NH₃ Emissions and Lifetime Estimated by Satellite Observations with Differential Evolution Algorithm

Abstract

As an important irritant trace gas in the atmosphere, ammonia (NH₃) significantly impacts human health and environment. Bottom-up emission inventories are widely used to estimate ammonia emissions and their geographical distributions over China. However, high uncertainties are still associated with emission inventories due to inaccurate emission factors used. The Differential Evolution (DE) algorithm is a population-based stochastic optimization algorithm used to solve complicated optimization problems. We quantify NH₃ emissions and lifetime from Infrared Atmospheric Sounding Interferometer (IASI) NH₃ observations together with MERRA-2 wind fields based on the DE algorithm. Two inland cities, Urumchi and Golmud in China, are chosen to study of the NH₃ emissions based on the distributions of NH₃ total columns and wind fields. The NH₃ emissions rate estimated is about 5.84 × 10⁻¹¹ and 4.19 × 10⁻¹¹ kg·m⁻²s⁻¹ in Urumchi and in the Golmud area from May to September from 2008 to 2023, respectively. The lifetime of NH₃ estimated in the two areas is 4.31 and 9.19 h, respectively. We compare the NH₃ emissions and lifetime estimated in this study with the values in other studies, and the results show the reliability of the method used. This work is one of few quantitative studies of NH₃ emissions from cities using satellite observations in China.

1. Introduction

Ammonia is a strong irritant gas and can cause respiratory problems when inhaled in high concentrations [1]. An increase in ammonia concentration in the atmosphere have negative effects on human health [2]. It also can cause skin and eye irritation at high concentrations [3]. In addition, atmospheric ammonia can react with other acidifying pollutants to form fine particulate matter, which can lead to respiratory and cardiovascular diseases [4]. Also, long-term exposure to high levels of ammonia can harm the environment, including plants and soil fertility [5]. So, it is important to monitor and control ammonia emissions in order to maintain a healthy and safe environment.

Sources of atmospheric ammonia are agriculture (animal waste, fertilizer application), industry (production of chemicals, manufacturing processes), biomass burning, vehicle exhaust, soils, natural vegetation, wild animals, and oceans. Sinks of atmospheric ammonia include dry deposition, wet removal by precipitation, and conversion to particulate ammonium salts via reaction with acids. Based on the activity data on fertilizer application, livestock farming, fertilizer production, and populations with emission factors of various ammonia sources, bottom-up emission inventories are developed to estimate China’s atmospheric ammonia emissions and its geographical distributions. However, there is large uncertainty in emission estimation due to the inaccurate emission factors used.

Satellite remote sensing has been widely used in estimating gas emissions. With the help of satellite sensors, we can observe the concentration distributions of various gases in the atmosphere, such as sulfur dioxide (SO₂), nitrogen dioxide (NO₂), Ozone (O₃), carbon dioxide (CO₂), carbon monoxide (CO), methane (CH₄), and ammonia (NH₃), etc. [6,7,8,9,10,11,12,13]. Based on the spatial and temporal distribution of gases measured by satellite, emissions can be estimated from various sources, such as industrial plants, vehicles, and agricultural activities [14]. Satellite remote sensing provides a valuable tool for monitoring emissions of pollution gases and greenhouse gases, which is critical for pollution control and mitigating climate change [15,16].

Satellite observations can resolve the emission sources with a higher resolution than the bottom-up inventory methods [17]. Detecting emissions from point source shows a significant advantage for short-lived gases, such as SO₂, NO₂, and NH₃, as the emissions of the point sources can be derived with a simple mass balance method [18,19,20]. Beirle et al. used NO₂ tropospheric columns observed by Ozone Monitoring Instrument (OMI) satellite, wind information, and the downwind decay to derive the emissions from megacities and continuously emitting point sources [21]. Liu et al. used the method proposed by Beirle et al. [21] to estimate the emissions of cities and NOx lifetime by the variations of plume with wind speed [22,23,24]. Zhang et al. used SO₂ columns observed by OMI and SO₂ profiles derived from the MOZART-4 model to evaluate ground-level SO₂ concentrations over China [25]. Kourtidis et al. developed a box model that calculates SO₂ emissions using relationships between gridded OMI satellite data of NO₂ and SO₂ with calm wind [26]. The anthropogenic emissions of NO₂ and SO₂ in China simulated by modeling were compared with the values derived from OMI data and ground level observations [27].

