Author's personal copy ARTICLE IN PRESSlar.wsu.edu/docs/2007_Jain_ClearSky_final_AtmEnv.pdf · Author's personal copy Atmospheric Environment 41 (2007) 6745 6761 Development of the

This article was published in an Elsevier journal. The attached copyis furnished to the author for non-commercial research and

education use, including for instruction at the author’s institution,sharing with colleagues and providing to institution administration.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copy

Atmospheric Environment 41 (2007) 6745–6761

Development of the ClearSky smoke dispersion forecast systemfor agricultural field burning in the Pacific Northwest

Rahul Jainc, Joseph Vaughana, Kyle Heitkampd, Charleston Ramosa,Candis Claiborna, Maarten Schreuderb, Mark Schaaf b, Brian Lamba,�

aLaboratory for Atmospheric Research, Department of Civil and Environmental Engineering, Washington State University,

Pullman, WA 99164-2910, USAbAir Sciences, Inc., 421 SW 6th Avenue, Suite 1400, Portland, OR 97204, USA

cGolder Associates, Calgary, BC, CanadadGeomatrix, Inc., 19203 36th Avenue West, Suite 101, Lynnwood, WA 98036-5772, USA

Received 19 December 2006; received in revised form 25 April 2007; accepted 25 April 2007

Abstract

The post-harvest burning of agricultural fields is commonly used to dispose of crop residue and provide other desired

services such as pest control. Despite careful regulation of burning, smoke plumes from field burning in the Pacific

Northwest commonly degrade air quality, particularly for rural populations. In this paper, ClearSky, a numerical smoke

dispersion forecast system for agricultural field burning that was developed to support smoke management in the Inland

Pacific Northwest, is described. ClearSky began operation during the summer through fall burn season of 2002 and

continues to the present. ClearSky utilizes Mesoscale Meteorological Model version 5 (MM5v3) forecasts from the

University of Washington, data on agricultural fields, a web-based user interface for defining burn scenarios, the

Lagrangian CALPUFF dispersion model and web-served animations of plume forecasts. The ClearSky system employs a

unique hybrid source configuration, which treats the flaming portion of a field as a buoyant line source and the smoldering

portion of the field as a buoyant area source. Limited field observations show that this hybrid approach yields reasonable

plume rise estimates using source parameters derived from recent field burning emission field studies. The performance of

this modeling system was evaluated for 2003 by comparing forecast meteorology against meteorological observations, and

comparing model-predicted hourly averaged PM2.5 concentrations against observations. Examples from this evaluation

illustrate that while the ClearSky system can accurately predict PM2.5 surface concentrations due to field burning, the

overall model performance depends strongly on meteorological forecast error. Statistical evaluation of the meteorological

forecast at seven surface stations indicates a strong relationship between topographical complexity near the station and

absolute wind direction error with wind direction errors increasing from approximately 201 for sites in open areas to 701 or

more for sites in very complex terrain. The analysis also showed some days with good forecast meteorology with absolute

mean error in wind direction less than 301 when ClearSky correctly predicted PM2.5 surface concentrations at receptors

affected by field burns. On several other days with similar levels of wind direction error the model did not predict apparent

plume impacts. In most of these cases, there were no reported burns in the vicinity of the monitor and, thus, it appeared

that other, non-reported burns were responsible for the apparent plume impact at the monitoring site. These cases do not

provide information on the performance of the model, but rather indicate that further work is needed to identify all burns

ARTICLE IN PRESS

www.elsevier.com/locate/atmosenv

1352-2310/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.atmosenv.2007.04.058

�Corresponding author.

E-mail address: [email protected] (B. Lamb).


and to improve burn reports in an accurate and timely manner. There were also a number of days with wind direction

errors exceeding 701 when the forecast system did not correctly predict plume behavior.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: CALPUFF; PM2.5; Biomass burning; Air quality model

1. Introduction

Humans have used fire in a variety of slash-and-burn cultivation and hunting schemes that bothpresaged and co-exist with agricultural burning aspracticed today (Pyne, 1982; Crutzen and Andreae,1990).

Air-quality and climate issues associated withagricultural field burning, modern or traditional, andcorrelates in forestry and range management havebeen investigated at a variety of scales: laboratory-scale burns, local case studies of burns of specificbiomass fuels, regional multi-investigator surface andaircraft campaigns, remotely sensed satellite imageryanalysis for regional or global data acquisition, globalemission inventories and global modeling experimentsfor various purposes. Instrumental and modelingapproaches to studying biomass-burning emissionsvary enormously with the scale of the investigationand the issues being addressed.

In this paper, we briefly review the state of sciencerelating to biomass burning and pyrogenic emis-sions, particularly relating to the somewhat lessstudied types of biomass burning such as cropresidue burning, and describe a novel modelingsystem called ClearSky for support of air qualityprotection at local to regional scales. The modelingsystem is described in Section 2 and the results ofClearSky 2003 performance evaluation are pre-sented in Section 3. The paper concludes with asummary in Section 4.

1.1. Laboratory, local and regional studies

Jenkins et al. (1996) performed laboratory burnsfor four-cereal crop residue and four forest fuels in awind tunnel, analyzing for 19 polycyclic aromatichydrocarbon (PAH) species. Up to 70% of the PAHwere found to partition into the particulate state,with PAH accounting for a particulate massfraction of up to 4000 [mg kg�1]. Christian et al.(2003) measured gaseous and particulate emissionfactors for some 16 biomass fuels, commonlyburned annually to biennially in Africa, Asia,Europe or North America. These fuels were burned

in a laboratory burn hut and characterized interms of elemental composition, emission ratios[mmolmol�1 CO2] and emission factors for avariety of gas and aerosol products. Emissions werecharacterized using open-path Fourier transforminfrared (OP-FTIR) spectroscopy, proton transferreaction mass spectroscopy (PTR-MS) and severalgas chromatography techniques including massspectroscopy (GC/MS), electron capture detector(GC/ECD), and flame ionization detector (GC/FID). The emission factors were also related to themodified combustion efficiency (MCE ¼ DCO2/(DCO+DCO2)) (Ward, 2001) through linear regres-sion with R2 of �0.5–0.9. In this laboratory workChristian et al. (2003) demonstrated quantificationof emitted species not found in previous, relatedfield investigations, specifically certain oxygenatedvolatile organic compounds (OVOC) determinedusing PTR-MS and OP-FTIR. Laboratory-scalestudies are important, in part, in establishingviability of techniques for making specific measure-ments in the field.

