Introduction

In situ coal fire causes all types of pollution, including air, water, and land (Gielisch & Kropp, 2018). In situ coal fire emits many harmful gases (oxides and dioxides of nitrogen, carbon, and sulfur), which causes multiple health problems like lung and skin diseases (Gielisch & Kropp, 2018). Furthermore, coal fire is one of the major contributors to anthropogenic greenhouse gas emissions (Crow et al., 2019; Gielisch & Kropp, 2018; Howarth, 2014; Karavalakis et al., 2016; Thompson et al., 2011; Venkatesh et al., 2011). Coal fires lead to global climate change (Munawer, 2018) due to heat-trapping by greenhouse gases in the tropospheric region (Astrup et al., 2009; Herndon, 2018; Kweku et al., 2018; Lashof & Ahuja, 1990). The uncontrolled burning of coal seams is a natural hazard worldwide (Wessling et al., 2008). Many researchers have revealed that the burning of fossil fuels is one of the significant threats to the environment and economy of a country (Braadbaart et al., 2012; Ding et al., 2013; Gurney et al., 2009; Heede, 2014; Jaeglé et al., 2005; Landry & Matthews, 2016). The primary reason for in situ coal fires is the self-combustion of coal due to illegal mining activities and abandoned coal mines. Coal burning at the subsurface level takes place with a variable oxygen supply. If the in situ coal is exposed to oxygen, spontaneous combustion of coal occurs, leading to many greenhouse gas emissions (Ozdeniz et al., 2014). The top ten greenhouse gas emitters are China, the USA, European Union, India, Russia, Japan, Brazil, Indonesia, Iran, and South Korea (Mengpin & Johanne, 2020). Thus, the gas emission characteristics are also different from the complete combustion in which the reaction takes place with a fixed oxygen supply. It is very difficult for researchers to estimate the amount of greenhouse gas emissions from in situ coal burning due to an invariable oxygen supply. The chances of complete combustion of in situ coal are less until the coal is exposed to the atmosphere or oxygen. In case of insufficient oxygen supply to coal, incomplete combustion of coal takes place. During the incomplete combustion of carbon and other elements are oxidized to form oxides of carbon, sulfur, etc. The combustion reactions that take place during in situ coal burning are as follows:

Combustion of coal in the presence of sufficient air

$$1\ gm\ C\ (Coal)+2.67\ gm\ {O}_{2}=3.67\ gm\ C{O}_{2}+\ Heat$$
(1)

Combustion of coal in the presence of insufficient air

$$1\ gm\ C\ (Coal)+{1.33\ gm\ O}_{2}=2.33\ gm\ CO+Heat$$
(2)

Combustion of sulfur

$$1\ gm\ S+{1\ gm\ O}_{2}=2\ gm\ {SO}_{2}+Heat$$
(3)

Coal burning emits harmful gases and emits heavy metals into the environment (UCS, 2008). The monitoring of greenhouse gas emissions is a challenging task. There are three ways (eddy covariance, measuring the temporal change of gas stocks, and measuring the concentration of atmospheric gases) to estimate the fluxes of greenhouse gases (Bréon & Ciais, 2010). Despite the vast in situ monitoring network, finer-scale estimation of global gas fluxes in the atmosphere is challenging. Also, in situ monitoring is not possible over areas such as large water bodies and dense forests due to the region’s inaccessibility. These limitations can be overcome by adopting the spaceborne remote sensing technique to monitor greenhouse gas (Chevallier et al., 2007; Palmer, 2008). The remote sensing-based approach offers uniform spatial (Liu et al., 2016; Miyazaki et al., 2017) and temporal (Lamsal et al., 2011; Richter et al., 2005) coverage in monitoring of these gases. Over the past few years, space-based observatory data have been effectively used to monitor and assess emission inventories in industrial regions by measuring various gaseous pollutants (Borrell et al., 2003; Hoff & Christopher, 2009; Martin, 2008; Palmer, 2008). The satellite instrument measures the columnar density of the gases based on the radiation absorption at a typical wavelength.

