and Sentinel-2 satellite-based vegetation indices (VIs) after appropriately accounting for the spatial alignment between the two time series datasets. The study was carried out for an agricultural field (118 ha) in the lowlands of northeastern Germany. The ability of different VIs to estimate daily net ecosystem exchange (NEE) and gross primary productivity (GPP) based on linear regression models was assessed. Most VIs showed high (&gt;0.9) and statistically significant (p&lt;0.001) correlations with GPP and NEE, although some VIs deviated from the seasonal pattern of CO2 exchange. By contrast, correlations between ecosystem respiration (Reco) and VIs were weak and not statistically significant, and no attempt was made to estimate Reco from VIs. Linear regression models explained generally more than 80 % and 70 % of the variability in NEE and GPP, respectively, with high variability among the individual VIs. The performance in estimating daily C fluxes varied among VIs depending on the C flux component (NEE or GPP) and observation period. Root mean square error (RMSE) values ranged from 1.35 g C m−2 d−1 using the green normalized difference vegetation index (GNDVI) for NEE to 5 g C m−2 d−1 using the simple ratio (SR) for GPP. This equated to an underestimated net C uptake of only 41 g C m−2 (18 %) and an overestimation of gross C uptake of 854 g C m−2 (73 %). Differences between the measured and estimated C fluxes were mainly explained by the diversion of the C flux and VI signal during winter when C uptake remained low, while VI values indicated an increased C uptake due to relatively high crop leaf area. Overall, the results exhibited similar error margins to mechanistic crop models. Thus, they indicated the suitability and expandability of the proposed approach for monitoring cropland C exchange with satellite-derived VIs.

Seasonal variations and driving mechanisms of CO2 fluxes over a winter-wheat and summer-maize rotation cropland in the North China plain

Recent remote-sensing-based global carbon, water and energy budgets over land still include considerable uncertainties. Most existing flux products of terrestrial carbon, water and energy components were developed individually, despite the inherently coupled processes among them. In this study, we present a new set of global daily surface downwelling shortwave radiation (SW), net radiation (Rnet), evapotranspiration (ET), gross primary productivity (GPP), terrestrial ecosystem respiration (TER) and net ecosystem exchange (NEE) datasets at 0.05° resolutions from 1982 to 2019, by improving a satellite-based and coupled-process model—the Breathing Earth System Simulator (BESS). The new version of BESS (v2.0) integrated a newly developed ecosystem respiration module, an optimality-based maximum carboxylation rate (Vcmax) model, and extended the temporal coverage of flux datasets from 1982 to 2019. We evaluated BESS products against the FLUXNET2015 dataset at the site scale, and against several remote sensing and/or machine learning products on a global scale. At the site scale, BESS products agreed well with FLUXNET measurements, capturing 84%, 53%, 65%, 51% and 31% of daily variation in Rnet, ET, GPP, TER and NEE, respectively. Interannual variation in BESS NEE showed relatively low consistency with the FLUXNET measurements, while the rest fluxes explained approximately half of the interannual variation. On a global scale, we found marked discrepancies in spatio-temporal patterns between BESS and several benchmark products. Over the period 1982–2019, BESSv2.0 estimated the mean annual global Rnet, ET, GPP, TER and NEE to be 340.97 ± 5.22 ZJ yr−1 (mean ± 1 SD), 67.67 ± 0.71 103 km3 yr−1, 125.74 ± 5.95 Pg C yr−1, 109.30 ± 3.16 Pg C yr−1, and − 16.28 ± 2.95 Pg C yr−1, respectively, with significant annual linear trends (P < 0.01) by −0.05 103 km3 yr−2 for ET, by 0.52 Pg C yr−2 for GPP, by 0.28 Pg C yr−2 for TER, and by −0.25 Pg C yr−2 for NEE. We further evaluated various coupled processes derived by BESS in terms of functional properties (i.e., Budyko relation, carbon-use efficiency, water-use efficiency, and light-use efficiency), which agreed well with FLUXNET observations, unlike the benchmark products. Overall, BESSv2.0 can serve as a set of reliable and independent products from other global satellite products, facilitating studies related to global carbon, water and energy budgets in a coupled and comprehensive manner.