While laboratory studies such as cited aboveprovide guidance on the emissions expected inactual fires, field observations on a local scale area necessary corollary. For example, Daniels (1987)observed that burn hut tests failed to reproduce theelemental particulate distribution observed in thefield, concluding that actual fires of high intensityentrained considerable dust in the smoke. Thus, thelocal scale is where the complex interaction ofecosystem, climate, meteorology and fire dynamicscan be observed most concisely.

Pollutants released during open biomass burninginclude particulate matter (PM), particulate matterwith aerodynamic diameter equal to or less than10 mm (PM10), particulate matter with aerodynamicdiameter equal to or less than 2.5 mm (PM2.5), CO,CO2, CH4 and non-CH4 hydrocarbons (NMHC)(Crutzen and Andreae, 1990). The focus of currentmodeling work is PM2.5. Table 1 summarizes theemission factors for PM2.5 EFPM2:5

� �determined

from several local-scale studies.Air Sciences Inc. (2003) evaluated combustion

efficiency, fuel, soil moisture and emission factors

ARTICLE IN PRESSR. Jain et al. / Atmospheric Environment 41 (2007) 6745–67616746


for CO, CO2, CH4 and PM2.5 for some 26 wheatstubble burns in Columbia County in WashingtonState. While a moderate correlation (R2

¼ 0.61) wasreported between EFPM2:5 and combustion efficiency(approximately MCE), the strongest correlationnoted was between EFPM2:5 and residue moisturewith R2 ranging between 0.50 and 0.79.

At a regional-scale biomass burning effectsinclude contribution to haze, emissions of precur-sors that contribute to oxidant episodes, andindirect atmospheric effects such as modifyingradiative balance and altering precipitation out-comes. Ward et al. (1992) used the automated FireAtmosphere Sampling System (FASS) to capturegas and particulate emissions within biomass fireevents for a range of Brazilian vegetative commu-nities. Results for EFPM2:5 ranged from �5 to 20[g kg�1 C] for the various fuels and combustionstages.

Scholes et al. (1996) and Barbosa et al. (1999)used AVHRR imagery and other data to mapvegetation density and fire occurrence for Africa.Barbosa et al. (1999) estimated as 704–2168 Tgannual biomass consumed. The emissions of PM2.5

were estimated at 3.3–12.4 Tg annually. Hely et al.(2003) used satellite-derived data to calculate likelyemissions (during the SAFARI 2000 intensive) forseveral species for South Africa. For �31,000SPOT-detected fires, emissions were estimated at96.9 Tg CO2, 4.6 Tg CO and 0.3 Tg PM2.5. Tanseyet al. (2004) produced a global SPOT-derived 1-kmresolution burned area inventory for the year 2000.

Dennis et al. (2002) reported that of the of the�200Mg yr�1 calculated as total Texas PM2.5

emissions, prescribed rangeland burning accountedfor �52% and agricultural burning for �32%.Sinha et al. (2003) determined EF for 50 gaseousand particulate species from flights on 11 late dryseason savanna fires in southern African, finding a

EFTPM (total PM defined as Dpp4 mm) of 10.077.5[g kg�1 C].

1.2. Agricultural field burning in the Pacific

Northwest

Many farmers in the Inland Pacific Northwestregion of the United States burn crop residue in thefields after harvest. However, unlike most localitieswhere such burning occurs, in this region publicsmoke exposure has led to legal and regulatorypressure to curtail such burning. The ClearSkysystem was developed in response to the perceivedneed for better guidance for farmers as to whenburning is inadvisable because of likely humanexposure to smoke. A distinct yet allied project, theBlueSky/RAINS system of the USDA/Forest Ser-vice has a similar mission with respect to forestburning (O’Neill et al., 2003). Both of these systemsexplore the potential for modeling of pyrogenicaerosol emissions as input into environmentalmanagement, possibly linking locally focused deci-sion-making across jurisdictional lines and even-tually rising to the level of regional coordination inair-quality protection for human health.

Agricultural field burning is an inexpensive meansused to remove crop residue, thereby reducing theuse of chemicals for combating plant diseases,insects and weeds, and thereby maintaining cropyield (Hardison, 1980; Gray and Guthrie, 1977;Higgins et al., 1989; Jenkins et al., 1996; Mast et al.,1984; Cook and Haglund, 1991). In the absence offield burning, the crop residue remains on the soilsurface and creates a cooler and wetter environmentin the upper layers of the soil profile, making the soilmore hospitable to development of pathogens. Incontrast, the soil surface made bare by burningtends to warm and dry faster, creating a soilenvironment less favorable for the development ofdiseases. Bare soil tends to freeze sooner, deeper andmore frequently in fall and early winter than soilthat is protected by surface residues. Cook andHaglund (1991) demonstrated that yields are betterthe following year, when the wheat seed is directlysown into the burned stubble of a previous wheatcrop or bluegrass sod than when the stubble is leftstanding and not burned prior to planting. Theyattributed the increase in the yield to reduction inroot diseases. However, Chastain et al. (1998) havedissenting views of the superiority of residueburning. They conducted an investigation to deter-mine the relationship between post-harvest residue

ARTICLE IN PRESS

Table 1

PM2.5 emission factors determined by local-scale studies

Fuel type Emission factor PM2.5 [lb ton�1]

Wheat stubbleb 5.370.5 (low fuel moisture)a

9.871.1 (high fuel moisture)a

Kentucky blue grass stubblec 66.0

aAverage of backing fire, head fire and strip head fire

emissions.bAir Sciences Inc. (2003).cJohnston and Golob (2004).

R. Jain et al. / Atmospheric Environment 41 (2007) 6745–6761 6747


management and stand age effect on cool seasonperennial grasses. This investigation concluded thatseed germination level and seed purity were notconsistently related to age of crop stand or theresidue management practice adopted.

As a standard practice, Kentucky bluegrass(KBG) seed fields and wheat are harvested in Julyand early August and the crop residue is burnedfrom August to September (Mazzola et al., 1997).During field burning months (primarily August andSeptember), annually, thousands of acres (morethan 100,000 acres in northern Idaho and easternWashington combined) of agricultural fields (pri-marily KBG and Wheat) are burned in northernIdaho and eastern Washington. Burn coordinatorsrepresenting various state and regional environmentprotection regulatory agencies plan and coordinatethe prescribed burning of the fields. The burncoordinators use forecast meteorology to identifyburn windows during the day when the farmers mayburn their fields. The farmers are allowed toconduct permitted field burns only if the forecastmeteorology suggests adequate dispersion of smokeplumes away from the population centers.

In spite of the caution exercised by the burncoordinators in giving out burn permits, high PMconcentrations have been recorded when smokeplumes from the field burning have hit urbanlocalities. Exposure to these high PM (particularlyPM2.5) concentrations poses a potential risk tohuman health and is also a public nuisance.