In the last few decades, many satellite sensors to monitor different gases. The European Space Agency (ESA) launched the satellite ERS-2 with a Global Ozone Monitoring Experiment (GOME) to measure O3 and NO2 (Burrows et al., 1997). Thereafter, ESA launched the ENVISAT satellite with the Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY) sensor to measure a wide range of pollution species (Bovensmann et al., 1999), GOME-2 (Callies et al., 2000), and Infrared Atmospheric Sounding Interferometer (IASI) (Clerbaux et al., 2009) to measure CO, NH3, and CH4 gases. National Aeronautics and Space Administration (NASA), United States of America (USA), launched an Earth Observing System (EOS) satellite - TERRA with three sensors viz. Moderate Resolution Imaging Spectroradiometer (MODIS) (Barnes et al., 1998), Multiangle Imaging SpectroRadiometer (MISR) (Diner et al. 1998), and Measurements of Pollution in the Troposphere (MOPITT) (Drummond et al., 1995) to measure the aerosol and its optical properties and CO. Thereafter, AQUA was launched by NASA with another sensor to monitor the earth’s atmosphere (Lambrigtsen et al., 2004). Furthermore, NASA launched a satellite with two Aura sensors (Ozone Monitoring Instrument (OMI) and Tropospheric Emission Spectrometer (TES)) to understand the changing chemistry of the earth’s atmosphere (Schoeberl et al., 2006), measuring the trace gases like O3, NO2, SO2, HCHO, Bromine monoxide (BrO), chlorine dioxide (OClO) (Levelt et al., 2006), and mapping of global 3-D distribution of tropospheric ozone (Beer et al., 2001). Japan Aerospace Exploration Agency (JAXA) launched the Greenhouse gases Observing Satellite (GOSAT) to study the transport mechanism of CO2 and CH4 (Hamazaki et al., 2005; Kuze et al., 2006). ESA launched Sentinel 5-Precursor (S-5P) satellite (Veefkind et al., 2012) with TROPOspheric Monitoring Instrument (TROPOMI) for monitoring air quality, climate, and the ozone layer. The S-5P sensor can measure key atmospheric constituents, including CO, SO2, NO2, CH2O, and CH4. Due to satellite data’s uniform spatial and temporal resolution, the present study demonstrates the application of remote sensing-based approaches to estimate the source strength of greenhouse gases in a coal fire region. The study used TROPOMI (TROPOspheric Monitoring Instrument) onboard satellite sensor data of Sentinel-5P for estimating the spatial and temporal profiles of three greenhouse gases (CO, SO2, and NO2) in Jharia coalfield (JCF).

According to the Greenpeace India report, Jharia is the most polluted city in India and is a critical area of high pollution due to coal fires. Therefore, the monitoring and assessment of greenhouse gases and their source strength identification are required for controlling the emissions from coal fires. Additionally, the columnar density of three gases (CO, SO2, and NO2) in the coal fire region of JCF was compared with the same in the nearby rural, urban, and forest regions for understanding the source strength of various pollutants.

Materials and method

Study area

JCF (Fig. 1) is one of India’s major coalfields and primary coking coal suppliers. It is situated in the state of Jharkhand with geographical coverage of 258 km2 and extends from 23°38′ N to 23°49′ N latitude and 86°8′ E to 86°30′ E longitude. JCF is situated in two districts (Dhanbad and Bokaro) of Jharkhand, and thus, the entire region of both districts is selected for a comparative study of pollutant levels according to different land use types.

