Search

The contemporary carbon balance over the Tibetan Plateau is highly uncertain with a ten-fold difference between various estimates. In a warming world, the potential exists for a large carbon release from its permafrost which could compromise China's 2060 carbon-neutral goal. Here, we used a satellite- and inventory-based approach, ecosystem models, and atmospheric inversions to estimate that the carbon sink was 33.12–37.84 TgC yr−1 during 2000–2015. The carbon sink induced by climate change and increasing CO2 levels largely overcompensated for a livestock grazing-induced carbon source of 0.38 TgC yr−1. By 2060, the carbon sink is projected to increase by 38.3–74.5% under moderate to high emissions scenarios, with the enhanced vegetation carbon uptake outweighing the warming-induced permafrost carbon release. The restoration of degraded grassland could sequestrate an additional 9.06 TgC yr−1, leading to a total carbon sink of 57.78–70.52 TgC yr−1. We conclude that the Tibetan Plateau's ecosystems absorbed two-and-a-half times the amount of its cumulative fossil CO2 emissions during 2000–2015 and that their carbon sinks will almost double in strength in the future, helping to achieve China's pledge to become carbon neutral by 2060. [ABSTRACT FROM AUTHOR]

• History development of land carbon sink modeling was overviewed. • Substantial uncertainties exist in estimating land carbon sink based on TEMs. • Ongoing observations, big data and model-data fusion provide new opportunities. Terrestrial ecosystems play an important role in regulating the balance of global carbon cycle by sequestrating CO 2 of atmosphere. Terrestrial ecosystem models are a critical tool for quantifying the magnitude, interannual variability and long-term trends of the land carbon sink across various spatial and temporal scales; however, despite extensive research, large uncertainties and challenges still persist. This review first summarizes decades of history in ecosystem model development in terms of model theory and methods. We then identify model uncertainties, including those arising from model algorithms, parameterization and forcing data. Finally, we propose new opportunities to improve ecosystem models for accurately simulating the land carbon sink, including emerging process-based knowledge from observations and big data, as well as model-data fusion methods. [ABSTRACT FROM AUTHOR]

Crop harvested carbon (HC) is one of the most important components of the carbon cycle in cropland ecosystems, with a significant impact on the carbon budget of croplands. China is one of the most important crop producers, however, it is still unknown on the spatial and temporal variations of HC. This study collected statistical data on crop production at the province and county levels in China for all ten crop types from 1981 to 2020 and analyzed the magnitude and long-term trend of harvested crop carbon. Our results found a substantial increase of HC in cropland from 0.185 Gt C yr ^−1 in 1981 to 0.423 Gt C yr ^−1 in 2020 at a rate of 0.006 Gt C yr ^−1 . The results also highlighted that the average annual carbon sink removal from crop harvesting in China from 1981 to 2020 was 0.32 Gt C yr ^−1 , which was comparable to the net carbon sink of the entire terrestrial ecosystems in China. This study further generated a gridded dataset of HC from 2001 to 2019 in China by using jointly the statistical crop production and distribution maps of cropland. In addition, a model-data comparison was carried out using the dataset and results from seven state-of-the-art terrestrial ecosystem models, revealing substantial disparities in HC simulations in China compared to the dataset generated in the study. This study emphasized the increased importance of HC for estimating cropland carbon budget, and the produced dataset is expected to contribute to carbon budget estimation for cropland ecosystems and the entire China.