2. Description of ClearSky

ClearSky is an automated, interactive, web-basedmodeling system developed to support smokemanagers in making burn/no-burn decisions regu-lating agricultural field burning. Jurisdictions sup-ported by the project include several countiesin northern Idaho, including the Coeur d’Aleneand Nez Perce tribal reservations, and easternWashington (Fig. 1). ClearSky consists of threecomponents: a web-based field-burning scenariogenerator, an automated CALPUFF modelingsystem, and a web-based application for reviewinganimations of CALPUFF results for burn scenar-ios. A schematic of the ClearSky modeling system isshown in Fig. 2.

ARTICLE IN PRESS

Fig. 1. Location of air quality monitoring stations in Northern Idaho and Eastern Washington (counties with air quality monitoring

stations are marked ) equipped with (left) Nephelometer and (right) TEOM PM2.5 instruments. 1. Sandpoint, 2. Athol, 3. Rathdrum,

4. Meyer Ranch, 5. Post Falls, 6. Coeur d‘ Alene 7. Fighting Creek, 8. Pinehurst, 9. Moscow, 10. Lewiston, 11. Grangeville.

R. Jain et al. / Atmospheric Environment 41 (2007) 6745–67616748


The project website provides registered users withaccess to a web-based field-burning scenario gen-erator tailored to their jurisdiction. The scenariogenerator references a database of field acreage,location, crop type, etc., to provide the user withchoices of available fields to burn in a hypotheticalburn scenario for the next day. The databases aremaintained by the user agencies and communicatedto update the ClearSky system as often as deemednecessary by the agencies concerned. Each scenarioof one to multiple fields has burn acreages and burntimes (ignition and fire-out times) set by the user,within limits. Upon submission, each scenarioreceives a unique scenario identifier that encodestotal acreage, field count, jurisdiction, or similarinformation. In addition to the user-submittedscenarios, default scenarios are defined for eachjurisdiction so that results will be available for userreview, regardless of the user’s ability to build andsubmit a next day scenario in a timely manner.Multiple burn scenarios can be generated withineach jurisdiction.

The ClearSky system automates meteorologicaldata preparation, emissions preparation, CAL-PUFF execution and presentation of the resultsvia the password-protected project website.

ClearSky uses daily numerical weather predic-tions from the University of Washington Depart-ment of Atmospheric Science (UW) mesoscale-modeling group (Mass et al., 2003). UW uses theMesoscale Meteorological Model version 5 (MM5)to produce runs for nested domains with grid scalesof 36, 12 and 4-km, with the outermost 36-km

model run being initialized at 00 UTZ (GreenwichMean Time) and 12 UTZ each day. ClearSkycurrently uses forecast hours 12 to 36 from a 4-kmsimulation nested within a 12-km run initialized at00 UTZ, resulting in a simulation from 4 AM PSTto 4 AM 1 day later.

The CALMET model (Scire et al., 2000) incor-porates the diagnostic wind field model (Douglasand Kessler, 1988) to develop wind fields for themodeling domain. The MM5 4-km model output isingested into the CALMET model as pseudo-observations with one MM5 data point availablefor each grid cell. No terrain corrections are madewithin CALMET but a 3-day divergence minimiza-tion calculation yields a mass consistent 3-day windfield. The MCIP meteorological processor (Byun andChing, 1999) is also run with MM5 input to make theMM5 boundary layer terms, such as mixing heightand friction velocity, available without CALMETrecalculation. The CALMET 3-day wind fields andMCIP boundary-layer terms are then combined into afinal CALMET output file. The CALMET domainfor ClearSky (Fig. 3) consists of 180 columns (E–W)by 121 rows (N–S) of 4-km grid cells, by 13 verticallayers of variable spacing. These layers have levelheights of z ¼ 20, 36, 73, 109, 146, 220, 295, 408, 523,677, 1034, 1746, 3451, where z is a terrain followingelevation in meters. The southwest corner of thedomain is at �302.0km easting, �422.0km northingin Lambert conic conformal coordinates, relative toan origin at 1211W, 491N.

Scenarios defining field burning are processed togenerate the emissions files required to run CALPUFF.

ARTICLE IN PRESS

MM5 CALMET CALPUFF

u, v 3-d fields

3D met field

u, v, w, T, :

BL variables

2D PM2.5

concentration file

IC/BC

landuse

terrain

landuse

terrain

PM2.5

emissions

scenarios

Fig. 2. Flow diagram for the ClearSky smoke dispersion forecast system. The modeling system employs a 4-km� 4-km horizontal grid for

meteorological input and 1-km� 1-km horizontal grid for receptor sampling.



The ClearSky modeling system simulates thesmoke plumes from field burning by using the puffdispersion model, CALPUFF, in a forecast mode.CALPUFF is a multi-layer, multi-species, non-steady-state Gaussian Lagrangian puff dispersionmodel that simulates the effect meteorologicalconditions that vary in time and space on pollutanttransport, transformation and removal.

CALPUFF simulates a smoke plume as a seriesof Gaussian puffs; concentrations at each receptorare calculated for each puff for each time step fromthe Gaussian puff dispersion equation. The totalconcentration at a receptor is the sum of contribu-tions from nearby puffs averaged over the samplingsteps within each hourly time step (Scire et al.,1999). As used by ClearSky, CALPUFF simulatesthe hourly average PM2.5 concentrations due to allfield burns within a scenario for the entire domain;each scenario requires a separate CALPUFF run.

The web-based scenario-results review systemallows users to view CALPUFF PM2.5 animationsfor each of their jurisdiction’s scenarios, includingthe default scenarios. These animations are gener-ated from GIF files produced using PAVE, avisualization package available through theCommunity Modeling and Analysis System (www.cmascenter.org), and viewed via a flexible animator.

2.1. Hybrid source and plume rise simulation in

ClearSky

There are four types of agricultural field burning:head fire, mass ignition, strip head fire, and backingfire. In each case, ignition is typically conductedusing drip torches carried on all terrain vehicles. Ahead fire is one that is ignited at the upwind edge ofthe unit to be burned and pushed across the unit bythe wind. Head fires are usually fast. Mass ignitionis a variation of the head fire technique. With thistechnique, the unit to be burned is engulfed by fireas quickly as possible. A backing fire is the oppositeof a head fire. It is one that is ignited on thedownwind edge of the unit to be burned. A striphead fire is another variation of the head fire inwhich the field to be burned is ignited in strips,starting at the downwind side of the unit to beburned and proceeding upwind. Each strip’s flamefront runs into the previously burned strip, whichcauses it to be extinguished (Air Sciences Inc.,2003).

