Satellite detection and model verification of NOx emissions from power plants in Northern China

We evaluate the recently increasing tropospheric NO2 columns in Northern China measured by the Ozone Monitoring Instrument (OMI) with an advanced power-plant NOx emission inventory and the NASA INTEX-B emission inventory, using a global chemical transport model (GEOS-Chem). In areas with newly built power plants the modeled and OMI-retrieved summertime average tropospheric NO2 columns increased by 55% and 47%, respectively, between 2005 and 2007. A monthly average increase of 1.79 Gg NOx emissions is calculated to lead to an increase of 1.0 × 1015 molecules cm − 2 in the modeled NO2 columns in the study areas. Good consistency (R2 = 0.61, slope = 1.18, n = 14) between the increased modeled and OMI-retrieved summertime average NO2 columns is found. These results suggest that NOx emissions from large power plants in Northern China can be identified and quantified using OMI retrievals with confidence. The NASA INTEX-B emission inventory appears to underestimate the NOx emissions from the industry and transportation sectors, making it more difficult to quantify power-plant emissions when they are co-located with large cities.


Introduction
Emissions of nitrogen oxides (NO x = NO 2 + NO) have been increasing dramatically during the past decade in China (Richter et al 2005, Zhang et al 2007, due to the rapidly growing economy powered by the generation of energy from fossil fuels. Since 2005 hundreds of large new electricity generating units have been constructed in China, and the total 5 Author to whom any correspondence should be addressed. capacity of coal-fired power plants has increased by 42%, from 391 GW in 2005 to 556 GW in 2007. Under increasing pressure from acid deposition and summer photochemical smog, which are related to tropospheric NO 2 , NO x emission reduction has become an urgent target of legislation and government policies in China. The magnitude and distribution of NO x emissions can change dramatically even in a couple of years in China, as a result of the booming construction activities as well as the application of emission control measures in various sectors. The Ozone Monitoring Instrument (OMI) onboard NASA's Aura satellite platform has proven to be an effective tool to detect NO 2 in the atmosphere at high temporal and spatial resolution and thus to investigate and improve NO x emissions , Kim et al 2009, Lin et al 2010. Recently, Zhang et al (2009a) reported increased OMIretrieved summertime tropospheric NO 2 columns in response to power-plant construction that had occurred between 2005 and 2007 in Inner Mongolia, China, showing the potential for large point-source identification from the OMI retrievals. Focused on the same area, Li et al (2010) recently analyzed OMI PBL SO 2 columns to identify the reduction of SO 2 emissions from the same power plants, due to the installation and effective operation of flue-gas desulfurization (FGD) systems. Both of these analyses compared the OMI column retrievals with emissions. The drawback to this approach is that the relationship between emissions and columns can be compromised by pollutant inflow and outflow from the region, as well as by other factors like meteorology and chemical transformation. Although Zhang et al (2009a) and Li et al (2010) found distinct signals of increased NO 2 or SO 2 columns associated with the newly built power plants, they were not able to give quantitative relationships between emissions and satellite retrievals, which would need a chemical transport model to account for the changes in chemistry, transport, and meteorology. In this letter we improve on the previous analytical methodologies by using the NO x pointsource emission estimates to calculate NO 2 columns with a chemical transport model and then comparing modeled and measured columns.
To investigate the change of NO x emissions in Northern China and evaluate the impact of newly built power plants on local NO 2 concentrations, tropospheric NO 2 columns are simulated with GEOS-Chem, a global chemical transport model (CTM), for the summers (June, July, and August) of 2005 and 2007. The modeled columns are then compared with OMI-retrieved tropospheric NO 2 columns. This study is the first in which the modeled NO 2 columns from individual point sources are evaluated by comparison with satellite measurements in China.

