Atmospheric Chemistry & Modeling





Tropospheric NO2 columns retrieved from satellite instruments are useful to infer NOx pollution, NOx emissions and atmospheric chemistry. Current satellite products are subject to limitations in assumptions of aerosol optical effects, surface reflectance anisotropy, vertical profiles of NO2, and cloud optical properties.

Here we develop an improved Peking University Ozone Monitoring Instrument NO2 product (POMINO) for China and surrounding areas. As of 2018/09/20, there are two versions available: POMINO v1 and POMINO v2.

Registered users as of 2019/05/01:

Registered Users


POMINO v1 (Lin et al., 2014; Lin et al., 2015):

This is the original "POMINO" algorithm.

POMINO v1 explicitly accounts for aerosol optical effects, angular dependence of surface reflectance, and dynamically varying atmospheric profiles of air pressure, air temperature and NO2 at a high horizontal resolution (25-50 km). The daily AOD data are simulated by nested GEOS-Chem and further constrained by MODIS (C5.1) data on a monthly basis. The daily BRDF data are from MCD43C2 (C5.1).

Prior to the NO2 retrieval, we retrieve cloud top pressure and cloud fraction using consistent assumptions about the states of the atmosphere and surface.

For our NO2 and cloud retrievals, we adopt from KNMI (via the SCDs of tropospheric NO2 (DOMINO v2) and O2-O2 dimer (OMCLDO2 v1.1.1.3), the TOA reflectance, and some other ancillary information.

We develop the AMFv6 code for air mass factor calculation, based on the radiative transfer model LIDORT v3.6. The AMFv6 code improves upon the code developed by Paul Palmer, Randall Martin et al., with various aforementioned capability added/extended to accommodate the calculation here. With AMFv6, radiative transfer is calculated explicitly for each satellite pixel with no need to use a look-up table. The calculation of AMFv6 is parallelized and is sufficiently fast so that one day of retrieval with global coverage would only take about three hours using 16 CPU cores.

POMINO v2 (Liu et al., 2019):

On top of POMINO v1, POMINO v2 further constrains the vertical profile of aerosol extinction by monthly climatology from CALIOP, uses the SCD data from QA4ECV, and updates to MODIS (C6) merged AOD and MCD43C2 (C6) daily BRDF.


The POMINO v1 product is consistent with MAX-DOAS NO2 data in China, with a R^2 of 0.96 as compared to the value at 0.72 for DOMINO v2. The improved consistency is related to explicit pixel-by-pixel radiative transfer calculation (instead of using a look-up table), consistent treatments of all parameters in retrieving clouds and NO2, explicit consideration of aerosol optical effects (rather than adjusting ‘effective’ clouds to implicitly account for aerosols), and consideration of surface reflectance anisotropy.

The POMINO v1 product is able to capture the high pollution situations (e.g., high aerosol and NO2 concentrations), in addition to the modest and low population situations.

The POMINO v2 product further reduces the bias against MAX-DOAS data, while maintaining the high correlation.

AMFv6 Code

Our AMFv6 code is available for public use. Currently, AMFv6 also allows users to (some of them may need further customization):

  1. Turn on/off explicit treatment of aerosol optical effects, or revise aerosol info using measurements
  2. Choose from a variety of surface reflectance data (currently OMI, MODIS black-sky, MODIS blue-sky, MODIS BRDF)
  3. Turn on/off dynamic atmosphere (e.g., time-varying air temperature, pressure profile, etc.)
  4. Retrieve clouds online, or read cloud data from a third party
  5. Read surface pressure measurement instead of using the one from model met field
  6. Output AMFv6 results in either ascii, binary or netcdf format
  7. Produce/use a look-up table (if desired in some cases)

POMINO v1 product

This is the original "POMINO" product.

As of 2018/09/20, data are available from 2004 through 2016. For newer data, see our POMINO v2 product below.


POMINO animation of monthly mean NO2 VCD maps (0.25 x 0.25 degree): 2004/10-2016/12

POMINO v1 animation

Level-3 data

POMINO Monthly or Daily Level-3 Data Download

Both daily and monthly Level-3 NO2 tropospheric VCD products are on a 0.25 x 0.25 degree grid, spatially aggregated from the Level-2 data.

Included in the Level-3 data are tropospheric NO2 AMF, tropospheric NO2 VCD, AOD at 550 nm, SSA at 550 nm, and other ancillary parameters.

The file "readme_POMINO_level3.txt" in the link above provides an introduction, including example reading programs in IDL and Fortran.

Level-2 data

POMINO Level-2 Data Download

Included are pixel-specific NO2 tropospheric VCD product and ancillary data.