In addition, Zhang et al. used a GEOS-Chem joint model and Tropospheric Emission Spectrometer (TES) satellite NH₃ columns to estimate anthropogenic NH₃ emissions in China, at the 0.5° × 0.67° horizontal resolution, and the calculated annual NH₃ emissions were 22.06 Tg in 2008 [28]. Van Damme et al. used a decade of daily observations from Infrared Atmospheric Sounding Interferometer (IASI) satellite, to produce a high spatial resolution map of global distribution of NH₃, and identify the world’s industrial and agricultural ammonia point sources, and the annual growth of NH₃ columns over China were about (24.7 ± 2.1) × 10¹³ molec cm⁻² yr⁻¹ from 2008 to 2018 [29]. Clarisse et al. identified the type of global NH₃ point sources by a wind-adjusted super resolution method using IASI satellite data [30]. Clarisse et al. found some atmospheric ammonia emanations by an oversampling method using IASI satellite-derived atmospheric NH₃ data [31]. Dammers et al. used IASI and Cross-track Infrared Sounder (CrIS) satellite observations to identify large NH₃ point sources and estimate their emissions, and the estimated ammonia emissions over China are about 1556 kt yr⁻¹ [32]. Li et al. improved fertilizer-related NH₃ emission inventories and derived the atmospheric NH₃ emissions in mainland China about 12.11 Tg in 2016 [33].

Previous studies have shown that ammonia has a lifetime of about a few hours to a few days [13]. Van Damme et al. used a box model to identify, categorize and quantify the world’s ammonia emission hotspots based on IASI satellite observations, and the average effective lifetime of ammonia is smaller than 12 h [13]. Dammers et al. used a wind rotation approach, the exponentially modified Gaussian (EMG) method and CrIS and IASI(A and B) satellite data to estimate NH₃ emissions and lifetimes from large agricultural and industrial point sources, and the computed NH₃ lifetime for large point sources is on average 2.35 ± 1.16 h [32]. Hauglustaine et al. used a global model to summarize the global budget of ammonia and ammonium, and the corresponding simulation lifetime of ammonia in the atmosphere was 15 h in 2000 [34]. Shephard et al. used a 2D exponentially modified Gaussian (EMG) plume model and CrIS satellite data from April to September from 2013 to 2017 to estimate the lifetime of ammonia, and the estimated lifetime was 2.66 h [35]. Abeed et al. used IASI satellite observations, ECMWF and MERRA-2 reanalysis data products, and a GEOS-Chem model to estimate agricultural ammonia volatilization over Europe, and the lifetime is estimated to be about 6 h in Europe [36]. Evangeliou et al. used IASI and CrIS satellite data, a chemistry transport model, and an Inverse distance weighting (IDW) interpolation method to calculate the global ammonia emissions during the period 2008–2017, and the calculated ammonia averaged lifetime is about 11 h [37]. Whitburn et al. used IASI satellite data in the years 2008–2015 and a simple box model to estimate the NH₃ emissions over Indonesia, and the derived lifetime of NH₃ is between 12 and 36 h [38].

Due to the limitations of the spatial and temporal resolution of satellite observations, it is difficult to accurately estimate emissions from small-scale sources, especially for those non-isolated sources. However, satellite observations can resolve isolated emission sources (no other large sources within 100~200 km radius) and estimate emissions from these sources accurately.

The purpose of this study is to quantify NH₃ emissions from several inland cities located in China and estimate the lifetime of atmospheric NH₃ based on IASI satellite observations and wind information. Section 2 introduces the data and methods used in this work. Section 3 gives the results and discussion. The NH₃ emissions and the lifetime are calculated for two cities during “ozone season” (May–September) from 2008 to 2023. Section 4 makes the conclusions.

2. Materials and Methods

2.1. IASI Data

IASI is a satellite payload designed for atmospheric sounding and climate monitoring. It is operated by the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) onboard the Metop series of polar-orbiting satellites. The MetOp-A, MetOp-B, and MetOp-C satellite of the series were launched in 2006, 2012, and 2018, respectively.