ClearSky models field burning as a head fire.Head fires have been observed to have a flamingfront followed by a smoldering area. A new hybridapproach which models the flaming front of the fireas a buoyant line source and the smoldering part of

ARTICLE IN PRESS

Fig. 3. Terrain elevation in the Pacific Northwest domain for automated MM5/CALMET/CALPUFF forecast operations. ClearSky was

operated in 2003 for the area enclosed within the rectangle.



the fire as a buoyant area source was adopted tomodel agricultural field burning.

2.2. Evaluation of the hybrid source and plume rise

treatment in ClearSky

A sensitivity analysis was carried out to studyplume rise treatment within CALPUFF. Keepingthe area of the field constant, the flaming stage ofthe field burn was modeled both as an area sourceand as a line source. Plume rise treatment inCALPUFF, for a line source and an area source,is a function of initial plume temperature, ambientwind speed and plume exit velocity. By varying eachof the above-mentioned parameters systematically,we were able to compare the plume rise froman area source to the plume rise from a line source.A set of wind speeds defined for each layer mid-point was designated as Ub, the base case windprofile. As part of the sensitivity analysis, this basecase wind profile was varied over a range of 0.5Ub

to 2Ub.Results from the sensitivity analysis are summar-

ized in Fig. 4 in terms of final plume rise as afunction of exit velocity, source temperature andambient wind speed (in terms of the multiplier of thebase wind speed profile). The sensitivity analysisclearly demonstrates that the plume rise for a givensource can vary significantly depending on thesource definition in CALPUFF. For a givenambient wind speed and source temperature, treat-ment as a line source results in lower final plumeheight compared to treatment as an area source.The area source treatment shows greater sensitivityof the plume rise to initial plume temperaturecompared to the line source treatment; and themaximum plume rise occurs for both treatmentswhen the plume exit temperature is highest and theambient wind speed is lowest. For both treatments,plume rise increases with exit velocity. Generally,the line source plume rise is lower than the areasource plume rise using identical parameters.

In the Air Sciences field studies for wheat andKBG burns, emissions were estimated from mea-surements using a carbon mass balance method andthe temperature and updraft velocity of the plumewas also measured using the Fire AtmosphereSampling System (FASS) for the four types ofignition. The FASS apparatus measured the tem-perature and updraft velocity as a function of time(min), with t ¼ 0 when the flame front reached theapparatus.

ARTICLE IN PRESS

y = 95x + 248

R2 = 0.99

y = 61.6x + 187

R2 = 0.99

0

100

200

300

400

500

600

700

0

Exit Velocity (m/s)

Plu

me R

ise (

m)

Area Source

Line Source

Area Source

Line Source

Area Source

Line Source

y = 0.28x + 243

R2 = 0.89

y = 0.17x + 187

R2 = 0.90

0

100

200

300

400

500

600

700

300 500 700 900 1100 1300 1500

Source Temperature (K)

Plu

me R

ise (

m)

y = -159x + 527

R2 = 0.94

y = -91.9x + 354

R2 = 0.93

0

50

100

150

200

250

300

350

400

450

500

0 0.5 1 1.5 2 2.5

Wind Speed Profile Multiplier

Plu

me R

ise (

m)

1 2 3 4 5

Fig. 4. Sensitivity of buoyant area and line source plume rise to

changes in exit velocity (top), source temperature (middle), and

ambient wind speed (bottom). The axis for the wind speed graph

is in terms of multipliers of the base case wind speed vertical

profile (i.e. 0.5Ub, Ub, and 2Ub).



Figs. 5 and 6 show typical behavior for the fourdifferent fire types. Fig. 5 shows the updraft velocitywith respect to time for the four ignition types. Forall the types of ignition, it is common to have anincrease in updraft velocity in the flame front, att ¼ 0. In the smoldering part of the fire (tb0), onlythe strip head fire showed significant increase inupdraft velocity.

Fig. 6 shows the temperature measured by theFASS apparatus with respect to time for the fourdifferent ignition types. The plot shows that thehighest temperature is at the flame front of the fire.Also, the head fire had the highest increase intemperature while backing fire had the lowest. Thebacking fire shows the highest increase in tempera-ture in the smoldering part of the fire.

Given the range of exit velocities and tempera-tures in the different portions of the burn, it isworthwhile to use the results from the sensitivityanalysis to estimate the uncertainty in plume rise foran estimate of the uncertainty in specifying exitvelocity and source temperature. Given the slopes of

the regression equations in Fig. 4, the uncertaintiesin final plume rise for line and area sources are8–14m for an estimated uncertainty of 1m s�1 inexit velocity and 60–95m for an estimated un-certainty of 50K in source temperature. Uncertain-ties in the wind speed profile of 0.33 (i.e. 1 out of3m s�1) yield an uncertainty in plume rise of30–50m for the line and area sources. These arerelatively small uncertainties in plume rise com-pared to the total plume rise.

Based upon the Air Sciences results, plume riseparameters were selected for the flame front (buoy-ant line source) and smoldering (buoyant areasource) portions of the burn as given in Table 2.The initial line width and flame height were assumedfrom qualitative observations of burning fields.Based on information provided by the burncoordinators, the rate of field burn was set to 100acres per hour and applied uniformly to all regions.The crop residue burned in Washington is primarilywheat stubble and the crop residue burned innorthern Idaho is primarily KBG. The PM2.5

ARTICLE IN PRESS

Mass Ignition

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

-5 0 5 10 15

Time

Up

dra

ft V

elo

cit

y (

m/s

)

Backing Fire

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

-5 0 5 10 15

Time

Up

dra

ft V

elo

cit

y (

m/s

)

Strip Head Fire

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

-5 0 5 10 15

Time

Up

dra

ft V

elo

cit

y (

m/s

)

Head Fire

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

-5 0 5 10 15

Time

Up

dra

ft V

elo

cit

y (

m/s

)

Fig. 5. Effect of ignition type on updraft velocity.



emission factor for wheat stubble, for a heading fire,was taken as 3.6 g PM2.5 kg

�1 fuel (7.2 lb ton�1) or0.36% of the residue load; and the emission factorfor KBG was taken to be 30 g PM2.5 kg

�1 fuel(66 lb ton�1) or 3.3% (Air Sciences Inc., 2003 andJohnston and Golob, 2004). The ratio between the

flaming and smoldering emissions was assumed tobe 80:20. An estimated residue load of 6919 kg ha�1

(2.8 tons acre�1) was used for calculating the emis-sions.