Each tar.gz file contains a month worth of data files. Each data file contains Level-2 data for tropospheric NO2 AMF, tropospheric NO2 VCD, AOD at 550 nm, SSA at 550 nm, and other ancillary parameters.

The file "readme_POMINO_level2.txt" in the link above provides an introduction, including example reading programs in IDL and Fortran.


POMINO v2 product

This product is added on 2018/09/20.

As of 2019/05/13, data are available from 2004 through 2018.


POMINO v2 animation of monthly mean NO2 VCD maps (0.25 x 0.25 degree): 2004/10-2017/12

Level-3 data

POMINO v2 Monthly or Daily Level-3 Data Download

Both daily and monthly Level-3 NO2 tropospheric VCD products are on a 0.25 x 0.25 degree grid, spatially aggregated from the Level-2 data.

Included in the Level-3 data are tropospheric NO2 AMF, tropospheric NO2 VCD, AOD at 550 nm, SSA at 550 nm, and other ancillary parameters.

See user guide for brief documentation of the variables included (NO2 VCD, AMF, AOD, SSA, etc.), as well as how to read the Level-3 data.

Level-2 data

POMINO v2 Level-2 Data Download

See user guide for brief documentation of the variables included (NO2 VCD, AMF, AOD, SSA, etc.).

Examples to read the Level-2 data in IDL and Fortran are provided in the link above.



Lin, J.-T. *, R. V. Martin, K. F. Boersma, M. Sneep, P. Stammes, R. Spurr, P. Wang, M. Van Roozendael, K. Clémer, and H. Irie: Retrieving tropospheric nitrogen dioxide from the Ozone Monitoring Instrument: Effects of aerosols, surface reflectance anisotropy, and vertical profile of nitrogen dioxide, Atmos. Chem. Phys., 14, 1441-1461, doi:10.5194/acp-14-1441-2014, 2014 (PDF)

Lin, J.-T. *, Liu, M.-Y., Xin, J.-Y., Boersma, K. F., Spurr, R., Martin, R., and Zhang, Q.: Influence of aerosols and surface reflectance on satellite NO2 retrieval: seasonal and spatial characteristics and implications for NOx emission constraints, Atmospheric Chemistry and Physics, 15, 11217-11241, doi:10.5194/acp-15-11217-2015, 2015 (PDF) (Supplement)

Liu, M.-Y., Lin, J.-T. * , Boersma, K. F. *, Pinardi, G., Wang, Y., Chimot, J., Wagner, T., Xie, P., Eskes, H., Van Roozendael, M., Hendrick, F., Wang, P., Wang, T., Yan, Y.-Y., Chen, L.-L., and Ni, R.-J.: Improved aerosol correction for OMI tropospheric NO2 retrieval over East Asia: constraint from CALIOP aerosol vertical profile, Atmospheric Measurement Techniques, 12, 1-21, doi:10.5194/amt-12-1-2019, 2019 (PDF)


PKUCPL: A Two-Way Coupler to integrating models

We develop a PKUCPL (PeKing University CouPLer) coupler to integrating multiple models in a manner of two-way coupling, i.e., allowing for feedbacks between models.

The idea of developing PKUCPL originated from the fact that current global simulations tend to overestimate the global tropospheric oxidative capacity.

Global chemical transport models, widely used for studying global air pollution and transport, are limited by coarse horizontal resolutions, not allowing for detailed representation of small-scale nonlinear processes over the pollutant source regions. Traditional one-way nested regional models take global model outputs as lateral boundary conditions without feeding back to the global model, and thus do not affect the simulated global atmospheric environment. Global simulations at coarse resolutions typically overestimate the tropospheric oxidative capacity, i.e., with an overestimate in OH and ozone concentrations, and an underestimate in CO concentration, MCF lifetime and CH4 lifetime.

We develop and use PKUCPL to integrate, in a manner of two-way coupling, the global GEOS-Chem CTM (at ~2 degree resolution) and its multiple nested models (at 0.5 or 0.25 degree resolutions) covering Asia, North America and Europe. As an example, see the coupling regions and the coupling flowchart here for global + three nested models.

Effects of two-way coupling between global and nested models of GEOS-Chem

Under the two-way coupling framework, PKUCPL takes global model results as lateral boundary conditions of nested models, and at the same time takes nested model results, conduct a regridding procedure, and replaces the global model results within the nested model domains. This allows for feedback between global and nested models. Thus, the two-way coupling allows smaller-scale nonlinear processes in the nested domains (which are normally major pollution source regions) to be better represented and be able to affect the large-scale atmospheric chemistry across the globe.

