Estimating Fuel Consumption for the Upper CoastalPlain of South Carolina• Scott L. Goodrick, Dan Shea, and John Blake

Recent changes in air quality regulations present a potential obstacle to continued use of prescribed fire as a land management tool. towering of the acceptabledaily concentration of particulate matter from 6510 35 jig/m' will bring much closer scrutiny of prescribed burning practices from the air quality community.To work within this narrow window, land managers need simple tools to allow them to estimate their potential emissions and examine trade-offs betweencontinued use of prescribed fire and other means of fuels management. A critical part of the emissions estimation process is determining the amount of fuelconsumed during the burn. This study combines results from a number of studies along the Upper Coastal Plain of South Carolina to arrive at a simple meansof estimating total fuel consumption on prescribed fires. The result is a simple linear relationship that determines the total fuel consumed as a function of theproduct of the preburn fuel load and the burning index of the National Fire Danger Rating System.

Keywords: prescribed fire, emissions, fuel consumption

E

stimates of fuel consumption per unit area (PC) and emis-sions of particulate matter <2.5 jim in diameter (PM, 5 ) arerequired to strategically manage the smoke from prescribed

fire programs in the South and to assess their impacts on air quality.New federal regulations (US Environmental Protection Agency2007) have lowered the 24-hour maximum PM, exposure for thepublic from 65 to 35 jig/m 3 . Although the annual limit has notchanged (15 gig/rn t ), the cumulative impact of all sources of PM,5emission may also push designated urban environments over theannual threshold and into nonattainment status. Failure to achieveair quality standards could severel y restrict prescribed fire programsdesigned to reduce hazardous fuels, to restore red-cockaded wood-pecker (RCW) habitat (Picoides borealis) and native pine savannacommunities, and to maintain wildlife habitat for game species. Thecurrenr recovery plan for the RCW (US Fish and Wildlife Service2003) places emphasis on Frequent prescribed fire in restoring andsListaining RCW habitat, and frequent prescribed fire is critical torestoration and conservation of grass-forb savanna communities(Glitzenstein et al. 2003).

The Augusta—Aiken area, between Georgia and South Carolina,is an urban zone potentially affected by prescribed burning at theSavannah River site (SRS) and other forcstlands in South Carolinaand Georgia. Recent assessments by the South Carolina Depart-ment of Health and Environmental Control (Lawson 2008) and theGeorgia Environmental Protection Division (Johnston 2008) indi-cate that Augusta—Aiken area annual PM, 5 levels are close to the15 gig/nl annual standard. Similar to many other federal agencies inthe South, the Department of Energy has a goal to recover the

restore pine savanna communities, manage wildlife habitatfor game species, and manage hazardous fuels (US Department ofEnergy 2005). To achieve these goals, the objective for the annualprescribed fire program is 22,500 ac, a significant increase from the

Manuscript received September 17, 2008: accepted June 18, 2009.

historical average ol 13,000 ac (Kilgo and Blake 2005). Therefore,assessing the impacts of the fire program on air quality and identi-lying strategies to mitigate those impacts are critical. Current re-gional smoke management guidelines (US Forest Service 1989) andsmoke management regulations within South Carolina (SouthCarolina Forestry Commission 2005) are designed to limit undesir-able smoke impacts. However, they do not provide data to estimateFC and the resulting PM1 emissions.

The basic method to estimate PM 2.5 emissions from prescribedfires involves three independent variables: PC, area burned, and thecontaminant emission fector. If these variables are measured or canhe calculated, then the following equation is used: I'M 75 (mass) =

FC (mass per unit area) X Area X PM 25 emission factor(mass/mass).

PM, emission factors for wildland and prescribed fire are sum-marized by Batryc and Battye (2002). Urhanski et al. (2008) re-cently published emission factors for a large number of southernprescribed fires, including five fires at the SRS. Emission factors varyby combustion stage and fuel type, but the bulk values for individualburns are far less variable than FC. The PC term integrates the fuelbed structure and total available fuel loading (TF), which is thegreatest source of uncertainty in determining PC, as well as environ-mental conditions affecting fire behavior, and is therefore the great-est source of variation in the emission equation (Peterson 1987,Sandberg et al. 2002). Empirical estimates of PC from prescribedfires in the South are available from a limited number of studies.Hough (1968, 1978) produced a large number of PC observationsand generated relationships predicting FC as a function of TF andhulk duff-titter moisture content (MC) in northern Florida andsouthern Georgia. Ferguson er al. (2002) related litter and dullconsumption to diurnal changes in weather and moisture variablesat Eglin Air Force Base in the Florida panhandle, and Snyder et al.