To evaluate the hybrid source approach used inClearSky, a modest field campaign was completedduring the summer/fall burn season in 2004. Thefocus of the campaign was to collect plume risemeasurements from agricultural field burns andthen compare those plume rise measurements to theplume rise predicted in ClearSky.

Plume heights were measured during nine fieldburns over four days of the 2004 agricultural burnseason. The dates for the plume height measurementcampaign included July 30, August 20, September 8,and September 29, 2004. Four wheat stubble fieldburns were observed in eastern Washington and fiveKentucky bluegrass field burns were observed innorthern Idaho. Field sizes varied from �13 to64 ha (32–157 acres).

During each field burn, plume height measure-ments were taken utilizing an aircraft and also from

ARTICLE IN PRESS

Backing Fire

-5 0 5 10 15

Time

Strip Head Fire

-5 0 5 10 15

Time

Mass Ignition

-5 0 5 10 15

Time

Head Fire

0

20

40

60

80

100

120

140

160

-5 0 5 10 15

Time

Tem

pera

ture

(c)

0

20

40

60

80

100

120

140

160

Tem

pera

ture

(c)

0

20

40

60

80

100

120

140

160

Tem

pera

ture

(c)

0

20

40

60

80

100

120

140

160

Tem

pera

ture

(c)

Fig. 6. Effect of ignition type on source temperature.

Table 2

ClearSky hybrid source plume rise parameters

Parameter value

Buoyant area source (smoldering fraction)

Effective height of emissions (m) 0.5

Source temperature (K) 324

Effective exit velocity (m s�1) 1.4

Initial vertical spread (m) 100

Buoyant line source (flaming fraction)

Line height (m) 0.5

Source temperature (K) 361

Line width (along wind) (m) 5

Effective exit velocity (m s�1) 2.2



the ground. For this study, the top of plume heightswere measured because of the difficulty in determin-ing the plume centerline height. The air-crew madeobservations of top of plume height and verticaltemperature profiles, and took photographsthroughout each field burn. The ground-crewobserved the top of plume height using a simpleinclinometer and deployed instruments to measuresurface air temperature, surface wind direction andspeed, vertical profiles of wind speed and direction,and obtained field burn photographs. Residueloadings before and after the burn were alsocollected from each field.

The top of plume height measurements wereestimated from the aircraft using the aircraftaltimeter. As the pilot circled around the burningfield, another crew member watched to see when theplane altitude and the top of plume height wereequal, and then recorded the plane altimeter andtheir geographic location with a waypoint on ahandheld Global Positioning Satellites (GPS) unit.The time, altitude, and waypoint number werewritten on a data collection sheet.

Top of plume heights were measured from theground using a handheld clinometer and a handheldGPS unit. The clinometer was used to measure theelevation angle between the ground (zero degrees)and the top of the plume. A ground data collectionsheet was used to record the time and clinometerreadings. The geographic location of the surface

location used for plume height measurements wasdetermined using the GPS unit. These datapermitted calculation of the distance to the centerof the field from the surface location as well as thetrigonometric estimation of the top of plume height.This top of plume height measurement techniqueassumed that the plume only traveled verticallyfrom the center of the field.

Modeled plume heights from CALPUFF weredetermined by rerunning ClearSky for each fieldburn. CALPUFF was configured to produce aplume log detailing each puff released by thebuoyant line source and the buoyant area sourcethat simulated each field burn. CALPUFF calcu-lated centerline plume heights for the buoyant lineand buoyant area sources and also calculated thecorresponding Gaussian vertical dispersion coeffi-cient (sigma-z). The top of plume heights wereestimated by adding the value of 2.15 times thesigma-z value (characterizing the vertical dispersionof the plume) to the centerline plume height. Forbuoyant line sources, CALPUFF calculates only thefinal plume height and final plume distance values,while CALPUFF calculates the entire evolution ofplume height with distance for buoyant areasources. Top of plume heights estimated fromCALPUFF and top of plume heights measured bythe ground-crew were both expressed above groundlevel (AGL). The top of plume heights observedwith the aircraft altimeter had to be converted from

ARTICLE IN PRESS

Top of Plume Heights for CALPUFF Buoyant Area and Buoyant Line Source with Updated

Emissions Parameters and Maximum Air and Ground Observations

0

500

1000

1500

2000

2500

3000

3500

Day

1, F

ield 1

Day

1, F

ield 2

Day

2, F

ield 1

Day

2, F

ield 2

Day

2, F

ield 3

Day

3, F

ield 1

Day

3, F

ield 2

Day

4, F

ield 1

Day

4, F

ield 2

Burn Day

To

p o

f P

lum

e H

eig

ht

(m)

Buoyant Area Source Buoyant Line Source Max Air Observation Max Ground Observation

Fig. 7. Ranges of both CALPUFF plume heights and measured plume heights for each field burn.



above mean sea level (MSL) to AGL for compar-ison with estimated top of plume heights. Theaverage elevation of each burn site was determinedfrom a topographic map and used to convert MSLtop of plume heights to AGL.

Fig. 7 shows a summary of maximum top ofplume heights measured and final top of plumeheights modeled using the ClearSky hybrid emissionparameters. Because of the uncertainties in assign-ing top of plume heights, there is considerablescatter in the estimated plume heights. However,there is general agreement between the estimatedplume heights and the plume rise values calculatedfrom the CALPUFF model for buoyant line andarea sources using the source parameters estimatedfrom the Air Sciences field data.

3. Performance evaluation of the ClearSky modeling

system

There are a number of issues to consider in theevaluation of the ClearSky smoke plume forecastingsystem. First, operational model runs should not beevaluated directly since the emission scenarios aresubmitted as ‘‘potential’’ burns and often thecombination of fields actually burned is quitedifferent from a particular scenario. As a result,evaluation of the system requires compilation ofdata for actual burns and re-running the modelingsystem. Second, there is a relatively sparse networkof PM2.5 monitoring sites within the modelingdomain, and the goal of the burn managers is toburn fields under conditions that avoid impact ofsmoke in populated areas where most of themonitors are located. As a result, there are relativelyfew instances of measured PM2.5 concentrations dueto field burning, and there are almost no instanceswhere the plume from a single burn impacts morethan one monitor. In addition, there may beconfounding sources, such as forest wildfires orun-permitted burns such that PM2.5 observations ata monitoring site may reflect more than the intendedsources. Finally, because ClearSky models plumesfrom relatively small sources individually, errors inpredicted wind directions of just a few degrees cancause modeled plumes to completely miss a down-wind monitoring site for a period when the actualplume has a significant impact at the site.

