Environmental Pollution xxx (2014) 1e9 Contents lists available at ScienceDirect Environmental Pollution journal homepage: www.elsevier.com/locate/envpol Vegetation fires and air pollution in Vietnam Thanh Ha Le a, *, Thi Nhat Thanh Nguyen a, Kristofer Lasko b, Shriram Ilavajhala c, Krishna Prasad Vadrevu b, Chris Justice b a University of Engineering and Technology, Vietnam National University, E3, 144 Xuan Thuy, Cau Giay, Hanoi, Viet Nam Department of Geographical Sciences, University of Maryland, College Park, USA c Sigma Space Corporation, Maryland, USA b a r t i c l e i n f o a b s t r a c t Article history: Received March 2014 Received in revised form 15 July 2014 Accepted 17 July 2014 Available online xxx Forest fires are a significant source of air pollution in Asia In this study, we integrate satellite remote sensing data and ground-based measurements to infer fireeair pollution relationships in selected regions of Vietnam We first characterized the active fires and burnt areas at a regional scale from MODIS satellite data We then used satellite-derived active fire data to correlate the resulting atmospheric pollution Further, we analyzed the relationship between satellite atmospheric variables and groundbased air pollutant parameters Our results show peak fire activity during March in Vietnam, with hotspots in the Northwest and Central Highlands Active fires were significantly correlated with UV Aerosol Index (UVAI), aerosol extinction absorption optical depth (AAOD), and Carbon Monoxide The use of satellite aerosol optical thickness improved the prediction of Particulate Matter (PM) concentration significantly © 2014 Elsevier Ltd All rights reserved Keywords: Active fires Biomass burning Atmospheric variables Particulate matter concentration Vietnam Introduction Vietnam spans a land area of around 33 million ha, of which 13.9 million are classified as forest (10.4 million of natural forests, 3.5 million of forest plantations) (Ha, 2013) The anthropogenic causes of vegetation fires in Vietnam include: (i) burning of agricultural land: straw and grass residue burning in rice fields is 20%; (ii) local people using fire for hunting, trapping, and catching wild animals in forests, especially the use of smoke to harvest bee's honey causes 55%; (iii) forest product exploitation for timber, wood, cooking, and smoking account for 15%; (iv) the remaining 10% of forest fires are caused by trading conflicts from forest resource exploitation by stakeholders who burn forests to harm other competitors (Hoang, 2007) Table lists the number of forest fire occurrences and the area burned in Vietnam from January 2013 to August 2013 (Forest Protection Department (FPD) 2013) Forest fires are a major cause of aerosol emissions consisting of black and organic carbon, including mineral ash Biomass burning also results in the release of precursor compounds of Ozone (Pfister et al., 2008) The emitted aerosols may affect radiative properties of * Corresponding author E-mail address: ltha@vnu.edu.vn (T.H Le) the atmosphere, act as cloud condensation nuclei, and affect human activity by causing respiratory ailments (Nel, 2005) According to the national environmental report, air quality in Vietnam is degraded, especially in big cities like Hanoi and Ho Chi Minh (Pham et al., 2010) Major air pollutants are dust particles, carbon monoxide (CO), and sulfur dioxide (SO2) Analysis from ground-based measurements of Total Suspended Particles (TSP) and Particulate Matter (PM) concentration indicates that their concentrations exceed acceptable limits in these big cities The report also revealed that emission sources include transportation, thermoelectricity, cement production, and other industrial services, which are constantly contributing to air pollution (Fig 1) From the national perspective, wildland fire and air pollution information is a fundamental but challenging prerequisite for understanding and monitoring forests and air quality from local to regional scales In recent years, remote sensing is being used as a tool for monitoring regional fire hazards and air pollutant emissions in Vietnam Since 2007, the FPD, Ministry of Agriculture and Rural Development of Vietnam has deployed a Moderate Resolution Imaging Spectroradiometer (MODIS) Direct Broadcast receiving station in Hanoi with the primary purpose of early forest