Environ Sci Pollut Res DOI 10.1007/s11356-014-3561-9
RESEARCH ARTICLE
Time series model prediction and trend variability of aerosol optical depth over coal mines in India Kirti Soni & Kulwinder Singh Parmar & Sangeeta Kapoor
Received: 27 February 2014 / Accepted: 3 September 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract A study of the assessment and management of air quality was carried out at 11 coal mines in India. Long-term observations (about 13 years, March 2000–December 2012) and modeling of aerosol loading over coal mines in India are analyzed in the present study. In this respect, the Box-Jenkins popular autoregressive integrated moving average (ARIMA) model was applied to simulate the monthly mean Terra Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD550 nm) over 11 sites in the coal mines region. The ARIMA model was found as the most suitable model with least normalized Bayesian information criterion (BIC) and root mean square error and high value of R2. Estimation was done with the Ljung-Box test. Finally, a forecast for a 3-year period from January 2013 to December 2015 was calculated which showed that the model forecasted values are following the observed trend quite well over all mining areas in India. The average values of AOD for the next 3 years (2013–2015) at all sites are found to be 0.575±0.13 (Raniganj), 0.452±0.12 (Jharia), 0.339±0.13 (Bokaro), 0.280 ± 0.09 (Bishrampur), 0.353 ± 0.13 (Korba), 0.308 ± 0.08 (Talcher), 0.370 ± 0.11 (Wardha), 0.35 ± 0.10 (Adilabad), 0.325±0.09 (Warangal), 0.467±0.09 (Godavari Valley), and 0.236±0.07 (Cuddapah), respectively. In addition, long-term lowest monthly mean AOD550 values are observed over
Bishrampur followed by Cuddapah, Talcher, Warangal, Adilabad, Korba, Wardha, Godavari Valley, Jharia, and Raniganj. Raniganj and Jharia exhibit the highest AOD values due to opencast mines and extensive mining activities as well as a large number of coal fires. Similarly, the highest AOD values are observed during the monsoon season among all four seasons over all the mining sites. Raniganj exhibits the highest AOD value at all seasons and at all sites. In contrast, the lowest seasonal AOD values are observed during the postmonsoon season over Raniganj, Talcher, Wardha, Adilabad, Warangal, and Godavari Valley. Similarly, over Jharia, Bokaro, Bishrampur, Korba, and Cuddapah, the lowest AOD values are found in the winter season. Increasing trends in AOD550 have been observed over Raniganj, Bokaro, Bishrampur, Korba, Talcher, and Wardha as well as over Adilabad and Godavari Valley, which is in agreement with previous works. Negative or decreasing AOD trend is found only over Jharia, Warangal, and Cuddapah without being statistically significant. Seasonal trends in AODs have also been studied in the present paper.
Responsible editor: Philippe Garrigues
Introduction
K. Soni (*) CSIR-National Physical Laboratory, Delhi, India e-mail:
[email protected] K. S. Parmar Department of Mathematics, University School of Basic and Applied Sciences, G. G. Singh Indraprastha University, Dwarka 110075, Delhi, India S. Kapoor Laxmi Narayan College of Technology & Science (LNCTS), Bhopal, MP, India
Keywords ARIMA . Trends . Aerosols . Aerosol optical depth . Coal mines . India
India is the world’s third largest coal consuming nation after China and the USA. Coal mining is one of the core industries in India that plays a significant role in the economic development of the country (Chaulya and Chakrabarty 1995; Kumar 1996); however, it degrades the environment. Maximum mining accomplishments contribute directly or indirectly to air pollution (Kumar et al. 1994; CMRI 1998). Sources of air pollution in the coal mining regions normally come from drilling, blasting, coal loading and unloading, coal handling
Environ Sci Pollut Res
plants, road transport, and losses from exposed overburden dumps, overburden loading and unloading exposed pit faces, and workshops (CMRI 1998). These air pollutants deteriorate air quality and ultimately affect people, vegetation, and wildlife in and nearby mining regions (Chaudhari and Gajghate 2000; Crabbe et al. 2000; Wheeler et al. 2000; Nanda and Tiwary 2001). India has abundant domestic reserves of coal. Most of these are in the states of Jharkhand, Orissa, West Bengal, Bihar, Chhattisgarh, and Madhya Pradesh. Numerous short-term and a few long-term studies have been performed focusing on ambient air quality in coal mine areas in India: Ghosh and Majee (2000, 2007) carried out suspended particulate matter (SPM) and respirable suspended particulate matter (RSPM) studies at Jharia (Dhanbad, Jharkhand) coal fields; Reddy and Ruj (2003) reported SPM, sulfur dioxide (SO2), and NOx status in Raniganj; Singh and Puri (2004) and Singh (2006) have carried out studies at six stations in Korba; Mukhopadhyay et al. (2010) also reported SPM, SO2, and NOx status in Bankola region (Raniganj); Jaiprakash