Soil heterotrophic respiration (Rh) refers to the flux of CO2 released from soil to atmosphere as a result of organic matter decomposition by soil microbes and fauna. As one of the major fluxes in the global carbon cycle, large uncertainties still exist in the estimation of global Rh, which further limits our current understanding of carbon accumulation in soils. Here, we applied a Random Forest algorithm to create a global data set of soil Rh, by linking 761 field observations with both abiotic and biotic predictors. We estimated that global Rh was 48.8 ± 0.9 Pg C year−1 for 1982–2018, which was 16% less than the ensemble mean (58.6 ± 9.9 Pg C year−1) of 16 terrestrial ecosystem models. By integrating our observational Rh with independent soil carbon stock data sets, we obtained a global mean soil carbon turnover time of 38.3 ± 11 year. Using observation‐based turnover times as a constraint, we found that terrestrial ecosystem models simulated faster carbon turnovers, leading to a 30% (74 Pg C) underestimation of terrestrial ecosystem carbon accumulation for the past century, which was especially pronounced at high latitudes. This underestimation is equivalent to 45% of the total carbon emissions (164 Pg C) caused by global land‐use change at the same time. Our analyses highlight the need to constrain ecosystem models using observation‐based and locally adapted Rh values to obtain reliable projections of the carbon sink capacity of terrestrial ecosystems. [ABSTRACT FROM AUTHOR]

Abstract Vegetation plays a key role in regulating the material and energy exchanges among the biosphere, the atmosphere, and the pedosphere. Modeling and predicting vegetation key variables such as leaf area index (LAI) and gross primary productivity (GPP) are crucial to understand and project the processes of vegetation growth in response to climate change. While a number of studies developed models to simulate vegetation GPP using satellite‐derived LAI, the requirement of satellite‐based model inputs largely limits the predicting power of these developed models. This study developed a machine learning scheme, utilizing both support vector regression (SVR) and random forests (RF), which are capable of modeling LAI and GPP time series using only meteorological variables. We first simulated the LAI time series directly using meteorological variables as inputs and then buffered its unrealistic day‐to‐day fluctuation, and further modeled the GPP time series using meteorological variables and modeled LAI time series. This scheme enhanced the interpretability of machine learning models by considering the nonnegligible coupling between LAI and GPP. We tested our methods for four main plant functional types across North America and evaluated the models using both satellite‐based and flux tower data. The results demonstrated that the machine learning models perform well on simulating the time series of both LAI and GPP. We identified that there is a need to improve the phenology representation in the Biome‐BGCMuSo model. The machine learning models provide an alternative way to predict time series of LAI and GPP using only meteorological variables across large geographic regions, and also provide benchmarking accuracies for future developments of the process‐based models.

Abstract Wetland methane (CH4) emissions comprise about one‐third of the global CH4 source. The latitudinal distribution and climate sensitivity of wetland CH4 fluxes are the key determinants of the global CH4‐climate feedback. However, large differences exist between bottom‐up estimates, informed by ground‐based flux measurements, and top‐down estimates derived from spaceborne total column CH4. Despite the extensive coverage of satellite CH4 concentration observations, challenges remain with using top‐down estimates to test bottom‐up models, mainly because of the uncertainties in the satellite retrievals, the model representation errors, the variable prior emissions, and the confounding role of the posterior error covariance structures. Here, we use satellite‐based top‐down CH4 flux estimates (2010–2012) to test and refine 42 bottom‐up estimates of wetland emissions that use a range of hypothesized wetland extents and process controls. Our comparison between bottom‐up models and satellite‐based fluxes innovatively accounts for cross‐correlations and spatial uncertainties typically found in top‐down inverse estimates, such that only the information from satellite observations and the atmospheric transport model is kept as a constraint. We present a satellite‐constrained wetland CH4 ensemble product derived from assembling the highest‐performance bottom‐up models, which estimates global wetland CH4 emissions of 148 (117–189, 5th–95th percentile) Tg CH4 yr−1. We find that tropical wetland emissions contribute 72% (63%–85%) to the global wetland total. We also find that a lower‐than‐expected temperature sensitivity agrees better with atmospheric CH4 measurements. Overall, our approach demonstrates the potential for using satellites to quantitatively refine bottom‐up wetland CH4 emission estimates, their latitudinal distributions, and their sensitivity to climate.