fire detection over all of Vietnam The system named FireWatchVN, which includes near real-time active fire hotspot detection, interactive web-mapping fire visualization, fire database and statistical analysis functions, is online (www.kiemlam.org.vn/FireWatchVN) since December 2008 (Nguyen et al., 2008) Regarding air pollution http://dx.doi.org/10.1016/j.envpol.2014.07.023 0269-7491/© 2014 Elsevier Ltd All rights reserved Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 Table Forest fires in Vietnam from January 2013 to August 2013 (FPD, 2013) Month January February March April May June July August Total Datasets and methodology Forest fire Occurrences observation satellites, ground-based station data, and information and communication technologies Burnt area (ha) 44 56 39 14 18 17 39 338 249 125 27.6 86.6 18 26 200 909.2 applications, the Center for Environmental Monitoring (CEM) General Environment is running an Environmental Monitoring Portal (http://cem.gov.vn/en-US/en/Home.aspx) which provides air pollutant data measured by automatic measurement stations and manual equipment Since automatic measurement stations for air quality are often established in big cities like Hanoi, Danang, etc., these air pollutant data alone are inadequate to assess wildland fire impacts, in particular, air quality The main purpose of this work is to understand the relationship between forest fire emissions and air quality in Vietnam Air pollution data are scarcely available at fire locations due to the limited number of ground-based measurement stations Therefore, we relied mostly on satellite remote sensing products for analyzing fire-pollution relationships In this study, the following tasks have been achieved: a) characterize the spatial and temporal trends of active fires at a regional scale; b) determine the statistical correlation between active fires and satellite atmospheric variables; (c) evaluate the relationships among ground-based air pollution indices (particulate matter concentration), meteorological parameters (wind speed, temperature, relative humidity, pressure and radiation), and satellite atmospheric variables over Hanoi, Vietnam In addition to the above analysis, we discuss the development of a geographic information system (GIS) useful to fire and air pollution managers in Vietnam The system is being developed and will contribute to address wildfire risk and related pollution over selected regions The system will be aimed at providing users with fire and air pollution detection, monitoring, alerting, planning, reporting and predicting capabilities through the use of Earth 2.1 Active fires data The active fires dataset from January 2004 to December 2012 (MCD14ML Collection active fire product), obtained from MODIS onboard NASA's Terra and Aqua satellites were used to provide a synoptic view of fire activity (Giglio et al., 2003) Both Fire count (FC) and average Fire Radiative Power (FRP) were used in this analysis (Giglio et al., 2003) 2.2 Satellite atmospheric variables The following six satellite-derived atmospheric variables were used in this study: MODIS Angstrom Exponent (MODIS AE): A daily Level-3 MODIS aerosol product whose calculation involves averaging the data from the two satellites AQUA and TERRA in every 1 Â1 pixel MODIS Aerosol Optical Depth level-3 (MODIS AOD): The daily Level-3 MODIS aerosol optical depth product reported in Hubanks et al (2008) is a global daily spatial aggregation of the Level-2 MODIS AOD (10-km spatial resolution) into a regular grid with a resolution of 1 Â1 UV Aerosol Index (UVAI) and aerosol extinction and absorption optical depths (AAOD): Daily products of the near-UV algorithm for Ozone Monitoring Instrument at two UV wavelengths 354 and 388, respectively (Herman et al., 1997; Torres et al., 1998) Measurements of Pollution in the Troposphere (MOPITT)'s Carbon Monoxide (MOPITT CO): A global daily product, namely, MOPITT Gridded Daily CO Retrievals (Near and Thermal Infrared Radiances) V005 with resolution of 1 Â1 PMMAPPER® AOT: The daily aerosol product provided by NASA, namely MOD04 L2, is derived from MODIS images using aerosol retrieval algorithms over land and ocean areas with spatial resolution at 10  10 km In an effort to improve the spatial resolution of the MODIS aerosol products, a commercial software package called PMMAPPER® was developed to