Efficient application of the Radiance Enhancement method for detection of the forest fires due to combustion-originated reflectance
Rehan Siddiqui, Rajinder K. Jagpal, Sanjar M. Abrarov, Brendan M. Quine
EEfficient application of the RadianceEnhancement method for detection of theforest fires due to combustion-originated reflectance
Rehan Siddiqui
1, 2, 3, 4 , Rajinder K. Jagpal
2, 3, 4 , Sanjar M. Abrarov
1, 2, 3, 5 , andBrendan M. Quine
1, 4, 51Dept. Earth and Space Science and Engineering, York University, 4700 Keele St., Canada, M3J 1P32Epic Climate Green (ECG) Inc., 23 Westmore Dr., Unit 310, Toronto, M339V 3Y73Epic College of Technology, 5670 McAdam Rd., Mississauga, Canada, L4Z 1T24Dept. Physics and Astronomy, York University, 4700 Keele St., Toronto, Canada, M3J 1P35Thoth Technology Inc., Algonquin Radio Observatory, Achray Road, RR6, Pembroke, Ontario,Canada, K8A 6W7
February 3, 2021
Abstract
The existing methods detect the cloud scenes are applied at rela-tively small spectral range within shortwave upwelling radiative wave-length flux. We have reported a new method for detection of the cloudscenes based on the Radiance Enhancement (RE). This method can beused to cover a significantly wider spectral range from 1100 nm to 1700nm (Siddiqui et al., 2015) by using datasets from the space-orbitingmicro-spectrometer Argus 1000. Due to high sunlight reflection ofthe smoke originated from the forest or field fires the proposed REmethod can also be implemented for detection of combustion aerosols.The approach can be a promising technique for efficient detection andcontinuous monitor of the seasonal forest and field fires. To the bestof our knowledge this is the first report showing how a cloud methodcan be generalized for efficient detection of the forest fires due tocombustion-originated reflectance.
Keywords : radiance enhancement, clouds, forest fire, radiative trans-fer model, line-by-line calculation, micro-spectrometer a r X i v : . [ phy s i c s . a o - ph ] F e b Introduction
Increase of carbon dioxide gas appearing as a result of intense productionof goods in industrial and agricultural sectors of economy is a main issueof the modern human society that causes uncontrollable rise of atmospherictemperature due to devastating greenhouse effect. In particular, just in a fewfive decades the concentration of CO has been rapidly raised from 288 ppmto more than 410 ppm in 2020 [1–3]. As a consequence, the global warming ofthe atmosphere variates the weather dynamics causing tremendous negativeimpact to the flora and fauna of the Earth [4].The importance of clouds and their significant roles in sustaining the tem-perature balance on the Earth cannot be overestimated [5–8]. However, astable positive rate 2.05 ± Fig. 1.
The space observation package of Argus 1000.Inset shows a photograph of the light and small-sizeArgus 1000 micro-spectrometer.
The ultra-light (about 0.25 kg only), small-size and inexpensive Argus1000 instrument was launched into space in 2008 from India as a payload ofthe CanX-2 nanosatellite [20, 21]. Being in space it provides the large-scaledatasets as shown in Fig. 1 that can be used to extract valuable informa-tion about atmospheric gas constituents like concentration of CO and othergreenhouse gases by retrieving of the IR spectral radiance [21–23]. Inset inFig. 1 demonstrates a photo of the Argus 1000 micro-spectrometer.3part from determination of CO and other greenhouse gases, the IR ra-diance data collected by Argus instrument from space enable us to developthe RE method for efficient detection of the cloud scenes. Recently we sug-gested that a detection method of the cloud scenes can be generalized fordetection of the forest and field fires [19, 24]. Fig. 2.
Map of Canada with forest fire locationsduring the period from 1980 to 2019 [25].