As a result, the two-way coupling improves the simulations of both regional chemistry and global transport. In particular, the two-way coupling significantly alleviates the overestimate of the tropospheric oxidative capacity in the global simulation, with large reductions in OH and ozone concentrations and enhancements in CO concentration, MCF lifetime and CH4 lifetime (Yan et al., 2014; Yan et al., 2016).

In Yan et al. (2014):

1. We use CO as a tracer to diagnose the consequence of the two-way coupling (Difference between a two-way coupled and a 'pure' global simulation: An animation from 2008.07.01 to 2008.08.15). Compared to a pure global model, the two-way coupled simulation increases the global tropospheric mean CO concentration in 2009 by 10.4%, with a greater enhancement at 13.3% in the Northern Hemisphere. Correspondingly, the global tropospheric mean hydroxyl radical (OH) is reduced by 4.2%, resulting in a 4.2% enhancement in the methyl chloroform lifetime. (See the global budget of tropospheric OH for 2009 here.) The resulting CO and OH contents and MCF lifetime are closer to observation-based estimates.

2. Various factors differentiate the two-way coupled model from the global model, including the small-scale variability of NOx, CO and VOC, the resolution- and meteorology-dependent natural emissions, and other nonlinear small-scale processes. See the percentage contributions of individual factors to the difference in January 2009 tropospheric CO between the two-way coupled model and the global model.

3. Comparisons with the tropospheric CO measurements over the Pacific Ocean during the HIPPO campaigns (Flight tracks and times) in various seasons between 2009 and 2011 show significant improvements by the two-way coupled simulation on the magnitude and spatiotemporal variability of CO. In particular, the two-way coupled simulation captures the measured vertical profiles of CO, with a mean bias of 1.1 ppb (1.4%) below 9 km compared to the bias at -7.2 ppb (-9.2%) for the global model alone. See the CO time-height distribution across five HIPPO campaigns and the CO vertical profiles in five individual HIPPO campaigns.

4. The two-way coupling also affects simulations of other species. See our preliminary results for 2009: Global tropospheric hydrophobic BC, Global tropospheric hydrophilic BC, Global tropospheric OH, Global tropospheric O3.

In Yan et al. (2016):

We focus on the year of 2009. The simulated global tropospheric mean OH concentration is reduced by 5% and the MCF and CH4 lifetimes are increased by 5%. The global tropospheric ozone mass is reduced by 10%, and the global CO mass is increased by 10%. All these changes help reduce positive/negative biases in the simulation of these variables.

PKUCPL Code applied to GEOS-Chem

PKUCPL has been adopted in GEOS-Chem stanadard model since version 10.

Users can choose to couple the global model with any numbers of nested models, with a straightforward setup.

Users can choose various resolutions for global (e.g., 2.5 long. x 2 lon., 5 long. x 4 lat.) or regional (e.g., 0.667 long. x 0.5 lat., 0.3125 long. x 0.25 lat.) models.

The computational complexity of two-way coupling is minimized by the PKUCPL coupler. Users can learn how to set up a two-way coupled simulation quickly (i.e., within 1-2 hours), following a simple manual. Individual global/nested models can run at different nodes, as desired.

The computational time of the coupled system is determined by and is comparable to that of the slowest individual model. Our test suggests that only 2% or so additional run time is needed beyond the slowest individual model (the nested model for North America).

For more details, see this wiki page.

PKUCPL for integrating other models

PKUCPL is available for public use. It is flexible and can be modified and applied to other types of models (e.g., between chemistry models and climate/meteorological models or between atmospheric and oceanic models of various complexities). Please contact us for collaborations.


Yan, Y.-Y., Lin, J.-T. *, Kuang, Y., Yang, D.-W., and Zhang, L.: Tropospheric carbon monoxide over the Pacific during HIPPO: Two-way coupled simulation of GEOS-Chem and its multiple nested models, Atmospheric Chemistry and Physics, 14, 12649-12663, doi:10.5194/acp-14-12649-2014, 2014 (PDF)

Yan, Y.-Y., Lin, J.-T. *, Chen, J., and Hu, L.: Improved simulation of tropospheric ozone by a global-multi-regional two-way coupling model system, Atmospheric Chemistry and Physics, 16, 2381-2400, doi:10.5194/acp-16-2381-2016, 2016 (PDF)



Why are visibility data useful for aerosol inference?

Multi-decadal aerosol data are necessary to understand how aerosols affect climate and climate changes on the regional and global scales. However, high-quality satellite- or ground-based aerosol measurements are not available for such long-term studies. Surface visibility measurements at thousands of stations worldwide can provide useful information for long-term aerosol inference, complementing satellite- and ground-based measurements. Here we combine visibility measurements and a chemical transport model simulation to derive a new gridded AOD dataset, by converting near surface station-specific visibility data to gridded AOD data.