Scott S. Goodrick (sgoodrickJifid. us), US Sn rest Service, Soutt'ern Research Station-R WU4104. 320 Green Street, Athens, GA 30602. Dan Shea andJohn Blake. US Forest Service,,Savannai, River, New Ellenton, S(72980. Support/br this work was providea' by the US Forest Service. Savannah River, under Interageiny Agreement DE-A109-00SR22188 withthe Department oJI-nogy ,S3van na/i River Operations Office. Appreciation is extended to Dr. Brian Sullivan for sharing at/down fitch data, the US Forest ,Service Missoula lireLaboratomyfirproviding unpublished data Jions the SRS emissions study, and Dr. Bernie Parresolfor statistical advice. lVi' a/mo thank the reviewers whose conimnents helped to greatlyimprove c!,eJinal manuscript, ibis article was written antI prepared by a US Government employee on official time, and it is therefore in the public domain and not copyrightable.

SOL'FH. I. A y is. For. 34(1) 2010 5

(2005) itiade F(. i c,o,uremenis for various I ucl types in the FloridaEverglades. Ottmar et al. (2006) measured TF and PC for individualfuel bed components in conjunction with woody fuel and duff MCon 30 prescribed fires in the South. Prichard et al. (2007) used thisdata to derive fuel consumption relationships for each individualFuel bed component as a function of loading, duff MC and thefraction of soil exposed after burning to improve FC estimates fromthe model CONSUME 3.0. Sparks et al. (2002) related FC to firebehavior for a shortleaf pine forest in Arkansas, and Waldrop et al.(2004) reported PC measurements for a Piedmont forest in SouthCarolina. Scholl and Waldrop (1999) provided fuel loading and FCfor eight forest stands in the Upper Coastal Plain of South Carolina,and Sullivan et al (2003) published litter and duff consumption inconjunction with a fire effects study in the same area. In agreementwith Sandberg et al. (2002), these PC observations for the South arehighly variable, ranging from <1 ton/ac to >10 tons/ac, Unex-plained variability in FC makes it difficult to estimate PC and sub-sequently emissions front prescribed burning program.

Research on PC within other regions shows FC is related to TPand some measure or index of fuel or duff MC, and occasionally firebehavior (Sandberg 1980, Little et al. 1986, Kauffman and Martin1989, Brown and Reinhardt 1991, Brown et al. 1991, Botelho et al.1994). The relationship between MC of 1,000-hour fuels or deadand down wood with a diameter of 3-8 in. is used as the environ-mental driver for FC in the model CONSUME 3.0, based on west-ern US regional data (Prichard et al. 2007). Various empirical equa-tions from the western United States and rules of thumb are used inthe submodel BURNUP as part of FOFEM (Reinhardt et al. 1997,Reinhardt 2003) to estimate PC. With the exception of the recentwork by Ferguson et al. (2002) and Prichard et at. (2007), therelationships between FC and TF, duff MC, and other environmen-tal variables have received limited attention in the South since thework of Hough (1968, 1978), which showed a strong predictiverelationship between average duff MC, TF, and PC. Ferguson et al.(2002) showed duffand litter MC by volume to be closely associatedwith percentage of litter and duff reduction in four burns conductedin longleaf pine. Actual duff MC is difficult and expensive to obtain;however, indices of duff or fuel MC can be measured or calculatedfrom local weather observations using the National Fire DangerRating System (NFDRS) (Cohen and Deeming 1985, Burgan1988).

Our objectives include (1) determining whether FC can be reli-ably predicted from measurements of TF and environmental vari-ables collected during fire research studies at SRS; (2) determiningwhether simple weather measurements, estimated fuel MC, orNFDRS indices can function as alternatives to direct measurementof duff-litter MC; and (3) combining FC with recently publishedemission factors for PM25 from the southeast to evaluate the im-pacts of a prescribed fire program and to identify strategies for re-ducing PM,,,emissions. The NFDRS indices of primary interestare the burning index (BI), energy release component (ERC), andthe Keetch-Byrani drought index (KBDI), as these quantities arecommonly used to assess wildfire potential in the South and aresignificantly related to the conditional probability of a wildfire atSRS (US Forest Service 2005).