IASI measures the infrared radiation emitted by the Earth’s atmosphere in the spectral range of 645–2760 cm⁻¹ with a high spectral resolution of 0.25 cm⁻¹. By analyzing the atmospheric radiation, IASI observations can be used to retrieve vertical profiles of temperature and humidity, as well as the concentration of various atmospheric trace gases, including carbon monoxide, ozone, methane, and sulfur dioxide. IASI provides global coverage with a revisit time of less than 12 h, making it a valuable tool for weather forecasting, air quality monitoring, and climate research. IASI measurements have been used to study the trend and variability of atmospheric composition, to evaluate the performance of climate models, and to investigate the impact of human activities on the Earth’s atmosphere [39,40,41,42].

The main source of ammonia emissions comes from the large-scale use of ammonia fertilizers in agriculture [43]. Agriculture is considered the dominant source of atmospheric ammonia and contributes over 81% of global NH₃ emissions [29]. “Ozone season” (May to September) is both the high-ozone season and the growing season of animals and plants, when ammonia concentrations are high [44]. NH₃ total column data of MetOp-A and MetOp-B over China during “ozone season” (May to September) from 2008 to 2023 are used in this work. The IASI NH₃ total columns (v4.0.0, ULB-LATMOS) measured in “cloud free” (cloud fraction below 30%) days are considered. For inter-annual comparisons and trends analyses, this work also uses the reanalyzed daily IASI ammonia (NH₃) total column dataset.

2.2. MERRA-2 Wind Field

Modern-Era Retrospective analysis for Research and Applications (MERRA-2) is a NASA atmospheric reanalysis product [45] which provides long-term, global atmospheric data from 1980 to the present. It is based on the Goddard Earth Observing System Model, Version 5 (GEOS-5), and assimilates a wide range of observations, including satellite, ground-based, and aircraft data.

MERRA-2 includes a wide range of atmospheric parameters, such as temperature, humidity, wind, pressure, precipitation, and aerosol concentrations. It provides hourly, daily, and monthly data at a horizontal resolution of 0.5 degree latitude by 0.625 degree longitude, making it a valuable resource for climate research, weather forecasting, and air quality studies.

In this study, the M2I1NXASM dataset in MERRA-2 was selected with a time resolution of 1 h, downloaded from https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/ (accessed on 20 February 2024). The U50M and V50M data representing the wind direction are used and linearly interpolated to a grid of 0.01° × 0.01° in this work.

2.3. Emission Estimation

Beirle et al. used wind field information to classify NO₂ data by different wind downwards and proposed the “line density” method [21]. This method can effectively avoid the neutralization effect of the opposite wind direction data and help to obtain the dispersion pattern of pollutants clearly. Ammonia, like nitrogen dioxide, is a short-lived gas, so ammonia emissions can be estimated using this method. The “line density” concept was proposed to simplify the two-dimensional problem of the spatial distribution of NH₃ concentration to a one-dimensional problem. Integrating the NH₃ columns across the respective main wind direction, the two-dimensional mean NH₃ columns distribution is reduced to one-dimensional “line densities”.

A simple model function M(x) as a function of the distance x was used to fit the observed NH₃ line densities.

$$M(x)=E\times (e\otimes G)(x)+B$$

(1)

where E is total NH₃ emissions, B is a constant background of NH₃, and e(x) is a truncated exponential function:

$$\{\begin{array}{cc}e(x)=\exp (-(x-X)/x_0)& x\ge X\\ e(x)=0& x<X\end{array}$$

(2)

where X is the location of the emission source relative to a city center, and x₀ is the e-folding distance downwind. G(x) is a Gaussian function with standard deviation $\sigma$ to describe the interference of different factors on the ideal distribution e(x).

$$G(x)=\frac{1}{\sqrt{2\pi }\sigma }\exp \left(-\frac{x^2}{2{\sigma }^2}\right)$$

(3)

The parameters, including total emissions E, e-folding distance x₀, constant background B, location of the emission source X, and standard deviation $\sigma$ of G(x), are estimated by the non-linear least-square fit or other algorithms. Least-square fit easily gets stuck in a local minimum, so we used the Differential Evolution optimization algorithm to estimate the parameters in this study.