et al. (2010) carried out monitoring and simulation of particulate matter (PM) in Dhanbad (Jharkhand) coal mine; and Trivedi et al. (2009, 2010a, b) carried out an air pollution study for the five mines. Chaulya (2004, 2005) has carried out SO2, NOx, and particulate matter studies from September 1998 to August 1999 in IB Valley Orissa, and Chaulya et al. (2012) also reported analytical results of physicochemical parameters and proximate analysis of coal dust collected from the road surface of four opencast coal mines located in different coalfields in India. Besides ambient air quality, a study of radio-elemental characteristics of fly ash from a thermal power plant of Chandrapur using coal from Padmapur mines is reported by Menon et al. (2011). Using satellite imagery, Mishra et al. (2012) studied air pollution concentration over Jharia. More recently, George et al. (2013) carried out studies at Chandrapur coal mine and reported PM10 in the ambient air and its comparison with other environments. All the studies mentioned above mainly focused on SO2, NOx, and particulate matter (SPM, RSPM, etc.) and for a limited period and specific area. Due to the significant importance of aerosol optical depth (AOD) in air pollution, for the first time, we present monthly long-term (March 2000 to December 2012) time series model prediction and trend analysis of AOD measurement over 11 selected coal mines in India. Coal and coal waste products release toxic-release chemicals, namely As, Pb, Hg, Ni, V, Be, Cd, Ba, Cr, Cu, Mo, Zn, Se, and Ra which are hazardous if released into the atmosphere. Coal use is one of the many human activities that generate greenhouse gas (GHG) emissions. There are severe health effects caused by burning coal. As a consequence of these activities, particulate matters and noxious gases are released into the atmosphere. Subsequently, suspended minute particulate matters and liquid droplet form aerosol. Atmospheric turbidity caused by aerosol is a whole indicator
of air pollution and it is measured by computing the AOD (Mishra et al. 2012). Consequently, aerosol studies over mining areas are of great importance mainly from the climatic point of view. In this respect, several studies on ambient air quality over the coal mines in India concerning particulate matter concentration (PM10, RSPM, etc.) have been performed during the last years; however, long-term studies of air quality are required to evaluate the environmental impact of coal mines (Jones 1993; Canter 1996; CMRI 1998; Chaulya et al. 1998, 2000; Ferreira et al. 2000). In order to install any control measure, the prediction of concentration of dust particles is essential (Ghosh 2002). This study is mainly important because satellites can provide observations for the whole domain of India. The Moderate Resolution Imaging Spectroradiometer (MODIS) measures radiances at 36 wavelengths from 0.41 to 14 μm, with a 2,330-km viewing swath allowing near-global daily coverage. AOD550 data from MODIS on-board the Terra satellite was used as the long-term objectives of the study at the satellite overpass time 10:30 a.m. Level 3 monthly data are produced by averaging the daily 10×10 km aerosol products at 1°×1° grid and are available in MODIS online visualization and analysis system (http://disc.sci.gsfc.nasa.gov/giovanni). In the present study, the autoregressive integrated moving average (ARIMA) modeling approach has been adapted to forecast aerosol optical depth over coal mines in India. The time series modeling is a useful tool in planning operating and decision-making of climatic fluctuations and has been commonly used for data generation forecasting estimation of missing data and extending hydrologic data records (Soltani et al. 2007). Many studies on time series mainly using the autoregressive (AR) model, moving average (MA) model, and the combination of two autoregressive moving average (ARIMA) models with air pollutant modeling have been made (Ballester et al. 2002; Abdel-Aziz and Frey 2003; Chelani and Devotta 2006; Portnov et al. 2009; Chattopadhyay and Chattopadhyay 2009; Liang et al. 2009; Abish and Mohanakumar 2013). With reference to future prediction of air pollutant concentrations, these studies have only concentrated on variables such as nitrogen dioxide (NO2), SO2, carbon monoxide (CO), and ozone (O3) instead of AOD. The present work mainly focuses on the time series analysis of the comprehensive AOD dataset obtained from Terra MODIS at 11 selected sites in India covering about a 13-year (March 2000 to December 2012) period. Moreover, statistical parameters like mean, standard deviation, coefficient of variation (CV), seasonal fraction (SF), and regression analysis were also calculated from the monthly AOD data. The AOD time series forecasting is an important analysis tool to extrapolate the current time series for future prediction of aerosol loading over the coal mines in India.