Anthropogenic NO x emissions
The 2006 NASA INTEX-B emission inventory (Zhang et al 2009b) is adopted as a baseline emission inventory for East Asia in GEOS-Chem. This inventory has been previously validated by several model simulation studies against satellite measurements (Zhang et al 2008) and in situ observations from the INTEX-B field campaign (van Donkelaar et al 2008). However, due to a lack of specific information at that time about point sources in China, the INTEX-B inventory only includes individual electricity generating units 300 MW; smaller units are treated as area sources and distributed by socioeconomic activity data at the province level. For this work we have acquired a new, more detailed power-plant NO x emission inventory, which is based on a Chinese power sector database for 31 provinces supported by CMEP (Chinese Ministry of Environmental Protection). This database contains more than 5700 generating units of all sizes, with detailed information about boiler type, technology, geographical location, coal consumption per unit electricity supply, and the month the unit came into operation.
Annual NO x emissions are calculated by unit according to the technology and operation information, following the equation of Zhao et al (2008): where i , j , k, m stand for province, generator unit, boiler size, and emission control technology; 1.4 is the mass scaling factor from standard coal to raw coal; E is the annual NO x emissions (Mg); U is the unit size (MW); T is the annual operation hours for unit j in province i ; F is the specific coal consumption per unit electricity supply (gce kWh −1 ); and EF is the emission factor. NO x emission factors for different types of units vary between 5.6 g kg −1 coal burned and 10.5 g kg −1 based on boiler size and the presence or absence of low-NO x burners (LNB) (Zhang et al 2007).  The current version of the GEOS-Chem model supports multiple-layer inputs for anthropogenic NO x emissions, which is critical for high-stack emitters such as power plants and large industrial sources. For power plants, we input the NO x emitted by generator units 100 MW into the second vertical layer (mid-altitude at ∼205 m) and by units >100 MW into the third vertical layer (mid-altitude at ∼335 m). For all other anthropogenic sources NO x emissions are input as surface emissions in the first level. The detailed description of default NO x emissions from soil, lightning, biomass burning, and aircraft adopted in the model can be found in Martin et al (2002). The chemical time step in the model is 30 min. A monthly varying tropopause height is used to derive the tropospheric NO 2 columns. More than six months presimulation is conducted to obtain the initial concentration fields for each year. Finally, daily modeled tropospheric NO 2 columns are averaged at the local time of 13:00-15:00. To obtain average modeled NO 2 columns coincident with the satellite measurements, modeled columns are re-gridded to 0.25 • × 0.25 • horizontal resolution and sampled by following the same daily satellite scenes used in the average OMI columns.

Satellite measurements
The Dutch-Finnish OMI on board NASA's EOS Aura satellite is a nadir-viewing imaging spectrograph measuring direct and atmosphere-backscattered sunlight in the ultraviolet-visible range from 264 to 504 nm (Levelt et al 2006). The EOS Aura spacecraft was launched on July 15, 2004, and circulates in a 98.2 • inclination, sun-synchronous polar orbit at 705 km altitude, with local overpass time at ∼13:45. As one benefit of the two two-dimensional CCD detectors and the wide field of view (114 • ), OMI measures the complete spectrum with a fine ground footprint (13 × 24 km 2 at nadir) along track and daily global coverage. For this study, we choose the OMI standard NO 2 Level-2G product (version 003) at NASA Goddard Earth Sciences Data and Information Services Center. We exclude scenes where the cloud fractions are greater than 30%, as the NASA science team did. The scenes with cloud-top pressures higher than 800 hPa in cloudy scenes are also rejected, because low clouds can give rise to inaccuracy if they intersect the NO 2 air mass (Mijling et al 2009). Scenes on the edges of each swath and anomalous rows reported by NASA are also filtered out. For the summertime average, we always have one-third of the grids left with usable data after application of the filter to individual OMI daily retrievals. Finally, the NO 2 columns are averaged at 0.25 • ×0.25 • horizontal resolution for the summers of 2005 and 2007.