2.4. Differential Evolution Optimization Algorithm

Differential Evolution (DE) is a stochastic, population-based optimization algorithm belonged to the family of Evolutionary Algorithms. DE was first proposed by Storn and Price [46], and it has been widely used in engineering design, machine learning, and financial modeling to solve some complicated optimization problems, as it is an efficient parallel search algorithm. DE is a population-based stochastic search technique that relies on mutation, crossover, and selection operators at each generation to evolve a collection of candidate solutions toward an optimal state. In this study, NH₃ total columns are calculated separately for calm and four different wind direction sectors, for a calm threshold of 2 m/s. For each wind direction sector, line densities are calculated by spatial integration in across-wind direction. So, the mean column distributions (two-dimensional, 2D) are reduced to one-dimensional “line densities” along the respective main wind direction. Then, we use DE algorithm to fit of a simple model function to the observed NH₃ line densities in order to determine the parameters (E, x₀, B, X, σ) by iteratively adjusting the five parameters to make the calculated line density close to the observed line density. The parameters include NH₃ emissions, and the lifetime (τ = x₀/w) of NH₃ is derived from the mean wind w and e-folding distance x₀, where w is the projected mean wind speed on the main wind direction. The basic idea behind DE is to maintain a population of candidate solutions and evolve them iteratively by applying three main operations: mutation, crossover, and selection. At each iteration, a new population is created by applying the mutation operation to the current population. The ith individual vector ${\stackrel{\rightharpoonup }{X}}_{i,G}$ of the population in generation G has D components.

$${\stackrel{\rightharpoonup }{X}}_{i,G}=\left[x_{1,i,G},x_{2,i,G},x_{3,i,G},\cdots \cdots ,x_{D,i,G}\right]$$

(4)

This involves generating a mutant vector ${\stackrel{\rightharpoonup }{V}}_{i,G}$ by adding a scaled difference vector between two randomly chosen population vectors to a third population vector.

$${\stackrel{\rightharpoonup }{V}}_{i,G}={\stackrel{\rightharpoonup }{X}}_{r_1^i,G}+F({\stackrel{\rightharpoonup }{X}}_{r_2^i,G}-{\stackrel{\rightharpoonup }{X}}_{r_3^i,G})$$

(5)

The crossover operation is used to combine the ${\stackrel{\rightharpoonup }{V}}_{i,G}$ with a randomly selected population vector. Then, produce a trial vector ${\stackrel{\rightharpoonup }{U}}_{i,G}$.

$${\stackrel{\rightharpoonup }{U}}_{i,G}=\left[u_{1,i,G},u_{2,i,G},u_{3,i,G},\cdots \cdots ,u_{D,i,G}\right]$$

(6)

$$u_{j,i,G}=\{\begin{array}{cc}{v}_{j,i,G}& \begin{array}{ccc}if(ran{d}_{i,j}[0,1]\le Cr& or& j=j_{rand})\end{array}\\ x_{j,i,G}& otherwise\end{array}$$

(7)

Finally, the selection operation is used to choose between the ${\stackrel{\rightharpoonup }{U}}_{i,G}$ and the corresponding original vector for the next generation.

$${\stackrel{\rightharpoonup }{X}}_{i,G+1}=\{\begin{array}{cc}{\stackrel{\rightharpoonup }{U}}_{i,G}& if\begin{array}{cc}f({\stackrel{\rightharpoonup }{U}}_{i,G})\le f({\stackrel{\rightharpoonup }{X}}_{i,G})& \end{array}\\ {\stackrel{\rightharpoonup }{X}}_{i,G}& otherwise\end{array}$$

(8)

Differential evolution (DE) has some advantages over least squares fitting (LSF). Firstly, DE is more robust than LSF in finding the global optimum solution. LSF gets stuck in a local minimum, while DE can explore the search space more effectively and find the global minimum. Secondly, DE does not require gradient information, which LSF needs. This makes DE suitable for optimization problems, where the objective function is noisy, discontinuous, or has many local minima. Lastly, DE is a simple algorithm that is easy to implement and only requires a few tuning parameters. In contrast, LSF requires careful selection of model parameters and regularization terms. Therefore, this study uses DE instead of least squares fitting to estimate the parameters of M(x).