Fig. 1
figure 1

Location of the study area

Full size image

Data used

The study used TROPOMI satellite sensor data of Sentinel-5P. TROPOMI sensor measures ultraviolet earthshine radiances at a higher spectrum. The first Copernicus mission satellite launched in October 2017, exclusively used to monitor atmospheric gases. A wide range of gases such as nitrogen dioxide (NO2), ozone (O3), formaldehyde (HCHO), sulfur dioxide (SO2), carbon monoxide (CO), and methane (CH4) can be monitored using a TROPOMI sensor. The data are provided at different time intervals depending on the level of processing like near real time (NRT), offline, and reprocessing. The present study used TROPOMI offline processed data of CO, SO2, and NO2. Vidot et al. proposed a Shortwave Infrared Carbon Monoxide Retrieval (SICOR) algorithm to estimate the CO from the S-5P sensor (Vidot et al., 2012). The retrieval of NO2 was proposed by (van Geffen et al., 2019), which is based on the DOMINO (Dutch OMI NO2) and QA4ECV (Quality Assurance for Essential Climate Variables) processing systems. Theys et al. (2017) have developed the retrieval of SO2 from the TROPOMI sensor. The study also used the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) Reanalysis 1 dataset of wind speed and direction for dispersion analysis. The region’s daily wind speed and direction data are available online at NOAA Physical Science Laboratory (PSL) of the USA. The CO, NO2, and SO2 concentrations in microgram per cubic meter along with metrological parameters like temperature (°C), relative humidity (%), solar radiation (W/m2), and barometric pressure (mmHg) were used for the multiple linear regression analyses. These data were obtained from the Central Pollution Control Board website from the environmental pollution data section (Source: https://cpcb.nic.in/automatic-monitoring-data/).

Methods

The flowchart of the working methodology is shown in Fig. 2. The working methodology involves three major stages viz. satellite data processing for retrieval of monthly means of columnar density for selected gaseous pollutants for the specified duration, visualization of retrieval data for the study region, and analyzing the data to understand the emission characteristics in different region climatic conditions.

Fig. 2
figure 2

Flowchart of the working methodology

Full size image

Sentinel-5P data available in the Google Earth Engine (GEE) are used for estimating the CO, SO2, and NO2 over the Jharia coalfield. The spatial profiles of the monthly mean of three pollutants were derived for the study region for the year 2019. The spatial maps of monthly means of columnar density were generated using the code editor in GEE and finally processed in ArcGIS for visualization and analysis. Preprocessing of the satellite data includes AOI extraction and cloud masking.

The monthly means were extracted for each month of 2019. Nine specific locations (Table 1) from four land use types (coal fire mining region, urban region, rural, and forest area) were chosen for comparative analysis of the monthly means of columnar density with land use types (Fig. 3). Out of nine stations, five (S1, S2, S3, S4, and S5), located in JCF, represent the coal fires; one (S6), situated in Bokaro steel city, represent industrial or urban region; two (S7, S8), situated in the two major hills with an elevation of 893 m and 1365 m, represent the forest area; and one (S9), located in a village, represents the rural area.

Table 1 Location details of different land use types in the study area
Full size table
Fig. 3
figure 3

Land use/landcover map of the study area

Full size image

The windrose plots for the same periods were also generated using NCEP-NCAR reanalysis data for the nearest location (86.250° N, 23.809° E) of the Jharia coalfield for understanding the dispersion behaviors of these gases. The wind speed and direction data are derived from the v-wind and u-wind components. The monthly windrose diagrams are generated for the year 2019. Similarly, the monthly average precipitation data for the year 2019 was used for analyzing its influence on the emission of gases.

The multiple linear regression analyses were performed to examine the dependency of TROPOMI-based columnar density and other metrological parameters (temperature, relative humidity, wind speed, solar radiation, and barometric pressure) with the ground-level pollution concentration. In this model, the concentration of gases obtained from CPCB was considered a dependent variable and columnar density of gases along with other meteorological parameters were regarded as independent or predictor variables.

Results and discussion

The spatial maps for CO, NO2, and SO2 were derived using GEE and processed using ArcGIS. The spatial maps from January 2019 to December 2019 for each month were derived for CO, NO2, and SO2. The monthly mean precipitation level along with the windspeed for 2019 was plotted (shown in Fig. 4) to study the effects of rainfall on greenhouse gas emissions. The windrose plots of each month of the year 2019 were also prepared and shown with the respective month’s spatial map.