Figure 2 shows a map of Canada with forest fire locations depicted bythe red spots [25]. In average a typical fire in a forest covers the range ofa few squared kilometers only. However, its impact to the forest is muchdevastating. In particular, the forest fire causes the smoke that extendsfor tenths and hundreds of squared kilometers increasing the temperature,spoiling air quality and preventing the sunlight locally for several months.Taking into consideration coverage areas of the spreading smokes, one canestimate from this map that about two thirds of the forest territories inCanada have been affected by forest fires during 1980 to 2019 years [25].The deficiency of the sunlight, pollution occurring due to smokes and anincreased temperature of the environment can kill pine trees even in a largerquantity than the actual fire that burns them down. As a result, beingunaffected directly by a fire these trees retain all material qualities including4igh density of wood and beam length. Consequently, these dead pine treeswith high quality of wood are logged in a commercially large scale.Conventional methods for forest fire detection include video-cameras,thermal imaging cameras, IR spectrometers to identify the spectral char-acteristics of smoke, light detection and ranging systems (LIDAR) and so on[26–28]. However, in contrast to the conventional methods, the RE techniquehas some advantages such as global coverage and continuous monitor due toperiodic observation from the space-orbiting nanosatellite in the real timemode. Another important feature is that the RE method can cover a widespectral wavelength range to detect enhancement of flux from the surfaceespecially due to clouds [18, 19].In this work we show how the RE technique can be used to detect effi-ciently the forest fires by using Argus 1000 space data. Our method is basedon a match between synthetic and observed radiance that accounts for highsurface reflectance appearing as a result of smokes produced by forest fires.To the best of our knowledge, a generalization of any water vapor cloudmethods for detection of the forest fires has never been reported in scientificliterature.
The RE methodology is based on least square match between space obser-vation and synthetic datasets. In order to generate synthetic data we usedline-by-line radiative transfer model GENSPECT [29]. This forward modelcomputes the radiance for the nadir and limb observations for the greenhousegases with all required parameters provided by the HITRAN molecular spec-troscopic database [30]. Some supplementary MATLAB files are additionallydeveloped to improve performance of the model. In particular, for computa-tion of the absorption coefficients we applied a newly modified code for morerapid and accurate computation of the Voigt function based on a new single-domain interpolation technique [31] for which the highly accurate referencevalues can be generated by using any of three rapid algorithms describedin our works [32], [33] or [33]. In contrast to the conventional algorithms[35, 36], in our implementation we interpolate the Voigt function in a singledomain itself in order to avoid unnecessary interpolation in computation ofthe absorption coefficients.The model GENSPECT accounts for different variables that include con-5entrations of the greenhouse gases, deviation of the nadir angle, zenith an-gle of sun, and so on [29]. In the latest updates we developed some func-tion files that also account for wavelength dependency of the reflectancedue to different surface albedo like clouds, pine-trees, vegetation and grass[37–40]. The radiative transfer model is run in a nested loop by increment-ing/decrementing values of the fitting variables until a best match is achieved.The RE methodology is based on the following formula [18, 19] RE i = 1 N N (cid:88) j =1 (cid:26) OBS i [ j ] − SY N i [ j ] SY N i [ j ] (cid:27) , where i is the index of wavelength sub-bands, j is the index of grid-pointsand N is the number of sub-bands that can be taken as 4. This methodologyshows the efficiency in determination of the cloud scenes. The correspondingCombined Radiance Enhancement (CRE) formula is given by [18, 19] CRE = N (cid:88) i =1 RE i . The RE and CRE values can be used to predict the cloud scenes. Specif-ically, when CRE is small and relatively close to zero, we can expect higherchances for cloud scene for a specific location due to high surface albedo.For example, if the surface albedo is relatively high, say above 0.6, then thebest match by RE method signifies that the specific observation is likely dueto thick cloud or any scattering particles such as ice pallets or aerosol. TheCRE is pre-defined for the wavelength bands and accounts for concentrationof the selected greenhouse gases [18, 19].Although the RE method cannot distinct the cloud scenes with aerosolsin from of solid particles and liquid droplets (including particulate mattersPM . [41]), it, nevertheless, can be advantageous in practical applications.If the weather forecast predicts no presence of clouds while the RE methodshows their availability, then we can conclude that these type of reflectancecould be due to aerosols like dust, industrial pollutants from big plants,hydroelectric stations or, more likely, combustion due to seasonal forest firesthat typically produce a large amount of smokes. Consequently, as a resultof high reflectance of the combustion-based aerosols, the RE method can alsobe used efficiently for detection of the forest fires.6 ig. 3. Flow-chart for computation of the RE method.