2.5. Selection of Study Area

Urumchi is the capital city of the Xinjiang Uygur Autonomous Region in the northwest of China. It is the largest city in Xinjiang region, and one of the most remote cities away from seas in the world [47]. It had a population of 4 million in 2022. At the end of 2022, there were 700 thousand grass-fed animals in Urumchi [48]. Urumchi is situated in a basin surrounded by the Tianshan Mountains to the north and south, and the city proper is located in the Dabancheng District, which lies in the eastern part of the basin. The city is at an elevation of approximately 800 m above sea level, and has a continental arid climate. The city is also situated along the ancient Silk Road trade route and serves as a hub for transportation and commerce in the region.

Golmud is a city located in the Haixi Mongol and Tibetan Autonomous Prefecture, in the province of Qinghai, China [49]. Golmud had a population of 0.14 million in 2022. At the end of 2022, there were 340 thousand grass-fed animals in Golmud [50]. It is situated in the northeastern part of the Qinghai-Tibet Plateau, at an elevation of 2809 m above sea level. The city is located near the Qilian Mountains to the north and the Kunlun Mountains to the south, surrounded by vast expanses of desert and grassland. The city has a dry and arid climate, with long and cold winter and a short and warm summer. It is an important transportation hub, connecting the western and eastern part of China, and also well known for its rich mineral resources, including salt, potassium, magnesium, and boron.

The city of Urumchi and Golmud are chosen for study for the following reasons: Firstly, NH₃ total columns in Urumchi and Golmud are higher than those in the surrounding areas. Secondly, there are no large NH₃ sources within 100 km radius of the two cities, so the interference from other sources is minimized. Figure 1 displays the locations of the two cities and the distributions of NH₃ total column over China based on the dataset from the study of Van Damme et al. [13]. This dataset is taken from ANNI-NH3-v2.1R-I data, which are the oversampled data of IASI NH₃ total columns for 9 years (2008–2016). The blue circles are the locations of Urumchi and Golmud.

3. Results and Discussion

The monthly averages of NH₃ total column in Urumchi and Golmud during from 2008 to 2023 were calculated based on IASI data, as shown in Figure 2. It can be seen that most NH₃ total columns in Urumchi were higher than those in Golmud. So the annual averages of NH₃ total columns in Urumchi are higher than the corresponding values in Golmud in Figure 3. Also, the annual averages of NH₃ total columns in Urumchi show high levels in 2013, 2016, and 2021, with the value of 1.8 × 10¹⁶, 1.7 × 10¹⁶, and 1.8 × 10¹⁶ molec. cm⁻², respectively. However, the variation of annual average of NH₃ columns in Golmud shows a different pattern, with a slowly increasing trend and no peaks in different years.

Monthly averages of NH₃ total columns throughout the observation period from 2008 to 2023 in Urumchi and Golmud were also calculated, as shown in Figure 4. NH₃ total columns show obvious seasonal variation in Urumchi, with high concentrations from March to October and low concentrations in winter, such as in January, February, November, and December. Monthly averages of NH₃ total columns are about 2.0 × 10¹⁶ molec. cm⁻² from March to October, and about 4.7 × 10¹⁵ molec. cm⁻² in winter in Urumchi. NH₃ total columns also show clear seasonal variation in Golmud, with high concentrations in summer and low concentrations in winter. The monthly average of NH₃ total columns reaches its maximum in July in Golmud, with a value of 1.0 × 10¹⁶ molec. cm⁻².

Since ammonia is mainly concentrated near the surface, each IASI observation is linked to wind data from the surface to the height of 50 m taken from MERRA-2. We applied the DE method to determine the NH₃ emissions and lifetime over Golmud and Urumchi city. Figure 5 plots the wind roses from MERRA-2 wind field data in Urumchi during “ozone season” (May–September) from 2008 to 2023. The main wind direction is west in Urumchi during the study period, so the variation of NH₃ distributions under west wind and calm conditions is used to estimate emissions.

The NH₃ emissions were estimated from the variation of NH₃ distributions under windy and calm conditions. Figure 6 shows the distributions of the mean NH₃ columns around Urumchi under calm and west wind conditions during “ozone season” (May–September) from 2008 to 2023. The outflow plume of NH₃ in Urumchi clearly displays the distribution patterns of NH₃ under different wind directions.

Figure 7 shows NH₃ line densities under calm wind (red) and west wind (green) around Urumchi, and abscissa is the distance x to the center of Urumchi city. The maximum of the line density (red) appears in the center of the city under calm winds, and 8 km east of the city center under west wind conditions (green). The shift of the maximum of the two line densities under the two conditions reflects the spread range of the NH₃ emission plume.