Vegetation plays a key role in regulating the material and energy exchanges among the biosphere, the atmosphere, and the pedosphere. Modeling and predicting vegetation key variables such as leaf area index (LAI) and gross primary productivity (GPP) are crucial to understand and project the processes of vegetation growth in response to climate change. While a number of studies developed models to simulate vegetation GPP using satellite‐derived LAI, the requirement of satellite‐based model inputs largely limits the predicting power of these developed models. This study developed a machine learning scheme, utilizing both support vector regression (SVR) and random forests (RF), which are capable of modeling LAI and GPP time series using only meteorological variables. We first simulated the LAI time series directly using meteorological variables as inputs and then buffered its unrealistic day‐to‐day fluctuation, and further modeled the GPP time series using meteorological variables and modeled LAI time series. This scheme enhanced the interpretability of machine learning models by considering the nonnegligible coupling between LAI and GPP. We tested our methods for four main plant functional types across North America and evaluated the models using both satellite‐based and flux tower data. The results demonstrated that the machine learning models perform well on simulating the time series of both LAI and GPP. We identified that there is a need to improve the phenology representation in the Biome‐BGCMuSo model. The machine learning models provide an alternative way to predict time series of LAI and GPP using only meteorological variables across large geographic regions, and also provide benchmarking accuracies for future developments of the process‐based models. Plain Language Summary: Vegetation growth interacts external environment mainly through phenology (the periodic events in vegetation biological life cycles) and photosynthesis processes. Leaf area index (LAI, an index quantifies the vegetation leaf area during phenological cycles) and gross primary productivity (GPP, a metric quantifies carbon sequestration in photosynthesis) are important to monitor the processes of vegetation growth in response to climate change. While a number of models simulate vegetation GPP using satellite‐derived LAI (retrieved from remote sensing variables), accounting that satellite‐derived data is not able for predicting simulation, we developed a machine learning scheme to simulate LAI and GPP only via meteorological variables, without utilizing remote sensing data. The developed scheme considers the strong coupling between LAI and GPP and thus enhances the interpretability. We conducted the models in four main vegetation types across North America and evaluated the models in both continental and site scales. The results demonstrated that the machine learning models perform well on simulating both LAI and GPP. We identified that there is a need to improve the phenology representation in the Biome‐BGCMuSo model (a biogeochemical process‐based model). The machine learning models provide an alternative way to predict LAI and GPP using only meteorological variables. Key Points: We developed an interpretable machine learning scheme to simulate leaf area index (LAI) and gross primary productivity (GPP) via only meteorological variablesThe results demonstrated that the machine learning models perform well on simulating the time series of both LAI and GPPThe machine learning models provide benchmarking accuracies for future developments of the process‐based models [ABSTRACT FROM AUTHOR]

Vegetation/Ecosystem Modeling and Analysis Project (VEMAP) Phase 2 model experiments investigated the response of biogeochemical and dynamic global vegetation models (DGVMs) to differences in climate over the conterminous United States. This was accomplished by simulating ecosystem processes using historical climate and atmospheric CO2 records from 1895-1993. We evaluated the behavior of six models (Biome-BGC, Century, GTEC, LPJ, MC1, and TEM) by comparing simulated runoff in 13 watersheds to gauged streamflow from the Hydro-Climatic Data Network. Metrics used to assess the "goodness of fit" between simulated and observed values were: (1) Pearson's r to evaluate the overall data set, (2) Kendall's T to gauge seasonality trends as derived from a time-series analysis of monthly runoff, and (3) three measures of absolute and relative error. We found small differences in performance among the six models over all watersheds. However, the models yielded highly divergent results depending upon the watershed analyzed. Performance of the ensemble of models in a watershed was positively correlated with observed streamflow: models in the wettest watersheds in this study were associated with the highest model correlations and largest absolute errors, and models in the driest watersheds were associated with the lowest correlations and smallest absolute errors. Mean relative error was small and nearly constant across watersheds. A bias estimator showed that the models tended to underestimate runoff in wet watersheds and overestimate runoff in dry watersheds. Analysis of long-term trends in runoff using a moving-average approach demonstrated the ability of the models to reproduce temporal variation in observed data, even though quantitative differences among models were large. Models relying on prescribed vegetation (Biome-BGC, Century, and TEM) outperformed the two DGVMs (LPJ and MC1); GTEC gave the poorest fit to observations due to the absence of an evaporation function and a snow routine. Across all 13 watersheds, TEM ranked the highest in model performance. The validation results presented here suggest that improvements in the simulation of hydrologic processes in land-surface models will come, in part, from a more realistic representation of subgrid-scale soil moisture and from a more detailed understanding and representation of subsurface processes.