Figure 3 shows flow-chart of the RE method for detection of the forestfire that compares space observation and synthetic data. As we can seefrom this figure, in the intermediate stages the synthetic data is also dividedinto four sub-bands and passed to the slit function smoothing that simulatesresolution of the Argus instrument. Once the RE computation is completed,the comparison of the space observation and synthetic data is performed andif there is a best match, then the corresponding observation data is chosen.The detailed description of the RE algorithmic implementation can be found7n the work [18].
The line-by-line radiative transfer tool GENSPECT [29] generates syntheticspectral radiance for the greenhouse gases that can be used for comparisonwith space observation data. Figure 4 shows the synthetic spectral radiancefor CO , H O , O and CH gases computed at constant reflectance. The greenshadow area is originally computed radiance while the red curve depicts theslit function smoothed radiance that simulates the limited resolution of thespace instrument. The arrows in this figure indicate absorption band due toO near 1260 nm, the wider band of absorptions due to water vapors from1300 nm to 1480 nm, two absorption bands near 1575 nm and 1600 nm aredue to CO and one narrow absorption band near 1650 nm due to CH [19,22]. Each of these wavelength bands can be used in the RE method for thecomparison with Argus observation data in order to detect cloud scenes andforest fires. Concentrations of H O and CO bands are especially essentialvariables for matching of the synthetic model with space dataset from Argus1000 micro-spectrometer. Fig. 4.
Synthetic radiance computed by radiative transfermodel GENSPECT [27] at a constant reflectance.
The geolocation of the Argus instrument has been determined with helpof Systems Tool Kit (STK) [42], [43] and [44] for location of the forest firesover Canada. The datasets from 2009 to 2015 for Argus instrument has8een processed and analysed for water vapour and combustion-based aerosolclouds. Figure 5 shows Argus 1000 path, observation numbers from 12 to 70for week 11 and pass 69, over British Columbia, Canada. As we can see fromthis figure, the red spots showing the actual forest fire locations intercept thetrace of space orbiting IR remote sensor.
Fig. 5.
Geolocation of the Argus 1000 micro-spectrometercorresponding to week 11, pass 69 and observation numbersfrom 12 to 70, over British Columbia, Canada.
Increase of atmospheric temperature gradually changes the canopy con-stituents of the forests; the burned and dead pine trees are not necessarilyreplaced by new generation of pine trees. Due to increasing atmospheric tem-perature, the chances for spreading of broadleaf trees and bushes are highersince for successful competition of young generation of the pine trees withother species a colder environment is highly preferable. Nowadays we canobserve explicitly that after seasonal forest fires in the Algonquin NationalPark, Ontario, Canada, the pine trees are replaced by broadleaf trees, bushesand grasses changing the appearance and nature of the forest. Moreover, withdevelopment of the agricultural sector the more and larger territories of theforests are transformed to prairies, which are intensively used now by farmersfor grazing cows, lambs, goats and horses. The density of trees in the forestsis also declined due to increasing demand of most valuable wood of pine treesin the market and environmental pollutions.Surface reflectance of forests is generally the wavelength dependent ratherthan a constant. Therefore, a more rigorous consideration requires that thereflectance has to include cumulative contribution of cloud, pine trees, veg-9 ig. 6.
Reflectance dependencies for thick cloud, vegetation,pine trees, grass and their cumulative weighted sum. etation (broadleaf trees and bushes) and grass. The visual analysis of theforest in British Columbia suggests that pine trees occupy 0.5 to 0.7 of theground area, while remaining area is occupied by broadleaf trees, bushes andgrasses. Figure 6 shows reflectance as a function of the wavelength for thickcloud, vegetation, pine trees and grass, which data can be obtained from[37–40].
Fig. 7.
Evolution of the cumulative reflectance computed byweighted sum at different contribution factors.