estimate the aerosol products at  km from MODIS images, named as PMMAPPER® AOT (Nguyen et al., 2010) The validation was carried out over Europe for the years 2007e2009 (Campalani et al., 2011) and over Hanoi from 2011 to 2012 (Nguyen et al., 2013) Aerosol Optical Thickness/Aerosol Optical Depth (AOT/AOD) is considered as one of the Essential Climate Variables (ECV) that influences climate, visibility and quality of the air AOT is representative for the amount of particulates present in a vertical column of the Earth's atmosphere Aerosol concentration can be measured directly by ground-based sensors or estimated by sensors onboard polar and geostationary satellites observing the Earth 2.3 Particulate matter concentration and meteorological data For the PM measurement data over Hanoi, we relied on ground measurements provided by CEM in Vietnam Environment Administration The data are for 21020 56.300 , 105 220 58.800 and include PM concentration at different sizes (1, 2.5 and 10 mm) averaged over 24 h and the corresponding hourly meteorological data (wind speed, wind direction, temperature, relative humidity, pressure and solar radiation) from August 2010 to July 2012 2.4 Gridded data We analyzed our data at 1 Â1 and Vietnam Since atmospheric variables MODIS AE, MODIS AOD, AAOD, UVAI, and MOPITT CO are available at the same regular 1 Â1 grid, they were used directly without any additional processing The FC per grid cell is calculated by taking the sum of all active fire points within a cell and FRP as an average for all active fires within a cell Daily satellite aerosol maps are at km2 spatial resolution, while PM1/2.5/10 and meteorological data are measured at ground stations and averaged in twenty-four hours and one hour periods, respectively Cloud-free satellite data were used within a given radius (R) of the ground station Co-located measurements of CEM instruments are selected and averaged within a specific time window (T) around the satellite overpasses The optimal thresholds of R and T were selected by experiments presented in Nguyen et al (2013) The final integrated variables include PM1/2.5/10, PMMAPPER® AOT, wind speed, temperature, relative humidity, atmospheric pressure, and solar radiation 2.5 Estimating PM concentrations Fig Contribution of major pollutant emission sources in Vietnam 2008 [VEA, 2010] We used integrated datasets consisting of PM1/2.5/10, PMMAPPER® AOT, Wind Speed (Wsp), Temperature (Temp), Relative Humidity (RelH), Pressure (Bar) and solar Radiation (Rad) to estimate PM Multiple Linear Regression (MLR) and Support Vector Regression (SVR) methods were tested Using the MLR technique, particulate matter concentration in different diameters was calculated using a linear equation as follows: Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 (a) (b) Fig Active fires in Vietnam by year (a) and month (b) from 2004 to 2012 PM ẳ b0 ỵ b1 AOT þ b2 Wsp þ b3 Temp þ b4 RelH þ b5 Bar ỵ b6 Rad (1) where PM is PM10, PM2.5 or PM1 mass concentrations (m gmÀ3), AOT is PMMAPPER® AOT at 0.553 mm, b0 is intercept for the PM equation whereas b1À6 are regression coefficients for the predictor variables including AOT, wind speed (m sÀ1), temperature (C), relative humidity (%), barometer (hPa) and radiation (W mÀ2), respectively The ε-SVR, first introduced by Vapnik (1996), will find a function that has at most ε deviation from the actually obtained target from the training data in order to be as ‘flat’ as possible to minimize the expected risk Regarding PM estimation, the ε-SVR with epsilon loss function and Radial Basic Function kernel provided by LIBSVM demonstrated in Chang and Lin (2011) was applied Results and discussions 3.1 Active fires in Vietnam For Vietnam, the number of active fires detected from MODIS for different years is given in Fig 2a and monthly variations in Fig 2b An average fire count of 16,086 per year was noted from 2004 to 2012 The peak year was 2010 with 21,750 fire counts A monthly fire progression is observed, with the peak occurrence in March (Fig 2b) Vietnam is divided into eight administration regions: Northwest, Northeast, Red River Delta, North-Central Coast, SouthCentral Coast, Central Highlands, Southeast, and Mekong River Delta Following annual reports from Vietnam Forestry Administration, the months having high possibility of fires (except in coastal areas) are