MethodsThe SRS is a 198,000-ac Department of Energy facility and

National Environmental Research Park located in Aiken, Barnwell,and Allendale counties in South Carolina. The environmental and

Iorcsi stand conditions at the SRS have been described in derail,including the prescribed fire and wildland fire history (Kilgo andBlake 2005). Typical of the southern region, a large portion of theSRS was farmed following European settlement. Between 1951 and1970, about 40% of the land base was reforested with various south-ern pine species. Currently, about 65% of the area is in upland pineor pine-hardwood plantations (129,000 ac), 8% mixed hardwood,and about 22% bottomland hardwoods and cypress-tupelo forests.A significant prescribed fire program was initiated in the late 1970son a nominal 5-year cycle. Since 2005, over 20,000 ac/year of on-dersrory prescribed burning has been accomplished.

We analyzed FC data collected from three separate understoryprescribed fire studies conducted at the SRS between 1995 and 1997(Scholl and Waldrop 1999, Sullivan et al. 2003, Urbanski et al.2008). All three studies were performed in the eastern half of theSRS in the RCW management area (US Department of Energy2005). In the first study, Scholl and Waldrop (1999) collected pre-and posthurn data on eight forest stands prescribed burned in 1995to develop a pre- and postburn photo series for managers. Scholl(1996) established 10 plots in each stand. At each plot, a 50-ft planertransect (Brown 1974) was used to obtain a nondestructive estimateof live fuels (grasses, forbs, shrubs, and vines) on four 10.8-1 t 2 plots,and litter and cone dry weight on 0.54-fr 2 subplots. Down and deadwoody fuel, including 1-hour (<0.25-in, diameter), 10-hour(0.26-1.0-1n. diameter), 100-hour (1.01-2.99-in. diameter), and1,000-hour fuels, were measured along the transect line. Nonde-structive live fuel percentage cover and height were converted tobiomass from equations developed from a separate series of destruc-tive harvests on 100 plots of 10.8 ft 2 each. The second study is asubset of the data from a large emissions project conducted in 1996that spanned 40 sites in the southeast (Urbanski et al. 2008). As partof this effort, five prescribed fires at SRS were monitored to estimateFC and emission factors for various combustion products, includingPM 25 . Detailed methods are given by Ottmar and Vihnanek(1997). For each burn, pre- and postburn fuel measurements werecollected on 10 transects (50 ft each) centered on each of the threeemissions towers. A total of 16 pre- and posrburn plots (10.8 ft2each) were located along each transect. Within each plot, litter, duff,cones, 1-hour, 10-hour, and live vegetation components (grasses,forbs, shrubs, arid vines) were harvested to estimate biomass. In athird study, conducted in 1997, a comprehensive fire effects exper-iment was performed by Sullivan et 1]. (2003) to determine how lireintensity and FC related to root damage and mortality in old-fieldlongleaf pine (Pinus Ja1rotr1s L.). The data were collected on four5-ac replicates of three burn intensities for a total of 12 burns.Within each 5-ac stand, 10 transects (50 ft each) were established,similar to the second study, and 16 plots (10.8 ft 2 each) were alsosystematically established. In contrast to the second study, each plotwas subdivided into four subplots to measure pre- and posthurn livefuels (grasses, forbs, shrubs, and vines), litter and duff, cones, and1-hour and 10-hour fuels.

The original data in the first study by Scholl and Waldrop (1999)were modified by eliminating the 100-hour and 1,000-hour com-ponents, since these components are rarely consumed during pre-scribed fires in the southeast, and comparable data on these fuelswere not available in the other studies. Data on cones, I-hour andI 0-hoir fuels, and live fuels not previously reported by Sullivan et al.(2003) for the third study were included in the current anal ysis. Theunpublished observations on l'F and FC data from the second study(Urbanski et al. 2008) were provided by the US Forest Service

