Abstract. The applications of novel deep learning (DL) techniques in atmospheric science are rising quickly. Here we build a hybrid DL model (hyDL-CO), based on convolutional neural networks (CNNs) and long short-term memory (LSTM) neural networks, to provide a comparative analysis between DL and Kalman filter (KF) to predict carbon monoxide (CO) concentrations in China in 2015–2020. We find the performance of DL model is better than KF in the training period (2015–2018): the mean bias and correlation coefficients are 9.6 ppb and 0.98 over eastern China and are −12.5 ppb and 0.96 over grids with independent observations (i.e., grids with CO observations that are not used in DL training and KF assimilation). By contrast, the assimilated CO concentrations by KF exhibit comparable correlation coefficients but larger negative biases. Furthermore, the DL model demonstrates good temporal extensibility in the test period (2019–2020): the mean bias and correlation coefficients are 95.7 ppb and 0.93 over eastern China and 81.0 ppb and 0.91 over grids with independent observations, while CO observations are not fed into the DL model as an input variable. Despite these advantages, we find a weaker prediction capability of the DL model than KF in the test period, and a noticeable underestimation of CO concentrations at extreme pollution events in the DL model. This work demonstrates the advantages and disadvantages of DL models to predict atmospheric compositions with respect to traditional data assimilation, which is helpful for better applications of this novel technique in future studies.
Abstract. Satellite and surface carbon monoxide (CO) observations have been widely used to investigate the sources and variabilities of atmospheric CO. However, comparative analyses to explore the effects of satellite and surface measurements on atmospheric CO assimilations are still lacking. Here we investigate the evolution of atmospheric CO over E. Asia in 2015–2020, via assimilating CO measurements from the Measurement of Pollution in the Troposphere (MOPITT) and China Ministry of Ecology and Environment (MEE) monitoring network. We find a possible inconsistency by assimilating satellite and surface measurements: the adjusted CO columns are about 3.29, 3.63 and 3.68 x 1018 molec/cm2 by assimilating surface CO measurements, in contrast to 2.80–2.93, 3.14–3.25 and 3.09–3.22 x 1018 molec/cm2 by assimilating MOPITT CO observations in 2015–2020 over E. China, North China Plain (NCP) and Yangtze River Delta (YRD), respectively. This inconsistency could be associated with possible representation errors due to differences between urban and regional CO backgrounds. Furthermore, assimilations of normalized surface CO measurements (to mitigate the influences of representation errors) indicate declines of CO columns by about 4.0, 4.5, and 4.0 x 1016 molec/cm2/y over E. China, South Korea and Japan in 2015–2020, respectively, in contrast to 1.7–2.1, 2.1–2.7, and 2.1–2.6 x 1016 molec/cm2/y by assimilating MOPITT CO measurements. This discrepancy reflects the different vertical sensitivities of satellite and surface observations in the lower and free troposphere. This work demonstrates the importance to integrate information from satellite and surface measurements to provide a more accurate evaluation for atmospheric CO changes.
Abstract. Satellite and surface carbon monoxide (CO) observations have been widely used to investigate the sources and variabilities of atmospheric CO. However, comparative analyses to explore the effects of satellite and surface measurements on atmospheric CO assimilations are still lacking. Here we investigate the assimilated atmospheric CO over East Asia in 2015–2020, via assimilating CO measurements from the Measurement of Pollution in the Troposphere (MOPITT) instrument and Ministry of Ecology and Environment of China (MEE) monitoring network. We find noticeable inconsistencies in the assimilations: the adjusted CO columns (Xco) are about 162, 173 and 172 ppb by assimilating surface CO measurements, in contrast to 138–144, 149–155 and 144–151 ppb by assimilating MOPITT CO observations over East China, the North China Plain (NCP), and the Yangtze River Delta (YRD), respectively. These inconsistencies could be associated with possible representation errors due to differences between urban and regional CO backgrounds. Furthermore, the adjusted surface CO concentrations are about 631, 806, and 657 ppb by assimilating surface CO measurements, in contrast to 418–427, 627–639 and 500–509 ppb by assimilating MOPITT CO observations over East China, NCP, and YRD, respectively; assimilations of normalized surface CO measurements (to mitigate the influences of representation errors) indicate declines of CO columns by about 2.2, 2.1, and 1.8 ppb yr−1, in contrast to 0.63–0.86, 0.97–1.29, and 1.0–1.27 ppb yr−1 by assimilating MOPITT CO measurements over East China, South Korea, and Japan, respectively. These discrepancies reflect the different vertical sensitivities of satellite and surface observations in the lower and free troposphere. This work demonstrates the importance of integrating information from satellite and surface measurements to provide a more accurate evaluation of atmospheric CO changes.
Abstract. The applications of novel deep learning techniques in atmospheric science are rising quickly. Here we build a hybrid deep learning (DL) model (hyDL-CO), based on convolutional neural networks (CNN) and long short-term memory (LSTM) neural networks to provide a comparative analysis between DL and Kalman Filter (KF) to predict carbon monoxide (CO) concentrations in China in 2015–2020. We find the performance of DL model is better than KF in the training period (2015–2018): the mean bias and correlation coefficients are 9.6 ppb and 0.98 over E. China, and −12.5 ppb and 0.96 over grids with independent observations. By contrast, the assimilated CO concentrations by KF exhibit comparable correlation coefficients but larger negative biases. Furthermore, DL model demonstrates good temporal extensibility: the mean bias and correlation coefficients are 95.7 ppb and 0.93 over E. China, and 81.0 ppb and 0.91 over grids with independent observations in 2019–2020, while CO observations are not fed into the DL model as an input variable. Despite these advantages, our analysis indicates a noticeable underestimation of CO concentrations at extreme pollution events in the DL model. This work demonstrates the advantages and disadvantages of DL models to predict atmospheric compositions in respective to traditional data assimilation, which is helpful for better applications of this novel technique in future studies.
Abstract. Adjoint of the GEOS-Chem model has been widely used to constrain the sources of various atmospheric pollutants. Here we provide an updated version (GC-Adjoint-HEMCO) of the adjoint of GEOS-Chem model to support the MERRA-2 meteorological data and Harmonized Emissions Component (HEMCO) emission inventories. State-of-the-art inventories, such as CEDS (Community Emissions Data System), MIX, NEI2011 (National Emissions Inventory), and GFED4 (Global Fire Emission Database), are supported in GC-Adjoint-HEMCO. We find good agreements in carbon monoxide (CO) emissions from various inventories, chemical sources and sinks, and surface and column CO concentrations between GC-Adjoint-HEMCO and GEOS-Chem (v12-8-1) forward simulations. Furthermore, observing system simulation experiments (OSSE) are employed to evaluate the performance of GC-Adjoint-HEMCO in 4D variational (4D-var) assimilations. We find underestimations by approximately 15 % in the a posteriori anthropogenic CO emissions over North America and Europe due to limited coverage of observations by smoothing the pseudo-CO observations with Measurement of Pollution in the Troposphere (MOPITT) averaging kernels. As an example application of GC-Adjoint-HEMCO, we constrain anthropogenic CO emissions in 2015 by assimilating MOPITT CO observations. The a posteriori anthropogenic CO emission estimates derived in this work match well with Jiang et al. (2017) in North America and Africa but are overestimated in Asia, South America and Australia, which could be associated with the different treatment of MOPITT CO observations over ocean grids and the large differences in CO chemical sources and sinks. The updated model developed in this work is a useful extension of the adjoint of GEOS-Chem model.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.