The forests can occasionally share the land with turbid rivers and lakes.However, it is relatively rare when forest fires occur in the neighborhoodof water. Therefore, we do not consider these events in our model. As an10xample, Fig. 7 illustrates the evolution of the reflectance occurring dueto contribution factor (CF) from the vegetation. In our model we considercumulative reflectance from the surface due to thick cloud, pine trees, vege-tation and grass. The computation was performed by weighted sum methodby using the contribution factors from all four types of surface reflectance.
Fig. 8.
Synthetic radiance (all slit function smoothed) com-puted at different wavelength dependent reflectance.
Fig. 9.
Synthetic radiance (smoothed) computed for cu-mulative wavelength dependent reflectance at CF = 0.3.
Figure 8 shows the radiance computed at constant reflectance, thick cloud,vegetation, pine trees, grasses and by weighted sum due to forest fire (FF).As we can see from this figure, the wavelength dependency of the reflectance11ignificantly changes the radiance. Particularly, the radiative transfer modelGENSPECT [29] generates the graphs where the right portion above 1400nm is suppressed. This suppression effect on the right part of the spectralregion plays an important role in retrieval process of the space observabledata. Despite this one can still observe the profound absorption bands near1575 nm and 1600 nm due to carbon dioxide greenhouse gas.Figure 9 shows the synthetic radiance (green shadow area) and slit func-tion smoothed radiance (red curve) that accounts for the wavelength depen-dent radiance. Comparing Figs. 4 and 9 with each other we can observesignificant changes in the radiance. However, it should be noted that despitethe suppression on the right portion of the graphs above 1400 nm, the twoabsorption bands of CO greenhouse gas near 1575 nm and 1600 nm stillremain profound. Fig. 10.
Comparison of synthetic radiance (smoothed)radiance and space observation radiance.
It has been found by experimental fitting that CF = 0 . in calculation is increase by 40%. Alongsidewith the RE model, the increased level of CO also supports an assumptionfor presence of the forest fire. The synthetic spectrum shown in Fig. 10has been computed by incorporating all four different types of surface re-12ectance, specifically due to aerosol cloud, pine trees, vegetation (broadleaftrees and bushes) and grass. Space observation radiance spectrum in Fig. 8corresponds to Argus week 11, pass 69, observation number 49. Fig. 11.
Absolute difference between synthetic and space ob-servation data: a) 2D plot and b) 3D plot.
Figures 11a and 11b show the absolute difference between synthetic andspace observation data in 2D and 3D plots, respectively. As we can seefrom these figures, the darkest blue areas correspond to CF = 0.3. Changeof vegetation CF higher than 0.6 or lower than 0.2 significantly increasesthe absolute difference. The weightages for the cloud, pine trees, vegetation(broadleaf trees and bushes) and grass are found to be 0.4, 0.3, 0.2 and0.1, respectively. The error analysis for this RE method shows a reasonableagreement between synthetic and Argus space observation radiance to detectthe forest fire location with given IR wavelength range.Table 1:
RE and CRE values for the week 11, pass 69 and observation numbers40 to 52.
Week Pass Obs. ( O ) RE ( H O ) RE ( CO ) RE ( CH ) CRE
11 69 40 0.7214 2.344764 2.260691 0.783366 6.11022111 69 41 0.549205 1.764849 1.727316 0.508873 4.55024311 69 42 -0.00688 0.753885 0.876892 0.039664 1.66356411 69 43 0.135247 0.429764 0.61495 0.072215 1.25217611 69 44 0.124571 0.213207 0.215502 -0.13654 0.41673911 69 45 -0.41421 -0.12293 -0.06696 -0.40725 -1.0113411 69 46 -0.39236 -0.11299 -0.06674 -0.40589 -0.9779911 69 47 -0.31398 -0.04154 0.030332 -0.33553 -0.6607211 69 48 -0.07514 0.165338 0.444729 0.013554 0.54847711 69 49 -0.78618 -0.30506 -0.12341 -0.46567 -1.6803211 69 50 -0.48266 -0.11653 -0.04798 -0.4106 -1.0577611 69 51 0.446414 0.882347 0.535246 -0.06945 1.79455711 69 52 0.74834 1.229831 0.88675 0.212886 3.077807 ig. 12. Frequency vs. CRE bar chart. The red lineindicates the threshold value.