Fig. 4
figure 4

Monthly average precipitation and rainfall chart

Full size image

Spatio-temporal variation of columnar density of CO

The spatial distributions of the monthly mean of CO (mol/m2) are shown in Fig. 5a–l for the year 2019. It can be inferred from the monthly mean map that a higher value of CO was observed in the southeastern regions, i.e., in most of the months except during monsoon months (July, August, and September). The monthly precipitations (shown in Fig. 4) indicate high precipitation levels in these 3 months. The results indicate that columnar density during monsoon months (July (0.035–0.041), August (0.033–0.037), September (0.033–0.040), and October (0.038–0.048)) are lower in comparison to the other months ((January (0.039–0.055), February (0.035–0.052), March (0.039–0.055), April (0.044–0.057), May (0.040–0.056), June (0.043–0.052), November (0.039–0.062), December (0.039–0.060)).

Fig. 5
figure 5

CO columnar density map (a) January (b) February (c) March (d) April (e) May (f) June (g) July (h) August (i) September (j) October (k) November (l) December

Full size image

The highest value was observed during the winter months (November and December), as the gas dispersion was low due to calm atmospheric conditions. It was also observed that CO mainly concentrated over the Jharia coalfield and Bokaro Steel Plant, which indicates that coal fire and the steel plant are the major sources of CO emissions.

The monthly means of CO were also extracted for nine locations of different land use types, as shown in Fig. 6. Out of these nine points, five locations represent the mining area (S1, S2, S3, S4, and S5), one represents the urban or industrial area (S6), two represent the hilly/forest area (S7 and S8), and one represents the village area (S9). In most of the locations, CO was highest in November and December. The higher concentration in these 2 months is low dispersion, as the wind speed was relatively lower during these months (shown in Fig 4). The monthly mean of CO were found to be higher in fire areas (S1, S2, S3, S4, and S5) and steel plant (S6) than the stations located in forest/hilly areas (S7 and S8) and village area (S9). The results indicate that coal fire and the Bokaro steel plant are the two major sources of CO emission in the study region.

Fig. 6
figure 6

Monthly means of columnar density of CO in different locations

Full size image

Spatio-temporal variation of columnar density of NO2

The monthly means of the columnar density map of NO2 (mol/m2) is derived in the Google Earth engine for the year 2019, as represented in Fig. 7a–l. In most cases, the derived monthly means indicate a higher value in the mining and urban regions (Bokaro Steel city) except in September. The maximum monthly mean for NO2 was found to be 0.00015591, 0.000158976, 0.000160008, 0.000146958, 0.000122054, 0.000110367, 0.0000798774, 0.0000737129, 0.0000996758, 0.000110612, 0.000160898, and 0.000146346 respectively in January, February, March, April, May, June, July, August, September, October, November, and December. In this case, the maximum values of the monthly mean are found to be lowest during the monsoon months (July, August, and September).

Fig. 7
figure 7

NO2 columnar density map (a) January (b) February (c) March (d) April (e) May (f) June (g) July (h) August (i) September (j) October (k) November (l) December

Full size image

It can also be inferred from the spatial maps that the highest NO2 was mainly concentrated over the Bokaro steel plant and Jharia coalfield across the study area in most of the months.

The columnar density of NO2 was extracted for the nine selected locations, as shown in Fig. 8. It was observed that NO2 was highest in the Bokaro steel plant region (S6) in each month for the study period in comparison to the mining regions (S1, S2, S3, S4, and S5). The monthly means of columnar density of NO2 were found to be lowest in village and forest areas (S7, S8, and S9). Therefore, it can be presumed that most of the emissions of NO2 are from vehicular emissions in the urban area (Bokaro steel plant) and coal fires.

Fig. 8
figure 8

Monthly means of columnar density of NO2 in different locations

Full size image

Spatio-temporal variation of columnar density of SO2

The spatial distributions of the monthly mean columnar density of SO2 (mol/m2) are shown in Fig. 9a–l for 2019. These values were extracted for nine locations (S1-S9) of different land use types to analyze the distributions and sources (Fig. 10). The SO2 was observed to be low during monsoon months (July to September). In these months, the columnar densities of SO2 in rural and forest regions (S7, S8, and S9) were found to be below the detection level. The negative values indicate a clean area or having low SO2 levels (Theys et al., 2017). In most cases, the columnar density of SO2 was found to be highest over the mining and urban areas (S1, S2, S3, S4, S5, and S6), but the values did not show any specific trend. The spatial maps were not showing any hotspot of higher SO2 by which an inference on emissions source can be identified.