With these considerations in mind, the ClearSkysystem was evaluated by re-running ClearSkyduring August–September 2003 for 20 burn daysselected when the total daily acreage burned was

more than 1000 acres (Table 3). Data from burnmanager documents were used to identify fields thathad been burned on each day and to prepare theemissions data files. Within Idaho, actual ignitiontimes were not recorded so the burn times wererandomly selected within a typical burn window forthe Idaho fires. A 4-km� 4-km horizontal meteor-ological grid and a nested 1-km� 1-km horizontalsampling grid were defined in CALPUFF for themodel evaluation runs. CALPUFF predicted PM2.5

concentrations were compared against observationsat twenty monitoring stations (shown in Fig. 1),defined as discrete receptors in CALPUFF.

As a first step in the evaluation, the meteorolo-gical performance of the MM5/CALMET portionof ClearSky was investigated using hourly predictedvalues for meteorological parameters (wind speed,wind direction and temperature) compared againstavailable surface station observations.

The model performance statistics are summarizedin Table 4 in increasing order for the mean absolutewind direction error for August–September burndays. The mean error in wind direction and windspeed and mean absolute error in wind speed arealso shown. There seems to be a relationshipbetween the complexity of the terrain near the

ARTICLE IN PRESS

Table 3

Burn days re-run for ClearSky evaluation and total number of

hectares (acres) burned on each of these days in northern Idaho

and eastern Washington counties

Date Northern

Idaho

Eastern

Washington

Total

ha ac ha ac ha ac

6-August 1040 2567 253 625 1293 3192

7-August 848 2094 0 0 848 2094

11-August 1092 2696 153 378 1245 3074

12-August 961 2374 71 176 1033 2550

13-August 385 950 36 88 420 1038

14-August 626 1546 116 286 742 1832

19-August 1144 2824 120 297 1264 3121

25-August 376 929 41 100 417 1029

26-August 1632 4030 130 322 1763 4352

3-September 659 1627 0 0 659 1627

5-September 738 1821 28 70 766 1891

11-September 1989 4910 770 1902 2759 6812

12-September 443 1094 1792 4425 2235 5519

15-September 3179 7849 11 26 3189 7875

16-September 936 2310 210 518 1145 2828

18-September 1097 2708 653 1612 1750 4320

19-September 1303 3218 617 1524 1921 4742

22-September 2188 5402 1209 2985 3397 8387

23-September 1900 4691 796 1965 2696 6656

25-September 1318 3254 599 1479 1917 4733



station and the magnitude of the absolute error inwind direction. For stations in relatively openterrain, such as Moscow, ID, the absolute error issmall, while for stations in deep valleys or near alake, such as Kamiah and Sandpoint, the absoluteerrors are large. This pattern also holds for meanerror and bias in wind direction, which implies thatthe modeling system does not account well for theeffects of local terrain on the overall wind patterns.

The bias in wind speed at most stations is smalland near 0.1m s�1 except for Rathdrum andBonners Ferry. The larger bias at these two stationscannot be readily explained. The absolute error inwind speed seems to be bi-modal with results near

1m s�1 (at five stations) and near 2m s�1 (atRathdrum, Sandpoint, Bonners Ferry and Spo-kane). Overall, the absolute mean error in winddirection was 601 and the absolute mean error inwind speed was 1.3m s�1.

In spite of the errors in forecast meteorology,there were a number of cases where PM2.5

concentrations attributed to field burning wereaccurately modeled with the ClearSky system. Forexample, on August 25, a field burn in northernIdaho produced a plume that was modeled to traveldirectly over the Athol monitoring site as shown inFig. 8. For this case, the modeled and observedPM2.5 surface concentrations were almost an exact

ARTICLE IN PRESS

Table 4

Summary of surface meteorological evaluation results for August–September burn days with stations ranked by absolute mean wind

direction (WD) error

Station Local terrain Mean absolute

WD error (deg.)

Mean WD error

(deg.)

Mean absolute

WS error (m s�1)

Mean WS error

(m s�1)

Moscow, ID Open, hilly 35 �3 1.0 0.1

Rathdrum, ID Open, flat 45 �3 1.9 �1.3

Reubens, ID Open, flat 49 9 1.1 �0.5

Lewiston, ID Open, upper broad valley 54 �17 0.9 0.2

Sandpoint, ID Near lake 57 �5 2.0 0.1

Bonners Ferry, ID Valley 78 �54 1.7 �1.0

Spokane, WA Open, near valley 79 6 2.0 �1.6

Fig. 8. Observed and predicted PM2.5 concentrations (left) at the Athol monitoring site in northern Idaho and the corresponding predicted

plume concentration contours (right, monitoring site is the small square) during 25 August 2006.



match both in time and magnitude as shown inFig. 8. Similar results were also obtained on August26 when the forecast meteorology was in goodagreement with observations as shown in Fig. 9 andthe modeled plume impact at the Pinehurst mon-itoring site in northeastern Idaho was a close matchto the observed concentrations shown in Fig. 10. Inthis case, the modeled peak concentration occurredan hour before the measured peak, but this can beattributed to the uncertainty in the actual fieldignition times. We also find cases such as shown inFig. 11 where the observed concentration pattern isbounded by modeled time series taken from pointswithin 1, 2, 3, and 4 km of the actual monitoringsite. In other cases (not shown), modeled plumescompletely missed the receptor, but the centerlineconcentrations were in close agreement with theobserved maximum concentrations. The cases where

the modeled concentrations showed a good matchwith observations suggests that the hybrid sourceapproach and the associated source parameters arereasonably accurate and capable of correctly pre-dicting surface PM2.5 concentrations.

These results show that on some days the modelperformed much better than on other days. Basedon the meteorological performance of the model oneach burn day the evaluation results can beclassified in four categories as summarized inTable 5. In the first category, when the predictedmeteorology is in good agreement with the observedwinds, ClearSky modeled plume impact from fieldburning at several monitoring stations where plumeimpact consistent with field burning was observed.The average absolute bias in wind direction forthese 8 days was approximately 131 and the averageabsolute error was 291. For these cases, the observed

ARTICLE IN PRESS

Fig. 9. Observed and predicted surface winds for two sites on 26 August 2003.



plumes did not always pass over the monitoring site,but sometimes missed the monitoring site by severalkilometers. However, the modeled centerline con-centrations for many of these cases were very closeto the maximum concentrations observed at themonitoring site. This match between the observedmaximum concentrations and modeled centerline

concentrations suggest the emission and plume riseparameters selected for ClearSky are reasonablyaccurate.

In the second category were four burn days whenthere was poor agreement between the modeledmeteorology and observed meteorology; for thesedays ClearSky did not model plume impact from

ARTICLE IN PRESS

0

5

10

15

20

25

30

35

40

45

9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00

PM

2.5

(u

g/m

3)

Local Time

Predicted

Observed

Fig. 10. Observed and predicted PM2.5 surface concentrations at the Pinehurst monitoring site in northeastern ID during 26 August 2003.