December, January, February, March and April Similar results were reported in Ha (2013) The results are quite consistent with analysis of active fires and monthly trends presented in Fig Results from  1 mapping of fires suggest varied spatial distribution throughout the country However, the peak burning cells are concentrated in two regions, one in the Northwest and the other in Central Highlands (Fig 3a) We also noted the Mekong Delta as one of the important regions of biomass burning, from crop residues (Fig 3b) Fig Number of active fires in Vietnam (a) and in the peak March (b) from 2004 to 2012 Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 Fig represents the land use map over Vietnam for the year 2005 Based on the map, it is evident that fires over the Northwest region and especially Mekong delta are from agriculture activities Meanwhile, active fires in the Central Highlands are from both forest and agriculture The rainfall and temperature maps were used to relate to active fires Rainfall maps were created from real-time Tropical Rainfall Measuring Mission (TRMM) multi-satellite precipitation data (Huffman et al., 2007, 2010) Fig 5a and b depicts the maps of annual total rainfall and total rainfall in March from 2004 to 2012 respectively Temperature maps were spatially interpolated from 100 ground stations data managed by National Centre for Hydro- Meteorological Forecasting (NCHMF) Fig 5c and d depicts the yearly average temperatures and March average temperature from 2004 to 2012 respectively Less rainfall and high temperature can be seen in the Northwest and in Central Highlands regions In March, high temperatures are observed in all Northwest, Central Highlands and Mekong delta regions coinciding with active fires 3.2 Active fires and atmospheric variables Table shows correlation coefficients between active fires and atmospheric variables over the study area Results indicate fires correlating with MODIS AOD, AAOD, UVAI and MOPITT CO (0.14, 0.21, 0.24, and 0.33 respectively) Separated by year, MOPITT CO has the highest correlation coefficient meanwhile UVAI and AAOD have similar behavior (Fig 6) Compared to fireeCO correlations reported for the forest fires of Himalayas (Vadrevu et al., 2013), the correlations for Vietnam seem relatively low High correlations were found for 2006, 2007 and 2010, for AAOD, UVAI and MOPITTCO (Fig 6) in contrast to the other years Fig shows correlation coefficient between FC and AAOD, UVAI and MOPITT CO for nine years of monthly average data The peak month is April and it is consistent for all AAOD, UVAI and MOPITT CO variables Correlation coefficients between FC and UVAI and MOPITT CO are strong in August and September as well, but the number of active fires in those months were small (36 and 37, respectively) However, the relationship between active fires/ biomass burning and atmospheric variables is not consistent with the number of active fires presented in Fig by both month and year From a spatial perspective, the correlation between the number of active fires and all AAOD, UVAI and MOPITT CO are strongest for cells in the Northwest followed by the Central Highlands (Fig 8) Correlation coefficients for the Northwest region and the Central Highlands region are shown in Figs and 10, respectively High correlation coefficients of FC and UVAI/MOPITT CO are also found in other regions but the number of fires in those areas is often small (Fig 3) 3.3 Atmospheric variables and PM concentrations Fig The land use map in Vietnam in 2005 The relationship between satellite AOD and meteorological parameters to ground-based particulate matter concentrations (PM1, PM2.5 and PM10) and the benefits of using PMMAPPER® AOT in estimating PM1, PM2.5 or PM10, are discussed in this section Because of the limited number of automatic ground stations, the case study was conducted in Hanoi, the capital city of Vietnam First, satellite and ground-based measurements are integrated in order to resolve varying spatial and temporal resolutions with integration thresholds from Nguyen et al (2013) On the integrated dataset, the correlation coefficient of PM1, PM2.5 and PM10 in relation to other factors has been calculated to assess their roles in PM estimation Fig 11 shows the dependence of PM on PMMAPPER® AOT increase in the order of their aerodynamic diameters (i.e 1, 2.5 and then 10 