The fitted parameters by the DE algorithm are emissions E, background B, the e-folding distance x₀, the fitted sources located in a polluted background relative to the Urumchi center X, and the standard deviation σ of the Gaussian function to describe the interference of different factors on the ideal distribution e(x). Figure 8 displays the fitted line density using the model function M(x) by the DE algorithm for the Urumchi area. The green line is the observed line density of ammonia under main wind direction, and the blue line is the fitted line density of ammonia.

NH₃ emissions can be determined from the NH₃ columns downwind decay. The lifetime (τ = x₀/w) of NH₃ is derived from the mean wind w and e-folding distance x₀. For Urumchi, the NH₃ emission estimated is 5.84 × 10⁻¹¹ kg·m⁻²s⁻¹ and the lifetime of NH₃ is computed to be 4.31 h during “ozone season” (May–September) from 2008 to 2023.

Figure 9 plots the wind roses from MERRA-2 wind field data in Golmud during “ozone season” (May–September) from 2008 to 2023. The main wind direction is also west in Golmud during the study period, so the variation of NH₃ distributions under west wind and calm conditions is used to estimate emissions from Golmud area.

Figure 10 displays the mean NH₃ columns around Golmud under calm and west wind conditions during “ozone season” (May–September) from 2008 to 2023. The outflow plume of NH₃ in Golmud clearly displays the different distributions of NH₃ columns under different wind directions. The different distributions of NH₃ columns reflect the different plume dispersion regions under the two wind conditions.

Figure 11 shows NH₃ line densities under west wind and calm wind around Golmud (Figure 11). The maximum of line density (red) appears in the center of the city under calm winds, and 7 km east of the city center under west wind conditions (green). The downwind shift of the maximum of line density under west wind also reflects the spread range of the emission plume. Figure 12 displays the fitted line densities using the model function M(x) by the DE algorithm for Golmud area, which is used for estimation of NH₃ emissions. The green line is the observed line density of ammonia under the main wind direction, and the blue line is the fitted line density of ammonia. The NH₃ emission is 4.19 × 10⁻¹¹ kg·m⁻²s⁻¹ and the lifetime of NH₃ is computed to be 9.19 h during “ozone season” (May–September) from 2008 to 2023 in Golmud.

The NH₃ emissions in Golmud is slightly lower than that in Urumchi, as the emission sources are different in the two different areas. Firstly, the intensity of agricultural activities in the two cities is different, as the farming area and animal numbers fed in Urumqi are larger than those in Golmud. Secondly, more inhabitants in Urumqi means more anthropogenic emissions, such as those from transport, biomass burning, waste burning, and industry. These factors lead to high emission intensity from Urumqi city. Furthermore, the different lifetime of NH₃ estimated in the two areas reflects the totally different sources and sinks. A shorter atmospheric lifetime of NH₃ in Urumqi is likely caused by a higher emission of acidifying compounds (mostly SO₂ and NO_x) in the atmosphere. Moreover, more precipitation on average in Urumqi means more wet deposition in this area. These factors may explain the difference in ammonia lifetime in the two cities. The accuracy of emissions and lifetimes estimated by this work are highly depended on the accuracy of MERRA-2 wind directions and the satellite observations. There are potential sources of uncertainty contributing to the estimation results. First is the uncertainty of total columns of ammonia from satellite data due to retrieval methods and hardware. Second is fit errors due to calculation of the line density. Third is from wind fields, as wind information is important during fit for determination of the fitted parameters.

Further, in order to validate our results, we compared our results with those in other studies.

Table 1. NH₃ emissions estimated compared with emission inventories.

Method	Source	Urumchi	Golmud
REASv3.2.1 [51]	anthropogenic sources	2.83 × 10−10 kg·m−2s−1	1.59 × 10−10 kg·m−2s−1
EDGAR V6.1 [52]	anthropogenic sources	5.43 × 10−11 kg·m−2s−1	4.48 × 10−12 kg·m−2s−1
MEIC-PKU-NH3 [53]	anthropogenic sources	5.06 × 10−11 kg·m−2s−1	2.22 × 10−12 kg·m−2s−1
This study	anthropogenic sources and natural sources	5.84 × 10−11 kg·m−2s−1	4.19 × 10−11 kg·m−2s−1