Boreal forests in the northern hemisphere provide important sinks for storing carbon dioxide (CO2). However, the size and distribution of these sinks remain uncertain. In particular, many remote-sensing models show a strong bias in the simulation of carbon fluxes for evergreen needleleaf forest. The objective of this study is to improve these predictive models for accurately quantifying temporal changes in the net ecosystem exchange (NEE) of conifer-dominated forest solely based on satellite remote sensing, including the Moderate Resolution Imaging Spectroradiometer (MODIS) Terra daytime land-surface temperature (LST), night-time LST′, enhanced vegetation index (EVI), land–surface water index (LSWI), fraction of absorbed photosynthetically active radiation (FPAR), and leaf area index (LAI). Considering that the component fluxes, gross primary production (GPP), and ecosystem respiration (Re), are strongly influenced by vegetation phenology, seasonality information was extracted from time-series MODIS EVI data based on non-linear least-squares fits of asymmetric Gaussian model functions with a software package for analysing the time-series of satellite sensor data (TIMESAT). The results indicated that models directly incorporating phenological information failed to improve their performance for temperate deciduous forest. Instead, three methods to retrieve the component fluxes – GPP and Re – including direct estimates, models incorporating the phenological information, and models developed based on the threshold value (LST 273 K), were explored respectively. All methods improved NEE estimates markedly and models developed based on the threshold value performed best, and provided a future framework for accurate remote sensing of NEE in evergreen forest.

The measurement of carbon exchange between vegetation and the atmosphere is vital to quantify the impact of environmental variables on the carbon sequestration capacity of forests, and to predict how they will respond to future climate. In this study we use proximal remote sensing, defined as observations made from non-contact radiometric or imaging sensors in close proximity to the forest canopy (10–20 m), as an intermediate upscaling tool between direct measurements of carbon fluxes and satellite-derived estimations of primary productivity in a tropical dry forest (TDF) in Jalisco, Mexico. Two broad-band vegetation indices (VIs), the normalized difference VI and the enhanced vegetation index 2 (EVI2), were calculated from proximally sensed canopy properties, validated with field estimates of the fraction of absorbed photosynthetically active radiation by photosynthetic tissue (fAPARgreen), and compared to estimates of gross primary productivity (GPP) and net ecosystem exchange of CO2, measured from a flux tower. The VIs captured the phenology of the TDF, both under typical summer rainfall and during an atypically-dry wet season in El Niño of 2009. The VIs also tracked a secondary leaf-flushing in the dry season of 2010. Our study suggests that (1) VIs are the best predictors of gross carbon uptake, able to explain up to 86% of variations in GPP; (2) VIs are accurate predictors of the photosynthetic capacity of green tissue, able to explain up to 99% of fAPARgreen variation; and (3) VIs and soil water content can be used to develop an empirical model that captures the seasonal trajectory of GPP from high respiration after the rain pulses, to rapid leaf development, and finally to slow senescence as the soil dries out. Proximal remote sensing constitutes a useful tool to link field-base measurements of carbon fluxes to satellite- or airborne-derived estimates of carbon exchange.