mm) whereas their relationship to PM with meteorological variables (Wsp, Temp, Rel H, Bar and Rad) decrease with their size (i.e 10, 2.5 and then mm) These results confirm the importance of satellite PMMAPPER® AOT for estimating groundbased PM In addition, temperature, radiation and wind speed should be more important than relative humidity and pressure Secondly, MLR and SVR techniques presented in Section 2.5 are applied to estimate PM1/2.5/10 values Data are grouped into two years (Year from August 2010 to July 2011 and Year from August 2011 to July 2012) with statistics shown in Fig 12 In two years, the total samples are comparable (55,131 and 49,908) but data distributions in Aug, Oct, Nov, Jan, Feb, Mar and Apr are very different Therefore, we considered data for different years, instead of months Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 Fig Precipitation and temperature maps summarized from data collected from 2004 to 2013 Yearly average rainfall (10À2 mm) (a) March average rainfall (10À2 mm) (b) Yearly average temperature (Celsius degree) (c) March average temperature (Celsius degree) (d) Table Correlation coefficient (R) between atmosphere variables (MODIS AOD, MODIS AE, AAOD, UVAI, MOPITT CO) with active fires (FC) and Fire radiative power (FRP) MODIS AOD MODIS AE AAOD UVAI MOPITT CO FC FRP 0.142 0.035 0.211 0.238 0.330 À0.002 0.123 0.071 À0.021 0.178 or seasons In addition, datasets with (w) or without (w/o) satellite AOT are created and considered One year of data was used for estimating models, whereas another year of data was used to test the models and vice-versa The final correlation coefficient (R) and Root Mean Square Error (RMSE) were reported in Table The use of satellite AOT improved PM1/2.5/10 prediction significantly In the case of PM10 estimation, correlation increased from 0.038 to 0.174 and 0.239 when MLR and SVR were applied The correlation coefficient of PM2.5 estimation increased from 0.429 to Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 Fig Correlation coefficient between FC and AAOD, UVAI, MOPITT CO separated by year Fig Correlation coefficient between FC and AAOD/UVAI/MOPITT CO over the Northwest region by year 0.598 and 0.593 while the same trend is also seen on PM1 estimation (from 0.608 to 0.659 and 0.694, respectively) Our results (Table 3) suggest that PM1 and PM2.5 estimation can be done effectively using MLR and SVR techniques, whereas PM10 estimation needs more data and further investigation Further, when using satellite AOT, we found SVR better than MLR for PM10 and PM1 estimation The correlation for PM10 and PM1 increases (from 0.17 to 0.24 and from 0.65 to 0.69, respectively) while error decreases (from 96.65 to 74.93 and from 22.93 to 22.34, respectively) Meanwhile, MLR and SVR perform in nearly the same way for PM2.5 estimation, which is presented by a slight difference of correlation (R ¼ 0.59) and error (RMSE are 31.07 and 31.67) In general, SVR outperformed MLR 3.4 Proposed system for forest fire and air pollution monitoring Fig Correlation coefficient between FC and AAOD, UVAI, MOPPIT CO separated by month Based on the fire interactive visualization system UETFIRMS (2013) and the personal discussions with FPD and CEM, we are Fig Correlation coefficient between FC and AAOD (a), UVAI (b), MOPITT CO (c) from 2004 to 2012 Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 Table MLR and SVR performance on PM10, PM2.5 and PM1 estimation PM2.5 PM10 MRLw/o PM1 MRL-w SVR-w MRL- MRL-w SVR-w MRL- MRL-w SVR-w w/o w/o R 0.03 0.17 RMSE 109.22 96.65 0.24 74.93 0.42 0.59 40.83 31.07 0.59 31.67 0.60 0.65 24.59 22.93 0.69 22.34 R: Correlation coefficient, RMSE: Root Mean Square Error (mg/m3) Fig 10 Correlation coefficient between FC and AAOD/UVAI/MOPITT CO over the Central Highlands region by year will also provide information on wildfire and air pollution detection, monitoring, alerting, planning, and reporting capabilities through the use of Earth observation satellites, ground-based stations and communication technologies To develop the above system, scientists from VNU are collaborating with the University of Maryland, College Park and the Global Observation of Forest Cover (GOFC) Fire Implementation Team As a result, a localized version of the Fire Information for Resource Management System (FIRMS) (Davies et al., 2009, FIRMS 2013) was installed