The space observed flux are generally gives enhanced radiance due tolarger oblique view angles as compared with nadir view angles because of highclouds and aerosols clouds thickness. Atmospheric path length and albedo isalso a major contributor to give high radiance enhancements. The latitudeangular dependence is also an important parameter in our calculations tofind RE [18].Table 1 shows the RE and CRE values of individual wavelength bands ofO , H O , CO and CH . The higher values in CRE correspond to the higherreflectance due to thick clouds or forest fire aerosols clouds. In this study weused selected bunch of space observed values from observation numbers 40to 52. The observation numbers 43 to 49 are in a good agreement with ourRE model for the forest fire detection.In our RE model we incorporate all important parameters as discussedin this section. Figure 12 demonstrates the frequency bar chart of aerosolcloud scene due to forest fire. The vertical red line separates forest fire cloudscene reflectance (CRE) with higher reflectance due to other surfaces. Futurework requires elaboration and analysis of the CRE values within range onthe right side from the red line in order to clarify the nature of the relativelyhigh reflectance.As a future development we work on RE methods and algorithms thatcan be used to distinguish water vapour/ice clouds from forest or wildfireclouds without weather forecast and image processing datasets. This canbe achieved, for example, by matching not only albedo but also enhanced14olumn of CO greenhouse gas concentration due to intense forest or fieldfires. In this work we generalize the RE method for detection of the cloud scenesto detection of the forest fires. This method can cover a wide spectral rangefrom 1100 nm to 1700 nm by using space observation datasets of Argus 1000micro-spectrometer. The RE method can be implemented for detection ofcombustion aerosols due to high sunlight reflectance of the smoke originatedfrom the forest fires. Our model accounts for the wavelength dependentreflectance and is developed by a new method based on the weightage sum.As the nanosatellite rotates periodically around the Earth, the proposedapproach may be a promising technique for continuous monitor of dynamicsof the seasonal forest fires.
Acknowledgments
This study is supported by Department of Physics and Astronomy at YorkUniversity, Epic College of Technology, Epic Climate Green (ECG) Inc. andThoth Technologies Inc. The authors would like to express their gratitudeto Dr. Robert Zee and his team from University of Toronto Institute forAerospace Studies for support, guidance and suggestions in operating of theCanX-2 spacecraft.
References [1] Apadula F, Cassardo C, Ferrarese S, Heltai D, Lanza A. Thirty years ofatmospheric CO observations at the Plateau Rosa Station, Italy. Atm2019;10; 418. https://doi.org/10.3390/atmos10070418 [2] Karnauskas KB, Miller SL, Schapiro AC. Fossil fuel combustion is driv-ing indoor CO toward levels harmful to human cognition. GeoHealth2020;4(5);e2019GH000237. https://doi.org/10.1029/2019GH000237 fertilization effect based on global monitor-ing land-atmosphere exchange with a theoretical model. Environ Res Lett2020;15;084009. https://doi.org/10.1088/1748-9326/ab79e5 [4] Scarpa G, Berrang-Ford L, Zavaleta-Cortijo C, Marshall L, Harper SL,Cade JE. The effect of climatic factors on nutrients in foods: evidencefrom a systematic map. Environ. Res. Lett. 2020;15;113002. https://doi.org/10.1088/1748-9326/abafd4 [5] Guo F, Shen X, Zou L, Ren Y, Qin Y, Wang X, Wu J. Cloud detec-tion method based on