Search

Agricultural nitrous oxide (N2O) emission accounts for a non-trivial fraction of global greenhouse gas (GHG) budget. To date, estimating N2O fluxes from cropland remains a challenging task because the related microbial processes (e.g., nitrification and denitrification) are controlled by complex interactions among climate, soil, plant and human activities. Existing approaches such as process-based (PB) models have well-known limitations due to insufficient representations of the processes or uncertainties of model parameters, and due to leverage recent advances in machine learning (ML) a new method is needed to unlock the "black box"to overcome its limitations such as low interpretability, out-of-sample failure and massive data demand. In this study, we developed a first-of-its-kind knowledge-guided machine learning model for agroecosystems (KGML-ag) by incorporating biogeophysical and chemical domain knowledge from an advanced PB model, ecosys, and tested it by comparing simulating daily N2O fluxes with real observed data from mesocosm experiments. The gated recurrent unit (GRU) was used as the basis to build the model structure. To optimize the model performance, we have investigated a range of ideas, including (1) using initial values of intermediate variables (IMVs) instead of time series as model input to reduce data demand; (2) building hierarchical structures to explicitly estimate IMVs for further N2O prediction; (3) using multi-task learning to balance the simultaneous training on multiple variables; and (4) pre-training with millions of synthetic data generated from ecosys and fine-tuning with mesocosm observations. Six other pure ML models were developed using the same mesocosm data to serve as the benchmark for the KGML-ag model. Results show that KGML-ag did an excellent job in reproducing the mesocosm N2O fluxes (overall r2Combining double low line0.81, and RMSECombining double low line3.6g€¯mgNm-2d-1 from cross validation). Importantly, KGML-ag always outperforms the PB model and ML models in predicting N2O fluxes, especially for complex temporal dynamics and emission peaks. Besides, KGML-ag goes beyond the pure ML models by providing more interpretable predictions as well as pinpointing desired new knowledge and data to further empower the current KGML-ag. We believe the KGML-ag development in this study will stimulate a new body of research on interpretable ML for biogeochemistry and other related geoscience processes.

The effects of dams on carbon dioxide (CO2) fluxes in downstream lakes remain elusive. Here we combined eddy covariance observations and random forest models to examine multi‐decadal variations in CO2 fluxes in the Poyang Lake, the largest freshwater lake in China, and quantified the contribution of the Three Gorges Dam (TGD), the world's largest hydraulic project. We found the lake fluctuated between CO2 source and sink in 1961–2016, and tended to be CO2 sink in the post‐TGD period (2003–2016) when vegetation expanded early and spatially due to declining water level. TGD can explain approximately 6% of the total differences in annual CO2 fluxes, with major contributions in the impoundment period (up to 22% in middle September to October). The results show a positive side of operational major hydraulic projects on lake carbon sink, and probably caution the negative side of carbon release after dam removal. Plain Language Summary: In the past century, dams have significantly altered the hydrological connectivity between rivers and lakes, which affect CO2 exchange in the downstream lake systems. As the largest freshwater lake in China, Poyang Lake has also undergone drastic hydrological changes, attributable largely to the operation of the Three Gorges Dam (TGD), the world's largest hydraulic project ever, in 2003. Based on flux observations and machine learning method, we show that annual lake CO2 exchange shifted toward carbon sink during 1961–2016. The TGD has a major impact on lake CO2 fluxes, especially during the impoundment stage in middle September–October, explaining 22% of the flux differences between the pre‐ and post‐TGD period. The results show a positive side of hydraulic projects albeit their adverse impact on ecological protection. Key Points: Poyang Lake as a CO2 source or sink significantly depends on water levelPoyang Lake became a CO2 sink since the Three Gorges Dam operation in 2003Dam explains 22% of differences in CO2 fluxes in autumn impoundment period [ABSTRACT FROM AUTHOR]