at University of Engineering and Technology, Vietnam National University, Hanoi and accessed through UETFIRMS (2013) This system provides an interactive visualization of global active fires and global MODIS burnt areas It also includes an email fire alert function This system is being integrated with FAIR monitoring system along with the other components for improved air pollution monitoring in the region Conclusion Fig 11 Correlation coefficient of PM1/2.5/10 to AOT and meteorological parameters developing an integrated system of Forest fire and Air pollution (FAIR) monitoring system as shown in Fig 13 FAIR will provide fire and air pollution managers with a unique tool to better manage the risk of wildfire and related air pollution over selected regions FAIR Active fires in Vietnam were spatio-temporally analyzed An average number of 16,086 fires per year were recorded (2004e2012) The peak fire season is during March and the peak year was 2010 with 21,750 active fires From a spatial perspective, the two important hotspot regions include Northwest and Central Highlands We found the highest correlation of fire counts with AAOD, UVAI and MOPITT CO We found significant correlations of active fires with AAOD, UVAI and MOPITT CO We found dependence of PM on AOT increase in the order of their aerodynamic diameters (i.e 1, 2.5 and then 10 mm) whereas relationship of PM with meteorological variables (Wsp, Temp, Rel H, Bar and Rad) was found to decrease with their size (i.e 10, 2.5 and then mm) The use Fig 12 Number of samples on monthly basis in two years (Year from August 2010 to July 2011 and Year from August 2011 to July 2012) Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 Fig 13 The system concept of FAIR for forest fire and air pollution monitoring of satellite AOT data improved the PM1/2.5/10 prediction significantly An integrated system to map and monitor fires and air quality at the national level, named “FAIR” is being developed FAIR will robustly incorporate both ground-based and satellite-based data for fire and pollution detection, prediction, planning and reporting to better manage fires and improve air quality in Vietnam Acknowledgment This research was possible due to the funds received from two different projects, “Forest fire information system, QGTD.13.26” and “Air pollution monitoring and warning system, QGTD.13.27” from Vietnam National University, Hanoi We are grateful to MEEO S.r.l (Ferrara, Italy) for providing satellite aerosol maps and Center for Environmental Monitoring, Vietnam Environment Administration for PM mass concentration, meteorological data We are grateful to the satellite product developers of MODIS active fires, MODIS AE, MODIS AOD, AAOD, UVAI, and MOPITT CO for freely sharing the data through NASA portals References Campalani, P., Nguyen, T.N.T., Mantovani, S., Bottoni, M., Mazzini, G., 2011 Validation of PM MAPPER aerosol optical thickness retrievals at 1x1 km2 of spatial resolution In: Proceeding of Software, Telecommunications and Computer Networks (SoftCOM), the 19th In-ternational Conference, pp 1e5 Chang, C., Lin, C., 2011 LIBSVM: a Library for Support Vector Machines Davies, D.K., Ilavajhala, S., Wong, M.M., Justice, C.O., 2009 Fire information for resource management system: archiving and distributing MODIS active fire data IEEE Trans Geosci Remote Sens 47, 72e79 FIRMS, 2013 https://firms.modaps.eosdis.nasa.gov/firemap/ Forest Protection Department (FPD), Ministry of Agriculture and Rural Development of Vietnam, 2013 Monthly Report on Forest Fire and Destruction Giglio, L., Descloitres, J., Justice, C.O., Kaufman, Y., 2003 An enhanced contextual fire detection algorithm for MODIS Remote Sens Environ 87, 273e282 Ha, C.T., 2013 Vietnam National Forest Status of 2012 Annually report of Ministry of Agriculture and Rural Development Herman, J.R., Bhartia, P.K., Torres, O., Hsu, C., Seftor, C., Celarier, E., 1997 Global distribution of UV-absorbing aerosols from Nimbus-7/TOMS data J Geophys Res 102, 16911e16922 Hoang, M.H., 2007 Status of Forest Fire Management and Control in Vietnam GEOSS AP Symposium Integrated Observation for Sustainable Development in the Asia-Pacific Region, Tokyo, Japan Hubanks, P., King, M., Platnick, S., Pincus, R., 2008 MODIS Atmosphere L3 Gridded Product Algorithm Theoretical Basis Document Collection 005 Version 1.1 Tech Rep