0

5

10

15

20

25

30

35

40

9:00 AM 12:00 PM 3:00 PM 6:00 PM 9:00 PM

Time (P.S.T.)

Co

ncen

trati

on

ug

/m3

Receptor Observed

1-km 2-km

3-km 4-km

Fig. 11. Observed and predicted PM2.5 concentrations at the Rathdrum site on 19 August 2003. Maximum concentrations predicted

within 1, 2, 3, and 4 km of the monitoring site are also shown.



field burning at monitoring stations where plumeimpact consistent with field burning was observed.For these days, the average absolute bias in winddirection was 721 and the average absolute error inwind direction was 811.

In the third category, on 6 days the modeledmeteorology was in good agreement with theobserved winds, but ClearSky did not predict plumeimpact at monitoring stations where apparentplume impact was observed. In most of these cases,there were no reported burns in the vicinity of themonitor and, thus, it appeared that other, non-reported burns were responsible for the apparentplume impact at the monitoring site. These cases donot provide information on the performance of themodel, but rather indicate that further work isneeded to identify all burns and to improve burn

reports in an accurate and timely manner. Addi-tional information on the timing of burns is alsoneeded. Finally, in the fourth category on two burndays, no plume impacts were observed at monitor-ing sites, and ClearSky did not predict plumeimpacts. While this is a positive result, it does notprovide information for a quantitative assessmentof model performance. However it should be notedthat in practice, correct and successful use of theClearSky system would generate cases in this fourthcategory.

4. Summary and conclusion

ClearSky, an automated web-based smoke dis-persion forecast system for agricultural field burn-ing, was developed to guide smoke management inthe Inland Pacific Northwest. The modeling systemoperates CALPUFF, the Gaussian Lagrangian puffmodel, in a forecast mode using MM5 4-km forecastmeteorology processed by the meteorological modelCALMET.

The modeling system was operated for the firsttime during the fall burning season of 2002, and,based on the post-evaluation results, an improvedversion of the model was operated in 2003. In thispaper, specific burn days were chosen as examples topresent the range of ClearSky 2003 evaluation results.

ClearSky evaluation results demonstrate that themodeling system has a potential to be a veryeffective tool for smoke management in the PacificNorthwest. On burn days when the MM5 predictedmeteorology is in good agreement with the observedmeteorology, ClearSky predicted plume impact atseveral monitors. The plumes, at times, did notalways pass directly over the monitoring stations,however, the predicted centerline concentrations formany of these cases were close to the maximumconcentrations observed at the monitoring sitewhich suggests that the hybrid source approachand associated source parameters are capable ofaccurate PM2.5 predictions, emission and plume riseparameters.

The model as expected does not predict plumeimpact from field burning when the forecastmeteorology is not representative of the ambientmeteorology within the airshed. Statistical analysisof predicted meteorology against observations wasperformed for seven surface meteorological stations.The results demonstrate that the errors in winddirection were closely related to the complexity ofterrain in the vicinity of each monitoring site.

ARTICLE IN PRESS

Table 5

Summary of burn days in terms of evaluation categories and

associated wind direction bias (absolute) bias and mean

(absolute) error

Day Category WD bias

(abs)

WD mean

error

7-August 1 18 18

11-August 1 19 27

12-August 1 3 26

19-August 1 7 22

25-August 1 21 50

26-August 1 2 17

15-September 1 11 47


Average – 13 29

14-August 2 90 115

3-September 2 95 95



Average – 72 81

6-August 3 4 22

13-August 3 11 15

5-September 3 4 10

11-September 3 1 20


19-September 3 6 9

Average – 7 18

12-September 4 4 6

25-September 4 1 8

Average – 3 7

Category 1 includes days when ClearSky predicted plume impact

and plume impact was observed; category 2 includes days when

the meteorology was poor and ClearSky did not predicted

observed plume impact; category 3 includes days when the

meteorology was good, but ClearSky did not predict observed

plume impact; and category 4 includes days when no plume was

predicted and none was observed.



To address the issues raised in the performance ofClearSky it is evident that improvements in theforecast meteorology are needed. One approach isto take advantage of ensemble meteorologicalforecasts to drive an ensemble of dispersion resultsand to produce a probability distribution of plumetrajectories and dispersion. We have evaluated thisapproach (Heitkamp, 2006), and these results willbe presented in a second paper. A second approachis to investigate the use of nested meteorologicaldomains with grid resolution down to 1 km or less.This may improve the results, particularly in areaswith more complex terrain, but it may still notaddress the effects of the random aspect ofconvective motions that occur during mid-daysummer and fall conditions. Further fieldwork isalso required to obtain more quantitative data forevaluation of the system.

Acknowledgments

The authors gratefully acknowledge support andcooperation provided by the Idaho Division ofEnvironmental Quality, the Washington Departmentof Ecology and members of the Northwest Interna-tional Air Quality & Environmental TechnologyConsortium (http://www.nwairquest.wsu.edu).

References

Air Sciences Inc., 2003. Cereal-grain residue open field burning

study. Report for: Washington Department of Ecology,

Washington Association of Wheat Growers, and US Envir-

onmental Protection Agency, Region 10. Project No. 152-02.

Barbosa, P.M., Stroppiana, D., Gregoire, J.-M., Pereira, J.M.C.,

1999. An assessment of vegetation fire in Africa (1981–1991):

burned areas, burned biomass, and atmospheric emissions.

Global Biogeochemical Cycles 13 (4), 933–950.

Byun, D.W., Ching, J.K.S., 1999. Science Algorithms of the EPA

Models-3 Community Multiscale Air Quality (CMAQ)

Modeling System. EPA/600/R-99/030.

Chastain, T.G., Young III., W.C., Garbacik, C.J., Meints, P.D.,

Silberstein, T.B., 1998. Residue Management and Stand Age

does not Affect Seed Quality in Grass Seed Crops. /http://cropandsoil.oregonstate.edu/seed-ext/Pub/1998/SR9908.htmlS.

Christian, T.J., Kleiss, B., Yokelson, R.J., Holzinger, R.,

Crutzen, P.J., Hao, W.M., Saharjo, B.H., Ward, D.E., 2003.

Comprehensive laboratory measurements of biomass-burning

emissions, 1: emissions from Indonesian, African, and other

fuels. Journal of Geophysical Research 108 (D23), 4719.