Fig. 9
figure 9

SO2 columnar density map (a) January (b) February (c) March (d) April (e) May (f) June (g) July (h) August (i) September (j) October (k) November (l) December

Full size image
Fig. 10
figure 10

Monthly means of columnar density of SO2 (10-5) in different locations

Full size image

Source apportionment and dispersion analyses of gaseous CO, NO2, and SO2

Though the maximum CO and NO2 concentrations were observed near the emission sources on most occasions, the locations of recorded maximum CO and NO2 concentrations varied with time. The prime reason behind this is the influence of horizontal dispersion of gaseous pollutants due to higher wind speed. The wind rose plots for each month are shown along with the spatial maps of CO, NO2, and SO2 (Figs. 5a–l, 7a–l, and 9a–l). The windrose diagrams were generated using WRPLOT software version 8.0.2. The windrose diagrams indicate that the region has significant wind speed variations and directions with the season. The wind speed in the regions has either calm (0–1 m/s) or in the range of 1–4 m/s, and thus, the horizontal dispersion of pollutants is low. In March 2019, the wind speed exceeded 4 m/s 10% of the time, and the prevailing wind direction is N-W and S-E. Thus, the gaseous pollutants from mining regions are transported towards the S-E direction. This leads to a higher columnar density of CO and NO2 in the S-E region. In June 2019, more than 10 % of wind reached a speed greater than 4 m/s, and the prevailing wind direction was from E-N-E to W-S-W and from E-S-E to W-N-W. Therefore, most of the gases are concentrated towards the west of the emission sources.

For analyzing the strength of TROPOMI-based estimations of CO, NO2, and SO2, a multiple regression analysis was done by considering respective columnar density as a dependent variable and meteorological data (wind speed, temperature, relative humidity, and barometric pressure) along with the respective pollutant concentrations (CO/NO2/SO2) as dependent variables. The daily concentrations of CO, NO2, and SO2 and meteorological data from 1 January 2019 to 31 December 2019 were obtained for Jorapokhar, Jharkhand (latitude: 23.707909, longitude: 86.414670) from the CPCB website (https://cpcb.nic.in/automatic-monitoring-data/). The results of the multiple regression analyses are presented in Table 2. The R2 values for CO, NO2, and SO2 were 0.316, 0.295, and 0.598, respectively. Though the R2 values are less in each case, the regression models are statistically significant in each case. All the models are statistically significant at a 5% significance level, as the F values are less than 0.05 in each case. The results further indicate that wind speed has a negative influence on each pollutant, as the coefficients are − 0.264, − 0.040, and − 0.018, respectively, for CO, NO2, and SO2. For CO, all the predictor variables have a negative influence except barometric pressure. Interestingly, temperature and relative humidity showed a positive influence, and the rest of the parameters showed negative influences on NO2 concentrations. For SO2, all the predictor variables showed negative influence except columnar density on observed concentration. It should be noted that all the dependent variables are statistically significant at a 5% significance level for the prediction of CO concentrations except solar radiation and columnar density. It was also observed no dependent variables are statistically significant at a 5% significance level for the prediction of NO2 concentrations except barometric pressure. For SO2, all the dependent variables are statistically significant at a 5 % significance level except wind speed. Thus, it can be inferred from the results that TROPOMI-based columnar depth can be used for tracking and estimations of SO2 only. The remote sensing-based approach for estimation of greenhouse gases emission for a small region is explained in the subsequent section.

Table 2 Multiple regression analyses results
Full size table

Estimation of greenhouse gases emission from in situ coal fire

In this study, a remote sensing-based approach is used to estimate greenhouse gas emissions from in situ coal fires. The demarcated area is represented in the coal fire map of JCF for 2019 (Fig. 11), which was obtained from Biswal and Gorai (2021). JCF has a large number of working mines, and few of them are affected by coal fires. The red areas showed the subsurface fires in 2019. In this study, a small area was considered to quantify the gas emission from in situ coal burning due to the availability of coal seam details.