spectral area ratios in MODIS data. Canadian JRemote Sens 2015;41(6);561–76. https://doi.org/10.1080/07038992.2015.1112729 [6] Fournier N, Stammes P, Graaf MD, Piters A, Grzegorski M, KokhanovskyA. Improving cloud information over deserts from SCIAMACHY oxygenA-band measurements’, Atm Chem Phys 2006;6;163–72. https://doi.org/10.5194/acp-6-163-2006 [7] Mitchell DL, Finnegan W. Modification of cirrus clouds to reduce globalwarming. Environ Res Lett 2009;4(4);045102. https://doi.org/10.1088/1748-9326/4/4/045102 [8] Siddiqui R, Jagpal R, Salem NA, Quine BM. Classification of cloud scenesby Argus spectral data. Int J Space Sci Eng 2015;3(4);295–311. https://doi.org/10.1504/IJSPACESE.2015.075911 [9] Tymstra C, Stocks BJ, Cai X, Flannigan MD. Wildfire managementin Canada: Review, challenges and opportunities. Prog Disast Sci2020;5;100045. http://dx.doi.org/10.1016/j.pdisas.2019.100045 [10] Ganteaume A, Camia A, Jappiot M, San-Miguel-Ayanz J, Long-FournelM, Lampin C, A review of the main driving factors of forest fire ig-nition over Europe, Environ Manag 2013;51;651–62. https://doi.org/10.1007/s00267-012-9961-z https://doi.org/10.1016/j.foreco.2009.01.047 [12] Parisien M-A, Barber QE, Hirsch KG, Stockdale CA, Erni S, WangX, Arseneault D, Parks SA. Fire deficit increases wildfire risk for manycommunities in the Canadian boreal forest. Natur Commun 2020;11;2121. https://doi.org/10.1038/s41467-020-15961-y [13] Tymstr C, Stocks BJ, Cai X, Flannigan MD. Wildfire managementin Canada: Review, challenges and opportunities. Prog Disast Sci2020;5;100045. https://dx.doi.org/10.1016/j.pdisas.2019.100045 [14] Holm SM, Miller MD, Balmes JR. Health effects of wildfire smokein children and public health tools: a narrative review. J Expo-sure Sci Environ Epidem 2021;31;1–20. https://doi.org/10.1038/s41370-020-00267-4 [15] Wind G, Platnick S, King MD, Hubanks PA, Pavolonis MJ, Hei-dinger AK, Yang P, Baum BA. Multilayer cloud detection with theMODIS near-infrared water vapor absorption band. J Appl Meteor Cli-mat 2010;58;10;2315–33. https://doi.org/10.1175/2010JAMC2364.1 [16] Tang H, Yu K, Hagolle O, Jiang K, Geng X, Zhao Y. A cloud detectionmethod based on a time series of MODIS surface reflectance images.Int J Digit Earth 2013;6;Suppl. 1;157–71. https://doi.org/10.1080/17538947.2013.833313 [17] Siddiqui R, Jagpal R, Quine BM. Short wave upwelling radiative flux(SWupRF) within near infrared (NIR) wavelength bands of O , H O , CO and CH by Argus 1000 along with GENSPECT line-by-line radiativetransfer model. Canadian J Rem Sens 2017;43(4);330–44. https://doi.org/10.1080/07038992.2017.1346467 [18] Siddiqui R. Efficient detection of cloud scenes by a space-orbiting Argus1000 micro-spectrometer, PhD Thesis, York University, Toronto, Canada2017. 1719] Siddiqui RK, Jagpal RK, Abrarov SM, Quine BM. Radiance enhance-ment and shortwave upwelling radiative flux methods for efficient de-tection of cloud scenes. Int J Space Sci Eng 2020;6(1);1–27. https://dx.doi.org/10.1504/IJSPACESE.2020.109745 [20] Rankin D, Kekez DD, Zee RE, Pranajaya FM, Foisy DG, Beat-tie AM. The CanX-2 nanosatellite: expanding the science abilities ofnanosatellites. Acta Astronautica 2005;57(2–8);167–74. https://doi.org/10.1016/j.actaastro.2005.03.032 [21] Jagpal RK. Calibration and Validation of Argus 1000 Spectrometer –a Canadian Pollution Monitor. PhD Thesis, York University, Toronto,Canada 2011.