lists the comparison of our results with the emission values reported in some emission inventories. Anthropogenic sources in the REAS emission inventory include power plant, domestic, fertilizer, manure management, industry, road transport, other transport, and miscellaneous sources. Anthropogenic sources in the EDGAR emission inventory include agriculture, power, industry, residential, transportation, and other sources. Anthropogenic sources in the MEIC-PKU-NH₃ emission inventory include power plant, industry, residential, transportation, agriculture, and wildfire. In this study, anthropogenic sources in the two cities include agriculture, power, industry, residential, transportation, and other sources, while the natural sources may include soils and plants. Table 1 shows that the NH₃ emission of Urumchi derived in this study agrees well with the values from the EDGAR and MEIC-PKU emission inventory, but is less than the value from the REAS emission inventory. While the NH₃ emission of Golmud derived in this study is lower than the value from the REAS emission inventory, and higher than the values from the EDGAR and MEIC-PKU emission inventory, the emission estimated in our study is close to the values reported in these emission inventories. The comparison results prove the reliability of the emission estimation method used in this study.

Moreover,

Table 2. NH₃ lifetime estimated compared with those reported in publications.

Reference	Data	Method	Lifetime
Van Damme et al., 2018 [13]	IASI	Box model	Smaller than 12 h
Dammers et al., 2019 [32]	CrIS, IASI	Wind rotation and EMG method	∼2.5 h
(Hauglustaine et al., 2014 [34]	AeroCom emission data set	LMDz-INCA global model	∼15 h
Shephard et al., 2020 [35]	CrIS	2-D EMG method	∼2.66 h
Abeed et al., 2023 [36]	IASI	GEOS-Chem model	∼6 h
Evangeliou et al., 2021 [37]	IASI, CrIS	Inverse distance weighting	∼11.6 h
Whitburn et al., 2016 [38]	IASI	Box model	between 12 and 36h
This study	IASI	Line density model	4.31 h and 9.19 h

displays the comparison of the estimated lifetime of NH₃ with the values reported in publications. The lifetime of NH₃ reported in other studies is in the range between 2.5 and 36 h, and the values estimated in Urumchi and Golmud of 4.31 and 9.19 h fall into this range, which also proves the reliability of the estimation method used in this study.

4. Conclusions

Ammonia is a short-lived trace gas and satellite data are sparse, which makes it difficult to estimate NH₃ emissions from urban sources. In this study, we used the DE optimization algorithm to estimate the ammonia emissions and lifetime over two inland cities in China, based on the IASI satellite data and wind fields for the period (May–September) from 2008 to 2023.

The NH₃ emission rate estimated is 5.84 × 10⁻¹¹ and 4.19 × 10⁻¹¹ kg·m⁻²s⁻¹ in the Urumchi and Golmud areas, respectively. The lifetime of NH₃ is 4.31 h and 9.19 h, respectively, in these two areas. The different emissions and different lifetime of NH₃ estimated in the two cities reflects the distinct sources and sinks in the two areas. Firstly, the intensity of agricultural activities related to manure management and fertilizer application in the two cities is different. Secondly, different numbers of inhabitants means different anthropogenic emissions from transport, biomass burning, waste burning, and industry. Thirdly, different levels of acidifying compounds (mostly SO₂ and NO_x) in the atmosphere mean different conversion rate to particulate ammonium salts via reaction with acids. Lastly, different weather conditions lead to different dry deposition and wet removal. All these factors result in the different emissions and lifetime of ammonia in the two cities. We compared the NH₃ emissions and lifetime estimated in this study with the values in other studies, and the results show the reliability of the method used. This work is one of few quantitative studies of NH₃ emissions and lifetime from cities in China using satellite observations. Our study provides a valuable method for quantification of urban ammonia emissions, helping to control ammonia emissions from cities and evaluate the effectiveness of control measures, especially for cities with high concentrations of ammonia.

In summary, satellite observations offer a valuable insight into ammonia emissions and atmospheric processes. While satellite observations can resolve isolated urban sources and estimate the emissions from cities, it is difficult for satellite data to distinguish anthropogenic emissions from natural emissions. This problem can be resolved by conventional emission inventories. Combining satellite observations with ground-based measurements and refining modeling based on emission inventories can contribute to a more comprehensive understanding of ammonia dynamics in the atmosphere.