An accurate estimation of vegetation Gross Primary Productivity (GPP), which is the amount of carbon taken up by vegetation through photosynthesis for a given time and area, is critical for understanding terrestrial-atmosphere CO2 exchange processes, ecosystem functioning, and as well as ecosystem responses and adaptations to climate change. Earliest studies, based on ground, airborne and satellite Sun-Induced chlorophyll Fluorescence (SIF) observations have recently revealed close relationships with GPP at different spatial and temporal scales and across different plant functional type (PFT). However, questions remain regarding whether there is a unique relationship between SIF and GPP across different sites and PFT and how can we improve GPP estimates using solely remotely sensed data. Using concurrent measurements of daily TROPOMI (TROPOspheric Monitoring Instrument) SIF (daily SIFd), daily MODIS Terra and Aqua spectral reflectance, and vegetation indices (VIs, notably NDVI (normalized difference vegetation index), NIRv (near-infrared reflectance of vegetation) and PRI (photochemical reflectance index)) and daily tower-based GPP across eight major different PFT, including mixed forests, deciduous broadleaf forests, croplands, evergreen broadleaf forests, evergreen needleleaf forests, grassland, open shrubland, and wetland, the strength of the linear relationships between tower-based GPP and SIFd at 40 ICOS (Integrated Carbon Observation Systems) flux sites was investigated, and the synergy between these variables to improve GPP estimates using a data-driven modelling approach was evaluated. The results revealed that the strength of the linear relationship between GPP and SIFd was strongly site-specific and PFT-dependent. Furthermore, the GLM (Generalized Linear Model) model, fitted between SIFd, GPP, site and vegetation type as categorical variables, further supported this site-and PFT-dependent relationship between GPP and SIFd. This study also showed that the spectral reflectance bands (RF-R), SIFd plus spectral reflectance (RF-SIF-R) models explained over 80 % of the seasonal and interannual variations in GPP, whereas the SIFd plus VIs (RF-SIF-VI) model reproduced only 75 % of the tower-based GPP variance. In addition, the relative importance results demonstrated that the spectral reflectance bands in the near-infrared, red and SIFd appeared as the most influential and dominant factors determining GPP predictions, indicating the importance of canopy structure, biochemical properties and vegetation functioning on GPP estimates. Overall, this study provides insights into understanding the strength of the relationships between GPP and SIF and the use of the spectral reflectance and SIFd to improve GPP across sites and PFT.

The purpose of this study was to evaluate 10 process-based terrestrial biosphere models that were used for the IPCC fifth Assessment Report. The simulated gross primary productivity (GPP) is compared with flux-tower-based estimates by Jung et al. [Journal of Geophysical Research 116 (2011) G00J07] (JU11). The net primary productivity (NPP) apparent sensitivity to climate variability and atmospheric CO2 trends is diagnosed from each model output, using statistical functions. The temperature sensitivity is compared against ecosystem field warming experiments results. The CO2 sensitivity of NPP is compared to the results from four Free-Air CO2 Enrichment (FACE) experiments. The simulated global net biome productivity (NBP) is compared with the residual land sink (RLS) of the global carbon budget from Friedlingstein et al. [Nature Geoscience 3 (2010) 811] (FR10). We found that models produce a higher GPP (133 ± 15 Pg C yr(-1) ) than JU11 (118 ± 6 Pg C yr(-1) ). In response to rising atmospheric CO2 concentration, modeled NPP increases on average by 16% (5-20%) per 100 ppm, a slightly larger apparent sensitivity of NPP to CO2 than that measured at the FACE experiment locations (13% per 100 ppm). Global NBP differs markedly among individual models, although the mean value of 2.0 ± 0.8 Pg C yr(-1) is remarkably close to the mean value of RLS (2.1 ± 1.2 Pg C yr(-1) ). The interannual variability in modeled NBP is significantly correlated with that of RLS for the period 1980-2009. Both model-to-model and interannual variation in model GPP is larger than that in model NBP due to the strong coupling causing a positive correlation between ecosystem respiration and GPP in the model. The average linear regression slope of global NBP vs. temperature across the 10 models is -3.0 ± 1.5 Pg C yr(-1) °C(-1) , within the uncertainty of what derived from RLS (-3.9 ± 1.1 Pg C yr(-1) °C(-1) ). However, 9 of 10 models overestimate the regression slope of NBP vs. precipitation, compared with the slope of the observed RLS vs. precipitation. With most models lacking processes that control GPP and NBP in addition to CO2 and climate, the agreement between modeled and observation-based GPP and NBP can be fortuitous. Carbon-nitrogen interactions (only separable in one model) significantly influence the simulated response of carbon cycle to temperature and atmospheric CO2 concentration, suggesting that nutrients limitations should be included in the next generation of terrestrial biosphere models.