Method to converting from station-specific visibility data to gridded AOD data

Step 1: Near surface Aerosol Extinction Coefficient (AEC) is calculated from a quality-controlled 3-hourly visibility measurement in the absence of precipitation and fogs.

Step 2: A temporally and spatially coincident AOD to AEC ratio modeled by GEOS-Chem is used to convert near surface AEC to column AOD. In this way, the knowledge of model aerosol profile is involved, instead of assuming a uniform exponential vertical distribution as in many preivous studies.

Step 3: The visibility converted and GEOS-Chem simulated AOD to produce a new “merged” AOD dataset on a longitude-latitude grid (current resolution is 0.667º long. × 0.5º lat.). For each day, we find for a given grid cell all stations within a 2º radius of the grid cell center, calculate the ratios of visibility converted over GEOS-Chem AOD, and then use the median value of the ratios to scale the modeled AOD at the grid cell.

Product and validation

The visibility-model merged gridded AOD data are obtained. The new data preserve the spatial distribution of model AOD while using the visibility measurements to correct for the model bias.

The merged AOD dataset is highly consistent with AOD data from MODIS, AERONET, CARSNET and CSHNET, with a low bias (< 0.05 over East China) and high spatial and temporal (diurnal, seasonal and interannual) correlation.


Monthly data are currently available for 2004/10 through 2013/04 over East China (101.25ºE–126.25ºE, 19ºN–46ºN), and are free for non-commercial use.

Data resolution: 0.667 longitude x 0.5 latitude degree.

Animation of monthly mean visibility-model merged AOD maps: 2004/10-2013/04



Data Download

Visibility-model merged monthly mean AOD data download. Please see the "readme" file inside for how to read the binary data.

Adjusted visibility-model merged monthly mean AOD data download. For comparison with monthly mean MODIS/Aqua AOD, we also provide monthly mean data adjusted based on the difference in days with valid MODIS data versus with valid visibiilty data. See our paper below for details. Please see the "readme" file inside for how to read the binary data.


Lin, J.-T. *, and Li, J.: Spatio-temporal variability of aerosols over East China inferred by merged visibility-GEOS-Chem aerosol optical depth, Atmospheric Environment, 132, 111-122, doi:10.1016/j.atmosenv.2016.02.037, 2016 (PDF)

Lin, J.-T. *, van Donkelaar, A., Xin, J., Che, H., and Wang, Y.: Clear-sky aerosol optical depth over East China estimated from visibility measurements and chemical transport modeling, Atmospheric Environment, 95, 257-267, doi:10.1016/j.atmosenv.2014.06.044, 2014 (PDF)




Brief Introduction

Many variables of interest, such as air pollutants and meteorological parameters, often exhibit complex spatial and temporal variabilities. In particular, these variables contain many temporal scales that are non-periodic and non-stationary, challenging proper quantitative characterization and visualization.

The EOF-EEMD analysis-visualization package we complied aims to evaluate the spatiotemporal variability across scales, which can be periodic/stationary or not. As shown in the figure below, the package consists, in order, of an EOF analysis (Lorenz, 1956), an EEMD analysis (Wu et al., 2009), a Hilbert Transform (HT) with Marginal Spectrum Analysis (MSA), and a visualization step to quantitatively depict the spatial-temporal scales of measurement or model data.

EOF-EEMD flowchart
Figure. The flow chart of EOF-EEMD analysis-visualization package. The red boxes represent quantities visualized. Source: Liu et al. (2018).

Download EOF-EEMD

We will post the code online soon. For now, please contact us.



Liu, M.-Y., Lin, J.-T. *, Wang, Y.-C., Sun, Y., Zheng, B., Shao, J., Chen, L.-L., Zheng, Y., Chen, J., Fu, M., Yan, Y.-Y., Zhang, Q., and Wu, Z.: Spatiotemporal variability of NO2 and PM2.5 over Eastern China: observational and model analyses with a novel statistical method, Atmospheric Chemistry and Physics, 18, 1–20, doi:10.5194/acp-18-1-2018, 2018 (PDF) (Supplement)

Wu, Z., Huang, N. E., and Chen, X.: The Multi-Dimensional Ensemble Empirical Mode Decomposition Method, Adv. Adapt. Data Anal., 1, 339–372,, 2009

Lorenz, E. N.: Empirical Orthogonal Functions and Statistical Weather Prediction, Dep. Meteorol. MIT, 1(Statistical Forecasting Project;Scientific Report No. 1), 49, 1956