The CO _2 efflux from soil (soil respiration (SR)) is one of the largest fluxes in the global carbon (C) cycle and its response to climate change could strongly influence future atmospheric CO _2 concentrations. Still, a large divergence of global SR estimates and its autotrophic (AR) and heterotrophic (HR) components exists among process based terrestrial ecosystem models. Therefore, alternatively derived global benchmark values are warranted for constraining the various ecosystem model output. In this study, we developed models based on the global soil respiration database (version 5.0), using the random forest (RF) method to generate the global benchmark distribution of total SR and its components. Benchmark values were then compared with the output of ten different global terrestrial ecosystem models. Our observationally derived global mean annual benchmark rates were 85.5 ± 40.4 (SD) Pg C yr ^−1 for SR, 50.3 ± 25.0 (SD) Pg C yr ^−1 for HR and 35.2 Pg C yr ^−1 for AR during 1982–2012, respectively. Evaluating against the observations, the RF models showed better performance in both of SR and HR simulations than all investigated terrestrial ecosystem models. Large divergences in simulating SR and its components were observed among the terrestrial ecosystem models. The estimated global SR and HR by the ecosystem models ranged from 61.4 to 91.7 Pg C yr ^−1 and 39.8 to 61.7 Pg C yr ^−1 , respectively. The most discrepancy lays in the estimation of AR, the difference (12.0–42.3 Pg C yr ^−1 ) of estimates among the ecosystem models was up to 3.5 times. The contribution of AR to SR highly varied among the ecosystem models ranging from 18% to 48%, which differed with the estimate by RF (41%). This study generated global SR and its components (HR and AR) fluxes, which are useful benchmarks to constrain the performance of terrestrial ecosystem models.

Climate change has consequences for terrestrial functioning, but predictions of plant responses remain uncertain because of the gaps in the representation of nutrient cycles and C-N-P interactions in ecosystem models. Here, we review the processes that are included in ecosystem models, but focus on coupled C-N-P cycle models. We highlight important plant adjustments to climate change, elevated atmospheric CO, and/or nutrient limitations that are currently not-or only partially-incorporated in ecosystem models by reviewing experimental studies and compiling data. Plant adjustments concern C:N:P stoichiometry, photosynthetic capacity, nutrient resorption rates, allocation patterns, symbiotic N fixation and root exudation (phosphatases, carboxylates) and the effect of root exudation on nutrient mobilization in the soil rhizosphere (P solubilization, biochemical mineralization of organic P and priming effect). We showed that several plant adjustments could be formulated and calibrated using existing experimental data in the literature. Finally, we proposed a roadmap for future research because improving ecosystem models necessitate specific data and collaborations between modelers and empiricists. [ABSTRACT FROM AUTHOR]