Results and discussion
Three regions in Northern China are selected for study, referred to as East, Central, and West, which contain 42 newly built electricity generating units, with a combined capacity of We assemble all of these new power-plant units into fifteen sub-regions and categorize them into three groups, as shown in table 1: city with newly built power plants (designated as Group A), city without newly built power plants (Group B), and rural area with newly built power plants (Group C), following Zhang et al (2009a) and Li et al (2010). The size of each sub-region is 0.5 • × 0.5 • , which is close to the original model resolution and large enough to allow for significant NO 2 transport/transformation, considering that NO 2 has a short lifetime of about 3.6 h in summer (Schaub et al 2007). Table 1 also shows the NO x emissions with fractions of power-plant emissions and NO 2 columns from GEOS-Chem and OMI. While the spatial distributions of NO 2 columns in figures 1 and 2 are broadly similar in their features, the model significantly underestimates the absolute NO 2 concentrations, by 46% on average, compared to the satellite measurements. This is especially obvious in the larger cities such as Taiyuan, Zhangjiakou, Shuozhou, and Datong. This kind of underestimation has been found in previous studies (Wang et al 2007, Zhao and Wang 2009, Lin et al 2010. There are several possible reasons for the underestimation. One possibility is that the INTEX-B emission inventory underestimates or omits NO x emissions from smaller sources in the industry and transportation sectors. This hypothesis is supported by three aspects of our work. First, NO x emissions from the power sector scarcely changed between 2005 and 2007 in cities such as Zhangjiakou, Shuozhou, and Datong (see table 1), and in these areas GEOS-Chem underestimates the absolute NO 2 columns from OMI by almost 75%, significantly lower than the average. Second, the increase rates of satellite measurements are higher than those from the model in Baotou, Wuhai, and Taiyuan, implying that we might underestimate the increase of emissions. Third, the fractions of power-plant NO x emissions in total NO x emissions shown in table 1 for most Group A cities remained largely unchanged between 2005 and 2007, while those for Group B cities declined less than anticipated. Another factor could be that soil emissions adopted in GEOS-Chem could be underestimated by a factor of two in summer (Jaeglé et al 2005), which would cause bias in the background NO 2 level. Finally, it has been suggested that the NASA OMIretrieved products overestimate NO 2 columns by 67%-74% in summer (Lamsal et al 2010), which might be derived in part from the annual average NO 2 vertical profiles used in the calculation of the air mass factor (AMF). The dense aerosol loading, especially in cities, could also lead to above-average contributions to errors in the NO 2 column amounts .
Nevertheless, despite the differences in absolute NO 2 columns between model and satellite, the changes between 2005 and 2007 in both measures are consistent. Modeled summertime average NO 2 columns increased dramatically in areas with newly built power plants: by an average of 40% in Group A cities and 63% in Group C rural areas. The OMI-retrieved NO 2 columns increased by 46% and 47%, respectively. Due to interference from the industry and transportation sectors, we do not expect all study areas to behave perfectly in the comparison. We focus on areas where NO x emissions from power plants are a large fraction of total emissions. The increase rates (see figure 3) are very consistent in areas with newly built power plants, except for Shangdu, Zhuozi, and Liangcheng, because there are no NO x emissions from power plants in 2005 in these three areas. The increase rates in Group B cities are substantially lower than the modeled ones except Shuozhou. This is expected to occur, considering the biases of emissions from the industry and transportation sectors. For all fifteen areas, the R-squared value (R 2 ) of the increase rates of modeled and OMI-retrieved NO 2 columns is 0.65, with a slope of 0.67. In contrast to the sharp increase of NO x emissions found by Zhang et al (2009a), the increase rates of modeled columns are more comparable with those of OMIretrieved columns in rural areas, which provides confidence in our approach.
The correlations of absolute changes between summertime average modeled NO 2 columns ( NO 2 -CTM), OMI-retrieved NO 2 columns ( NO 2 -OMI), and monthly average NO x emissions ( NO x ) are shown in figure 4. Good correlation is found between NO x and NO 2 -CTM (R 2 = 0.77), with a slope of 1.79, which indicates that increased NO x emissions of 1.79 Gg/month will lead to increased local NO 2 columns of 1.0 × 10 15 molecules cm −2 in the summertime in these areas. NO 2 -OMI and NO 2 -CTM are also well correlated (R 2 = 0.61) if we eliminate Taiyuan, a very large capital city with many small industrial sources that are likely omitted from or underestimated in the INTEX-B inventory. The correlation between NO 2 -CTM and NO 2 -OMI (R 2 = 0.61, n = 14) in this work is of the same order of magnitude as the correlation between NO x and NO 2 -OMI (R 2 = 0.68, n = 9) in Zhang et al (2009a); however, comparisons in rural areas are much improved in this work, as compared with Zhang et al (2009a), which means that we have a better representation of regional NO x and the emissions from smaller sources through the use of the GEOS-Chem model. Because the estimated NO x emissions in these areas are mainly contributed by newly built power plants, the emissions from which are well known, the consistency in NO 2 -OMI, NO 2 -CTM, and NO x means that the satellite measurements can be used to infer NO x emissions from large point sources. Further study should be focused on the impact of newly added NO x emissions from power plants on local NO 2 vertical profiles and aerosol loadings, which could play critical roles in the calculation of AMF used to retrieve satellite measurements.
We conclude that rapid changes in NO x emissions from large power plants in Northern China can be identified and quantified with confidence, as a result of the consistency between modeled and satellite-observed tropospheric NO 2 columns.
The NASA INTEX-B emission inventory may underestimate NO x emissions from the industry and transportation sectors, inhibiting the reliable simulation of NO 2 concentrations in large cities. This suggests that NO x emissions from small sources need to be re-estimated for Northern China. Our intention now is to apply this kind of analysis for a large number of point sources throughout China to examine and accurately quantify the NO x emissions in China.