[22] Jagpal RK, Quine BM, Chesser H, Abrarov S, Lee R. Calibration andin-orbit performance of the Argus 1000 spectrometer – the Canadianpollution monitor. J Appl Rem Sens Lett 2010;4(1);049501. https://doi.org/10.1117/1.3302405 [23] Jagpal RK, Siddiqui R, Abrarov SM, Quine BM. Carbon dioxide re-trieval of Argus 1000 space data by using GENSPECT line-by-lineradiative transfer model. Environ Natur Resour Res 2019;9(3);77–85. https://doi.org/10.5539/enrr.v9n3p77 [24] Siddiqui R, Jagpal RK, Abrarov SM, Quine BM. A new approach todetect combustion-originated aerosols by using a cloud method. AGUFall Meeting 2020, OAE, Dubai.[25] Natural Resources Canada. [26] Alkhatib AAA. A review on forest fire detection techniques. Int. J. Dist.Sensor Networks 2014;10(3);597368. https://doi.org/10.1155/2014/597368 [27] Wang SD, Miao LL, Peng GX. An improved algorithm for forest firedetection using HJ data. Environ. Sci. 2013;13;140–50. https://doi.org/10.1016/j.proenv.2012.01.014 [28] Benali A, Russo A, S´a ACL, Pinto RMS, Price O, Koutsias N,Pereira JMC. Determining fire dates and locating ignition points with18atellite data. Remote Sens 2016;8(4);326. https://doi.org/10.3390/rs8040326 [29] Quine BM, Drummond JR. GENSPECT: a line-by-line code with se-lectable interpolation error tolerance. J Quant Spectrosc Radiat Transfer2002;74;147–65. https://doi.org/10.1016/S0022-4073(01)00193-5 [30] Hill C, Gordon IE, Kochanov RV, Barrett L, Wilzewski JS, RothmanLS. HITRANonline: An online interface and the flexible representation ofspectroscopic data in the HITRAN database. J Quant Spectrosc RadiatTransfer 2016;177(4);4–14. https://doi.org/10.1016/j.jqsrt.2015.12.012 [31] Abrarov SM, Quine BM. Siddiqui R, Jagpal RK. A single-domain im-plementation of the Voigt/complex error function by vectorized interpo-lation. Earth Science Res 2019;8(2);52–63. https://doi.org/10.5539/esr.v8n2p52 [32] Abrarov SM, Quine BM. Efficient algorithmic implementation of theVoigt/complex error function based on exponential series approximation,Appl Math Comput 2011;218(5);1894–902. https://doi.org/10.1016/j.amc.2011.06.072 [33] Abrarov SM, Quine BM, Jagpal RK. A sampling-based approxima-tion of the complex error function and its implementation withoutpoles, Appl Numer Math 2018;129;181–91. https://doi.org/10.1016/j.apnum.2018.03.009 [34] Abrarov SM, Quine BM. A rational approximation of the Dawson’s in-tegral for efficient computation of the complex error function. Appl MathComput 2018;321(15);526–43. https://doi.org/10.1016/j.amc.2017.10.032 [35] Fomin BA. Effective interpolation technique for line-by-line calculationsof radiation absorption in gases. J Quantit Spectrosc Radiat Transfer1995;53(6);663–9. https://doi.org/10.1016/0022-4073(95)00029-K [36] Sparks L. Efficient line-by-line calculation of absorption coefficientsto high numerical accuracy. J Quantit Spectrosc Radiat Trans-fer 1997;57(5);631–50. https://doi.org/10.1016/S0022-4073(96)00154-9 https://doi.org/10.5194/acp-13-3133-2013 [38] Li S, Suna D, Goldberg MD, Sjoberg B, Santek D, Hoffman JP, De-Weese M, Restrepo P, Lindsey S, Holloway E. Automatic near real-time flood detection using Suomi-NPP/VIIRS data. Rem Sens Environ2018;204;672–89. https://dx.doi.org/10.1016/j.rse.2017.09.032 [39] MODIS land. https://modis-land.gsfc.nasa.gov/ [40] Baldridge AM, Hook SJ, Grove CI, Rivera G. The ASTER spectrallibrary version 2.0. Rem Sens Environ 2009;113;711–15. https://doi.org/10.1016/j.rse.2008.11.007 [41] Christopher SA, Gupta P. Satellite remote sensing of particulate mat-ter air quality: the cloud-cover problem. J Air Waste Manag Assoc2010;60(5);596–602. https://doi.org/10.1016/j.atmosenv.2006.03.016 [42] Systems Tool Kit (STK). [43] Google Earth. [44] EOSDIS Worldview. https://worldview.earthdata.nasa.gov/https://worldview.earthdata.nasa.gov/