Our limited ability to accurately simulate leaf phenology is a leading source of uncertainty in models of ecosystem carbon cycling. We evaluate if continuously updating canopy state variables with observations is beneficial for predicting phenological events. We employed ensemble adjustment Kalman filter (EAKF) to update predictions of leaf area index (LAI) and leaf extension using tower-based photosynthetically active radiation (PAR) and moderate resolution imaging spectrometer (MODIS) data for 2002-2005 at Willow Creek, Wisconsin, USA, a mature, even-aged, northern hardwood, deciduous forest. The ecosystem demography model version 2 (ED2) was used as the prediction model, forced by offline climate data. EAKF successfully incorporated information from both the observations and model predictions weighted by their respective uncertainties. The resulting estimate reproduced the observed leaf phenological cycle in the spring and the fall better than a parametric model prediction. These results indicate that during spring the observations contribute most in determining the correct bud-burst date, after which the model performs well, but accurately modeling fall leaf senesce requires continuous model updating from observations. While the predicted net ecosystem exchange (NEE) of CO2 precedes tower observations and unassimilated model predictions in the spring, overall the prediction follows observed NEE better than the model alone. Our results show state data assimilation successfully simulates the evolution of plant leaf phenology and improves model predictions of forest NEE. [ABSTRACT FROM AUTHOR]

Terrestrial ecosystem primary productivity is a key indicator of ecosystem functions, including, but not limited to, carbon storage, provision of food and fiber, and sustaining biodiversity. However, measuring terrestrial ecosystem primary productivity in the field is extremely laborious and expensive. Optical remote sensing has revolutionized our ability to map terrestrial ecosystem primary productivity over large areas ranging from regions to the entire globe in a repeated, cost-efficient manner. This progress report reviews the theory and practice of mapping terrestrial primary productivity using optical remotely sensed data. Terrestrial ecosystem primary productivity is generally estimated with optical remote sensing via one of the following approaches: (1) empirical estimation from spectral vegetation indices; (2) models that are based on light-use-efficiency (LUE) theory; (3) models that are not based on LUE theory, but the biophysical processes of plant photosynthesis. Among these three, models based on LUE are the primary approach because there is a solid physical basis for the linkage between fraction of absorbed photosynthetically active radiation (fAPAR) and remotely sensed spectral signatures of vegetation. There has been much inconsistency in the literature with regard to the appropriate value for LUE. This issue should be resolved with the ongoing efforts aimed at direct mapping of LUE from remote sensing. At the same time, major efforts have been dedicated to mapping vegetation canopy biochemical composition via imaging spectroscopy for use in process-based models to estimating primary productivity. In so doing, optical remote sensing will continue to play a vital role in global carbon cycle science research. [ABSTRACT FROM PUBLISHER]

Uncertainties in the 20th century carbon budget associated with the treatment of land use change (LUC) are assessed using the Canadian Centre for Climate Modelling and Analysis (CCCma) first-generation Earth System Model (CanESM1). Eight coupled climate carbon cycle simulations are performed using different reconstructions of 1850-2000 land cover derived from historical information on changes in cropland and pasture area. The simulations provide estimates of the emissions associated with LUC, the relative contribution of changes in cropland and pasture to LUC emissions and the uncertainty associated with differences among historical data sets of crop area as well as in the manner in which the historical land cover data are constructed. The resulting estimates of the amount of biomass deforested over the 1850-2000 period range from 63 to 145 Pg C with cumulative implied LUC emissions ranging from 40 to 77 Pg C. These values of LUC emissions are considerably lower than Houghton's estimate of 156 Pg C. The year 2000 atmospheric CO concentration ranges between 371.1 ± 3.7 ppm depending on the data set used and the manner in which historical land cover is constructed. This compares to the observed value of 369.6 ppm at Mauna Loa and is 17.3 ± 6.3 ppm larger than for simulations without LUC. Although increases in cropland result in the expected increase in LUC emissions, changes in pasture area decrease these emissions because of carbon sequestration in soils. [ABSTRACT FROM AUTHOR]

The article presents a study which presents a new approach for measuring the magnitude and spatial pattern of nitrous oxide emissions from tropical forest soils of the Amazon Basin. It states that the study develops an empirical relationship between carbon dioxide and nitrous oxide emissions from tropical soils based on seven years of field measurements made in forests of Rondonia, Brazil. Also, the study combines the relationship with monthly figues of forest soil respiration across the basin.