Coal mining ranks as the largest anthropogenic CH4 source in China with emission factors (EFs) varying up to 30‐fold among inventories when applied to different provinces. The lack of independent evaluation of coal mining CH4 EFs in China is one of the main uncertainties in estimating national total CH4 emissions. Shanxi province, which supplies 25% of the national coal production, is the largest coal mining CH4 emission region in China and even among the world's largest coal production regions. This area is also a significant anthropogenic CO2 source because of high‐density power and industrial activities. Given the large uncertainties in CH4 and CO2 inventories from provincial to city scales, questions remain whether state‐of‐the‐art inventories have accurately estimated these emission hotspots. Here, we evaluate CH4 and CO2 emissions from one of the world's largest coal production regions near Taiyuan City, the capital of Shanxi province, China. CH4 and CO2 concentrations were measured from March 2018 to February 2019 from a 30‐m tower. These data were used within an inverse modeling framework to simulate both CH4 and CO2 concentrations and to evaluate EFs for this region. Results show generally good agreement between observed and simulated CH4 concentrations. However, the CO2 simulations were much lower compared to the observations. Given the minor role of NEE‐induced CO2 enhancements, we believe that the large difference is attributed to the underestimation of anthropogenic CO2 emissions. In general, the derived posteriori anthropogenic CH4 emissions were 85.2(±18.1)% of a priori emissions, where fugitive CH4 from coal mining accounted for ∼92.7% of total anthropogenic emissions. The derived coal mining EF was 23.2(±4.9) m3 CH4/ton coal, close to the default value of high CH4‐content coal, but twofold the province average that were reported by previous observation‐based studies in Shanxi province, indicating large spatial inhomogeneity in the coal mining CH4 EF. The posteriori CO2 emissions were 1.6‐fold of the a priori emissions, highlighting underestimation of CO2 emissions in industrial cities and some potential large emission sources that are missing from state‐of‐the‐art inventories. Finally, we also emphasize the use of satellite observations and denser tower‐based networks are essential in resolving the spatial inhomogeneity of greenhouse gas emissions. Plain Language Summary: The understanding of anthropogenic CH4 and CO2 emissions is basis for climate mitigation especially for global top emitting countries, but the largest issue before addressing above question is that many previous studies have found considerable bias of greenhouse gas emission for almost all inventories from city to regional scales. These facts hindered the government to make and evaluate corresponding mitigation policies. Here, to quantify CH4 and CO2 emissions at one of global largest CH4 and CO2 hotspot in China, we conducted 1 year tower‐based atmospheric CH4 and CO2 concentration measurements and used atmospheric inversion method to constrain and evaluate their emissions, we found coal mining CH4 emission factor has less bias but CO2 emissions were underestimated by 1.6‐fold, highlighting underestimation of CO2 emissions in industrial cities and some potential large emission sources that are missing from state‐of‐the‐art inventories. Our findings indicate more work is needed for urban government to fully understand their greenhouse gases emissions. Key Points: CH4 emission factor for coal mining was 23.2(±4.9) m3 CH4/ton coal, close to the default value of high CH4‐content coal in ChinaThe posteriori CO2 emissions were 1.6‐fold of the a priori emissions, indicating CO2 emissions in industrial cities were largely underestimatedSome large CO2 emissions are missing and more work is needed for urban government to fully understand their greenhouse gases emissions [ABSTRACT FROM AUTHOR]