Cook, R.J., Haglund, W.A., 1991. Wheat yield depression

associated with conservation tillage caused by root pathogens

in the soil not phytotoxins from straw. Soil Biology &

Biochemistry 23, 1125–1135.

Crutzen, P.J., Andreae, M.O., 1990. Biomass burning in the

tropics: impact on atmospheric chemistry and biochemical

cycles. Science 250, 1669–1678.

Daniels, A., 1987. Diffusion and emissions of smoke from

agricultural burning in Hawaii. Water, Air & Soil Pollution 34

(2), 111–124.

Dennis, A., Fraser, M., Anderson, S., Allen, D., 2002. Air

pollution emissions associated with forest, grassland and

agricultural burning in Texas. Atmospheric Environment 36,

3779–3792.

Douglas, S., Kessler, R., 1988. Users Guide to the Diagnostic

Wind Field Model (version 1.0), System Applications Inc.,

San Rafael, CA, 48pp.

Gray, P.M., Guthrie, J.W., 1977. Burning and other cultural

practices relative to populations of seed-borne pathogens of

Poa pratensis. Seed Science & Technology 5, 545–553.

Hardison, J.R., 1980. Role of fire for disease control in grass seed

production. Plant Disease 64, 641–645.

Heitkamp, K., 2006. Evaluation of the ClearSky smoke disper-

sion ensemble forecast system for agricultural field burning in

eastern Washington and northern Idaho. M.S. Thesis,

Washington State University, Pullman, WA.

Hely, C., Caylor, K., Alleaume, S., Swap, R.J., Shugart, H.H.,

2003. Release of gaseous and particulate carbonaceous

compounds from biomass burning during the SAFARI 2000

dry season field campaign. Journal of Geophysical Research

108 (D13), 8470.

Higgins, K.F., Arnold, D.K., James, L.P., 1989. Prescribed

Burning Guidelines in the Northern Great Plains. US Fish

and Wildlife Service, Cooperative Extension Service, South

Dakota State University, US Department of Agriculture

EC 760. Northern Prairie Wildlife Research Center

Home, Jamestown, ND, Page. /http://www.npwrc.usgs.gov/

resource/tools/burning/burning.htmS (Version 16JUL97).

Jenkins, B.M., Jones, A.D., Turn, S.Q., Williams, R.B., 1996.

Emissions factors for aromatic hydrocarbons from biomass

burning. Environmental Science and Technology 30,

2462–2469.

Johnston, W.J., Golob, C.T., 2004. Quantifying post-harvest

emissions from bluegrass seed production field burning

(Washington State University, Department of Crop & Soil

Science). Final report for Grass Seed Cropping Systems

for a Sustainable Agriculture, Washington Department of

Ecology, Idaho Department of Environmental Quality,

Washington Turfgrass Seed Commission, and Coeur d’Alene

Tribe.

Mass, C.F., Albright, M., Ovens, D., Steed, R., Grimit, E., Eckel,

T., Lamb, B., Vaughan, J., Westrick, K., Storck, P., Colman,

B., Hill, C., Maykut, N., Gilroy, M., Ferguson, S.A., Yetter,

J., Sierchio, J.M., Bowman, C., Stender, R., Wilson, R.,

Brown, W., 2003. Regional environmental prediction over the

Pacific Northwest. Bulletin of the American Meteorological

Society 84, 1353–1366.

Mast, T.J., Hsieh, D.P.H., Seiber, J.N., 1984. Mutagenicity and

chemical characterization of organic constituents in rice straw

smoke particulate matter. Environmental Science and Tech-

nology 18, 338–348.

Mazzola, M., Johnson, T.E., Cook, R.J., 1997. Influence of field

burning and soil treatment on growth of wheat after

Kentucky bluegrass, and effect of Rhizoctonia cerealis on

bluegrass emergence and growth. Plant Pathology 46,

708–715.

ARTICLE IN PRESSR. Jain et al. / Atmospheric Environment 41 (2007) 6745–67616760


O’Neill, S.M., Ferguson, S.A., Peterson, J., Wilson, R., 2003. The

BlueSky smoke modeling framework. In: Fifth Symposium on

Fire and Forest Meteorology, American Meteorological

Society, Orlando, FL /www.blueskyrains.orgS.

Pyne, S.J., 1982. Fire in America: A Cultural History ofWildland and

Rural Fire. Princeton University Press, Princeton, NJ, 654pp.

Scholes, R.J., Kendall, J., Justice, C.O., 1996. The quantity of

biomass burned in southern Africa. Journal of Geophysical

Research 101 (D19), 23,667–23,676.

Scire, J.S., Strimaitis, D.G., Yamartino, R.J., 1999. A Users

Guide for the CALPUFF Dispersion Model (Version 5).

Earth Tech Inc.

Scire, J.S., Robe, F.R., Fernau, M.E., Yamartino, R.J., 2000. A

Users Guide for the CALMET Meteorological Model

(Version 5). Earth Tech Inc.

Sinha, P., Hobbs, P.V., Yokelson, R.J., Bertschi, I.T., Blake,

D.R., Simpson, I.J., Gong, S., Kirchstetter, T.W., Novokov,

T., 2003. Emissions of trace gases and particulates from

savanna fires in southern Africa. Journal of Geophysical

Research 108 (D13), 8487.

Tansey, K., Gregoire, J.-M., Binaghi, E., Boschetti, L., Brivio,

P.A., Ershov, D., Flasse, S., Fraser, R., Graetz, D., Maggi,

M., Peduzzi, P., Periera, J., Silva, J., Sousa, A., Stroppiana,

D., 2004. A Global inventory of burned areas at 1 km

resolution for the year 2000 derived from SPOT vegetation

data. Climate Change 67, 245–377.

Ward, D.E., 2001. Combustion Chemistry and Smoke. In:

Johnson, E.A., Miyanishi, K. (Eds.), Forest Fires:

Behavior and Ecological Effects. Academic Press, New York

(Chapter 3).

Ward, D.E., Susott, R.A., Kauffman, J.B., Babbitt, R.E.,

Cummings, D.L., Dias, B., Holben, B.N., Kaufman, Y.J.,

Rasmussen, R.A., Setzer, A.W., 1992. Smoke and fire

characteristics for Cerrado and deforestation burns in

Brazil—BASE-B experiment. Journal of Geophysical Re-

search 97 (D13), 14,601–14,619.

ARTICLE IN PRESSR. Jain et al. / Atmospheric Environment 41 (2007) 6745–6761 6761

Documents

Author's personal copy ARTICLE IN PRESSlar.wsu.edu/docs/2007_Jain_ClearSky_final_AtmEnv.pdf · Author's personal copy Atmospheric Environment 41 (2007) 6745 6761 Development of the