ATBD-MOD-30, NASA Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 T.H Le et al / Environmental Pollution xxx (2014) 1e9 Huffman, G.J., Adler, R.F., Bolvin, D.T., Gu, G., Nelkin, E.J., Bowman, K.P., Hong, Y., Stocker, E.F., Wolff, D.B., 2007 The TRMM multi-satellite precipitation analysis: quasi- global, multi-year, combined-sensor precipitation estimates at fine scale J Hydrometeor (1), 38e55 Huffman, G.J., Adler, R.F., Bolvin, D.T., Gu, G., Nelkin, E.J., 2010 The TRMM multisatellite precipitation analysis (TMPA) (Chapter 1) In: Hossain, F., Gebremichael, M (Eds.), Satellite Rainfall Applications for Surface Hydrology Springer Verlag, ISBN 978-90-481-2914-0, pp 3e22 National Environmental Report 2010eOverview of Vietnamese Environment Download at: http://www.vea.gov.vn/Files/BaoCaoMTQG2010.rar Nel, A., 2005 Air pollution-related illness: effects of particles Science 308, 804e806 Nguyen, H.Q., Nguyen, H.M., Nguyen, T.H., Nguyen, T.H., Tran, H., Dang, V.T., 2008 Satellite Data Acquisition and Utilization for Forest Fire Management in Vietnam GeoInformatics for Spatial-Infrastructure Development in Earth and Allied Sciences (GIS-IDEAS) Nguyen, T., Mantovani, S., Bottoni, M., 2010 Estimation of aerosol and air quality fields with PMMAPPER, an optical multispectral data processing package In: ISPRS TC VII Symposium100 Years ISPRSeAdvancing Remote Sensing Science, vol XXXVIII (7A), pp 257e261 Nguyen, T.N.T., Ta, V.C., Le, T.H., Mantovani, S., 2013 Particulate matter concentration estimation from satellite aerosol and meteorological parameters: datadriven approaches In: Proc Of Knowledge and Systems Engineering, pp 351e362 Pfister, G., Wiedinmyer, C., Emmons, L., 2008 Impacts of the fall 2007 California wildfires on surface ozone: Integrating local observations with global model simulations Geophys Res Lett 35, L19814 http://dx.doi.org/10.1029/ 2008GL034747 Pham, K.N., Bui, C.T., Le, K.S., 2010 National Environmental Report 2010: Survey of Vietnam Environment, Ministry of Natural Resources and Environment of the Socialist Republic of Vietnam Torres, O., Bhartia, P.K., Herman, J.R., Ahmad, Z., 1998 Derivation of aerosol properties from satellite measurements of backscattered ultraviolet radiation: theoretical basis J Geophys Res 102, 17099e17110 UETFIRMS, 2013 http://gsht.uet.vnu.edu.vn:8080/firemap Vadrevu, K.P., Giglio, L., Justice, C., 2013 Satellite based analysis of fireecarbon monoxide relationships from forest and agricultural residue burning (2003e2011) Atmos Environ 64, 179e191 Vapnik, V., 1996 The Nature of Statistical Learning Theory Springer-Verlag, Newyork Abbreviations AAOD: Aerosol extinction Absorption Optical Depth AOD: Aerosol Optical Depth AOT: Aerosol Optical Thickness CEM: Center for Environmental Monitoring CO: Carbon monoxide FAIR: Forest fire and AIR pollution FC: Fire count FIRMS: Fire Information for Resource Management System FPD: Forest Protection Department FRP: Fire Radiative Power GOFC: Global Observation of Forest Cover MLR: Multiple Linear Regression MODIS: Moderate Resolution Imaging Spectroradiometer MODIS AE: MODIS Angstrom Exponent MODIS AOD: MODIS Aerosol Optical Depth level-3 MOPITT: Measurements of Pollution in the Troposphere (MOPITT) MOPITT CO: MOPITT's Carbon Monoxide NCHMF: National Centre for Hydro-Meteorological Forecasting PM: Particulate Matter SO2: Sulfur dioxide SVR: Support Vector Regression TSP: Total Suspended Particles UETFIRMS: Alocalized version of the FIRMS installed at University of Engineering and Technology, Vietnam National University, Hanoi UVAI: UV Aerosol Index VEA: Vietnam Environment Administration VOC: Volatile Organic Compound Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution (2014), http://dx.doi.org/ 10.1016/j.envpol.2014.07.023 ... active fires in Vietnam (a) and in the peak March (b) from 2004 to 2012 Please cite this article in press as: Le, T.H., et al., Vegetation fires and air pollution in Vietnam, Environmental Pollution. .. between FC and AAOD/UVAI/MOPITT CO over the Central Highlands region by year will also provide information on wildfire and air pollution detection, monitoring, alerting, planning, and reporting capabilities... “FAIR” is being developed FAIR will robustly incorporate both ground-based and satellite-based data for fire and pollution detection, prediction, planning and reporting to better manage fires and