Agricultural nitrous oxide (N2O) emission accounts for a non-trivial fraction of global greenhouse gases (GHGs) budget. To date, estimating N2O fluxes from cropland remains a challenging task because the related microbial processes (e.g., nitrification and denitrification) are controlled by complex interactions among climate, soil, plant and human activities. Existing approaches such as process-based (PB) models have well-known limitations due to insufficient representations of the processes or constraints of model parameters, and to leverage recent advances in machine learning (ML) new method is needed to unlock the 'black box' to overcome its limitations due to low interpretability, out-of- sample failure and massive data demand. In this study, we developed a first of its kind knowledge-guided machine learning model for agroecosystems (KGML-ag), by incorporating biogeophysical/chemical domain knowledge from an advanced PB model, ecosys, and tested it by simulating daily N2O fluxes with real observed data from mesocosm experiments. The Gated Recurrent Unit (GRU) was used as the basis to build the model structure. To optimize the model performance, we have investigated a range of ideas, including: 1) Using initials of intermediate variables (IMVs) instead of time series as model input to reduce data demand; 2) Building hierarchical structures to explicitly estimate IMVs for further N2O prediction; 3) Using multitask learning to balance the simultaneous training on multiple variables; and 4) Pretraining with millions of synthetic data generated from ecosys and fine tuning with mesocosm observations. Six other pure ML models were developed using the same mesocosm data to serve as the benchmark for the KGML-ag model. Results show that KGML-ag did an excellent job in reproducing the mesocosm N2O fluxes (overall r2 = 0.81, and RMSE = 3.6 mg N m-2 day-1 from cross-validation). Importantly KGML-ag always outperforms the PB model and ML models in predicting N2O fluxes, especially for complex temporal dynamics and emission peaks. Besides, KGML-ag goes beyond the pure ML models by providing more interpretable predictions as well as pinpointing desired new knowledge and data to further empower the current KGML- ag. We believe the KGML-ag development in this study will stimulate a new body of research on interpretable ML for biogeochemistry and other related geoscience processes. [ABSTRACT FROM AUTHOR]

Indirect nitrous oxide (N 2 O) emissions from streams and rivers are a poorly constrained term in the global N 2 O budget. Current models of riverine N 2 O emissions place a strong focus on denitrification in groundwater and riverine environments as a dominant source of riverine N 2 O, but do not explicitly consider direct N 2 O input from terrestrial ecosystems. Here, we combine N 2 O isotope measurements and spatial stream network modeling to show that terrestrial-aquatic interactions, driven by changing hydrologic connectivity, control the sources and dynamics of riverine N 2 O in a mesoscale river network within the U.S. Corn Belt. We find that N 2 O produced from nitrification constituted a substantial fraction (i.e., >30%) of riverine N 2 O across the entire river network. The delivery of soil-produced N 2 O to streams was identified as a key mechanism for the high nitrification contribution and potentially accounted for more than 40% of the total riverine emission. This revealed large terrestrial N 2 O input implies an important climate-N 2 O feedback mechanism that may enhance riverine N 2 O emissions under a wetter and warmer climate. Inadequate representation of hydrologic connectivity in observations and modeling of riverine N 2 O emissions may result in significant underestimations.

China is the largest CO 2 emitting country on Earth. During the COVID-19 pandemic, China implemented strict government control measures on both outdoor activity and industrial production. These control measures, therefore, were expected to significantly reduce anthropogenic CO 2 emissions. However, large discrepancies still exist in the estimated anthropogenic CO 2 emission reduction rate caused by COVID-19 restrictions, with values ranging from 10% to 40% among different approaches. Here, we selected Nanchang city, located in eastern China, to examine the impact of COVID-19 on CO 2 emissions. Continuous atmospheric CO 2 and ground-level CO observations from January 1st to April 30th, 2019 to 2021 were used with the WRF-STILT atmospheric transport model and a priori emissions. And a multiplicative scaling factor and Bayesian inversion method were applied to constrain anthropogenic CO 2 emissions before, during, and after the COVID-19 pandemic. We found a 37.1-40.2% emission reduction when compared to the COVID-19 pandemic in 2020 with the same period in 2019. Carbon dioxide emissions from the power industry and manufacturing industry decreased by 54.5% and 18.9% during the pandemic period. The power industry accounted for 73.9% of total CO 2 reductions during COVID-19. Further, emissions in 2021 were 14.3-14.9% larger than in 2019, indicating that economic activity quickly recovered to pre-pandemic conditions., (Copyright © 2022 Elsevier Ltd. All rights reserved.)