Fig. 11
figure 11

Colliery map showing the coal fire pockets in 2019

Full size image

The selected area comprises six collieries (Kenuadih, Godhar, Kusunda, Alkusha Ena, and Dhansar), as indicated in Fig. 11. The reserves in this area are confined to the Barakar coal measures (Lower Permian), which contain twenty-four coal seams (Mineral Exploration Corporation Limited, 2013). The in situ coal extends to a depth of 610 m and has a total reserve of 11,728 million tonnes (Mehta et al., 1957). The total surface area coverage of six collieries (study area) was estimated to be 10.2 km2. The quality of coal in this reserve was classified as high-grade metallurgical and sub-bituminous. The specific gravity of coal was assumed to be 1.346 kg per cubic meter. Past studies indicated that coal fires were existed in seams XI and XII of JCF from 2008 to 2016 (Hamilton et al., 2009; Mohalik et al., 2016). This suggests that seams XI and XII were under fire and were not completely extinguished from 2008 to 2016. The study estimated the gas emission from 2009 to 2019 based on the coal fire maps of 2009 and 2019 (Fig. 12) (for a detailed description of the coal fire maps, refer to the manuscript of Biswal & Gorai, 2020, 2021). The average thickness values of coal seams XI and XII are 7.31 m and 3.63 m, respectively (Mohalik et al., 2016).

Fig. 12
figure 12

Coal fire map in 2009 and 2019 in AOI

Full size image

The area coverages of coal fire for 2009 and 2019 were estimated to be 1.91 km2 and 0.89 km2 respectively. Thus, it is assumed that entire coal is burned for the area of 0.276 km2 (= 1.91–0.89) in 10 years (2009 to 2019). Therefore, the volume of coal burned can be determined as follows:

$$\begin{aligned}&\text{Volume of coal burned during }2009\mathrm{\ to\ }2019 \\&\quad= \text{Surface area of coal burned}\\&\qquad\times \left(\text{thickness of XI seam}+\text{thickness of XII seam}\right)\\&\quad=276000 {\mathrm{m}}^{2}\times (7.31\mathrm{ m}+3.63\mathrm{ m})= 3019440{\mathrm{ m}}^{3}\end{aligned}$$

The weight of coal burned can be determined by multiplying the volume of coal burned to the specific gravity of coal (1.346 kg/m3), which was estimated as 4,064,166.24 kg.

Equations (2) and (3) indicate that 1 g of coal burning under complete combustion emits 2.33 g and 2 g of CO and SO2, respectively. Therefore, 4,064,166.24 kg of coal burns emits approximately 9.4695 Mt of CO and 8.128 Mt of SO2, respectively. In the same way, CO2 emission can also be estimated.

Conclusions

The study demonstrates the use of Sentinel-5P observations in deriving the spatial profiles of three gaseous pollutants (CO, NO2, and SO2) over JCF. The study selects nine locations of four land use types (mining region, steel plant city, forest/hilly area, and village area) for identifying the major emission sources. The study results show that CO was high in the fire regions as compared to the non-fire areas. However, the columnar density of NO2 was high in the Bokaro steel city as compared to other regions. Coal-mining regions exhibit a higher value for NO2 but less than the urban area. The reason is that the urban area in this location mainly comprises industries whose NO2 emissions might be more than the emissions from coal fires. This indicates that coal-mining activities are the major source of CO, whereas the vehicular and steel industry is the major source of NO2. The spatial profile of SO2 does not reveal the specific emission source, as there was no specific trend observed with different land use types. The point data at nine locations indicates that SO2 was low in the rural area as compared to other land use types. Thus, the TROPOMI data can efficiently identify the major emission sources for air quality management in a region. The greenhouse gas emission results indicated that approximately 9.4695 Mt of CO and 8.128 Mt of SO2 has been emitted in 10 years (2009 to 2019) only from six collieries based on the assumptions that entire coal is burned with complete combustion.