Reliable modeling of carbon and water fluxes is essential for understanding the terrestrial carbon and water cycles and informing policy strategies aimed at constraining carbon emissions and improving water use efficiency. We used an assimilation framework (LPJ-Vegetation and soil moisture Joint Assimilation, or LPJ-VSJA) to improve gross primary production (GPP) and evapotranspiration (ET) estimates globally. The terrestrial biosphere model that we used is the integrated model – LPJ-PM coupled from the Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ-DGVM) and a hydrology module (i.e., the updated Priestley–Taylor Jet Propulsion Laboratory model, PT-JPLSM). Satellite-based soil moisture products derived from the Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active and Passive (SMAP) and leaf area index (LAI) from the global Land and Ground satellite (GLASS) product were assimilated into LPJ-PM to improve GPP and ET simulations using a Proper Orthogonal Decomposition-based ensemble four-dimensional variational assimilation method (PODEn4DVar). The joint assimilation framework LPJ-VSJA achieved the best model performance (with an R2 of 0.91 and 0.81 and an RMSD reduced by 50.4 % and 38.4 % for GPP and ET, respectively, compared with those of LPJ-DGVM at the monthly scale). The assimilated GPP and ET demonstrated a better performance in the arid and semiarid regions (GPP: R2 = 0.73, ubRMSD = 1.05 g C m−2 d−1; ET: R2 = 0.73, ubRMSD = 0.61 mm d−1) than in the humid and sub-dry humid regions (GPP: R2 = 0.61, ubRMSD = 1.23 g C m−2 d−1; ET: R2 = 0.66; ubRMSD = 0.67 mm d−1). The ET simulated by LPJ-PM that assimilated SMAP or SMOS had a slight difference, and the ET that assimilated SMAP soil moisture data was more improved than that assimilated SMOS data. Our global simulation modeled by LPJ-VSJA was compared with several global GPP and ET products (e.g., GLASS GPP, GOSIF GPP, GLDAS ET, GLEAM ET) using the triple collocation (TC) method. Our products, especially ET, exhibited advantages in the overall error distribution (estimated error (μ): 3.4 mm month−1; estimated standard deviation of μ: 1.91 mm month−1). Our research showed that the assimilation of multiple datasets could reduce model uncertainties, while the model performance differed across regions and plant functional types. Our assimilation framework (LPJ-VSJA) can improve the model simulation performance of daily GPP and ET globally, especially in water-limited regions.

For decades, the productivity of tropical montane cloud forests (TMCF) has been assumed to be lower than in tropical lowland forests due to nutrient limitation, lower temperatures, and frequent cloud immersion, although actual estimates of gross primary productivity (GPP) are very scarce. Here, we present the results of a process-based modeling estimate of GPP, using a soil–plant–atmosphere model, of a high elevation Peruvian TMCF. The model was parameterized with field-measured physiological and structural vegetation variables, and driven with meteorological data from the site. Modeled transpiration corroborated well with measured sap flow, and simulated GPP added up to 16.2 ± SE 1.6 Mg C ha−1 y−1. Dry season GPP was significantly lower than wet season GPP, although this difference was 17% and not caused by drought stress. The strongest environmental controls on simulated GPP were variation of photosynthetic active radiation and air temperature (T air). Their relative importance likely varies with elevation and the local prevalence of cloud cover. Photosynthetic parameters (V cmax and J max) and leaf area index were the most important non-environmental controls on GPP. We additionally compared the modeled results with a recent estimate of GPP of the same Peruvian TMCF derived by the summing of ecosystem respiration and net productivity terms, which added up to 26 Mg C ha−1 y−1. Despite the uncertainties in modeling GPP we conclude that at this altitude GPP is, conservatively estimated, 30–40% lower than in lowland rainforest and this difference is driven mostly by cooler temperatures than changes in other parameters.