Environ Sci Pollut Res
Data and methodology
CVð%Þ ¼
MODIS dataset The main dataset consists of Terra MODIS (version 5.1, level 3, 1°×1°) monthly AOD550 observations centered over the mining area in India. More specifically, long-range (about 13 years) MODIS datasets during the period March 2000 to December 2012 have been obtained at 11 sites in six states in India, namely Raniganj in West Bengal, Jharia and Bokaro in Jharkhand, Bishrampur and Korba in Chhattisgarh, Talcher in Odisha, and Wardha in Maharashtra. Adilabad Warangal Godavari Valley and Cuddapah in Andhra Pradesh (see Fig. 1 and Table 1). According to the Central Pollution Control Board (CPCB) report (2012), most of the cities in the states, namely West Bengal, Chhattisgarh, Odisha, Jharkhand, and Maharashtra fall under critical pollution level category, while Andhra Pradesh is at the high pollution level with respect to SO2, NO2, and RSPM. The analysis of three pollutants during 2010 in each state showed that with respect to SO2, Jharkhand had the maximum annual average concentration (23 μg/m3) followed by Maharashtra (17 μg/m3). With regard to NO2, West Bengal had the maximum annual average concentration (64 μg/ m3) followed by Delhi (55 μg/m3), and concerning PM10, Delhi had the maximum annual average concentration (261 μg/m3) followed by Jharkhand (193 μg/m3). The AOD uncertainty over land is ΔAOD=±(0.05+15 %) (Levy et al. 2010). The total dataset has been classified into four seasons namely winter (December–January–February), pre-monsoon (March–April–May), monsoon (June–July– August–September), and post-monsoon (October– November) on the basis of meteorology over northern India (Ramachandran and Kedia 2012). Statistical analysis Statistical analysis is used to calculate average values, standard deviation, CV, and SF of the AOD data series over coal mines in India. The seasonal fraction of AOD denotes the mean seasonal contribution (percentage) to the total annual AOD and is defined as the ratio of the sum of AOD in each season to the sum of AOD in all seasons during a year, that is: SFð%Þfor AOD ¼
AODS 100 % AODY
ð1Þ
where AODs is the sum of AOD in a particular season and AODy is the sum of AOD in all months of a year. Coefficient of variation in (%) is used to analyze the temporal variability of aerosol optical depth. It is the ratio of standard deviation to the mean of the dataset (Rangarajan 1997; Parmar and Bhardwaj 2013a, b) and is defined as:
Standard Deviation 100 % Mean
ð2Þ
The percentage (%) variation in AOD during a study period has been calculated from the following formula (Kaskaoutis et al. 2012b): ! aN xð % Þ ¼ 100 % ð3Þ x¯ where x is the variable, a is the slope value of the linear regression analysis, and N is the whole number of records in the given period. Using the SPSS version 17, a linear regression model of first order was used to determine the best fit line. The confidence intervals at the 95 % level were calculated to test the statistical significance of the trends using the compared t test and p value (