6 Soun -i.J. APi'L. FOR. 34(1) 2010

Missoula Fire laboratory for analysis. The calculated EC for allstudies was determined by subtracting 'IT for the described compo-

nents pre- and postburn.Although specific weather and ignition variables were collected

in all three studies, onl y direct fire behavior observations and fuelMC measurements were made during the second study (Ottniar andVihnanek 1997). Therefore, a common set of environmental andfire behavior variables from historical weather records are used forthe analysis. Variables for each prescribed fire were obtained for theSRS Remote Access Weather Station (383101-SAVRIV) from theNational Interagency Fire Management Integrated Database(NIFMID) database. Weather observations and NFDRS indiceswere calculated using Fire Family Plus, version 3.0.5.0. The primaryindices of interest in this stud y included the ERC, KBD1, and BI.The ERC can be viewed as a composite fuel moisture index thataccounts for the fuel MC of the various common fuel bed compo-nents (I-hour fuels, 10-hour fuels, etc.). The KBDI is a simplemeasure of drought designed as an indicator of duff and upper soilMG (Keetch and Byram 1968). The KBDI is also an integral com-ponent of the 1988 revision of NFDRS designed to improve per-formance for the southern United States (Burgan 1988). The B!provides an overall measure of fire behavior as it integrates fuelconditions (ERC) and the potential for fire spread (as indicated bythe NFDRS's spread component). The Bland ERG were calculatedfor each day of ignition based on fuel model P, which is representa-tive of many parts of the southeast coastal plain, including SRS, andfuel model G, which is commonly used as a reference fuel modelacross the United Stares, as it is the only fuel model to have fuel loadsin all fuel bed components. Values for BI and ERC were calculatedon the basis of the 1978 and 1988 model formulations (Burgan1988). The KBDI was also calculated for each day . The observed

10-hour fuel moisture (1 OH R), the calculated 1,000-hour fuelmoisture (1,000HR), and the observed relative humidity (RH) atmidday were also obtained from NIFMID data. Days since last rain>¼ in. (DSR) were obtained from the SRS annual meteorologicalreports (Parker and Addis 1993).

The data were analyzed with SAS (SAS Institute 2003). Pearsonscorrelation and ordinary least-squares stepwise linear regressionwere used to test the significance of main variables (TE, 131, ERG,KBDI, DSR, RH, IOHR, and 1,000HR) on FC at an a level of

0.05. The effect of the three research studies on the slope and inter-cept of the regression relationships was determined by creating anew variable, Study, and coding it as a dummy variable (0, 1, or 2)for each study. When initial examination of residuals suggested apossible interaction between TF and some of the environmentalindices, four new variables were created (TF X III, TF >< ERG,

TF X DSR, and TF X KBDI), and the analysis was repeated. Weperformed the regression analysis with TF in separate groups foreach of the NFDRS indices (BI, ERG, and KBDI), including the1978, 1988, and model P and C formulations of BI and ERG. TheEG analysis was run with independent variables TF, BI (1978, fuelmodel P), DSR, 1OHR, 1,000HR, and RH, followed by TF, BI(1988, fuel model P), DSR, 1OHR, 1,000HR, and RH, until all ofthe various BI and ERG values were used. Similarly, the FC analysiswas run with independent variables TF, KBDI, DSR, 1,000HR,and RH, but not 1 OHR because of the very strong correlation be-tween KBDI and 1OHR. The regression analysis to predict PC wasalso run with each non-NFDRS environmental variable (DSR,IOHR, 1,000HR, and RN) in combination with IF.

Although we ran identical statistical analysis using both the 1978

and 1988 formulations of NFDRS fuel models P and C, the r2

values for comparable regression models were consistently lower byabout 5-22% for the 1988 formulation, with the difference gener-

ally smaller for model G than for model P. The largest reason for this

difference in performance may be due to the inclusion of the KBDIas an additional variable in the BI and ERG calculations to increaseavailable fuel in standard fuel models to improve drought represen-tation in humid environments. The benefits from these changes on

wildfire potential are greatest in summer and fall, whereas prescribedburning generally occurs in winter and spring. Coupled with addi-tional changes in how fine fuel moisture, greening of vegetation, and

wind speed are incorporated, the 1988 formulation may not be asapplicable as the 1978 version to predict FC from prescribed burn-ing at SRS. To simplify the presentation of results, only results for

the 1978 formulation of NFDRS models P and C are reported.

ResultsThe ranges of 'IT, PC, prescribed fire ignition methods, and

environmental conditions are representative of prescribed burning

in the Upper Coastal Plain in South Carolina (Table 1). The TFranges from 2.6 to 12.3 tons/ac and PG from 1.0 to 8.6 tons/ac. The

stands are primarily pine stands between 20 and 50 years of age.Ignition methods include both hand (head and hacking fires) andaerial ignition. Burns were conducted over a period between Januaryand May, which is the traditional prescribed burning season for this

region.The Pearson's correlation values indicate that none of the envi-

ronmental or fire behavior variables are correlated with TP (Table 2)and therefore are not confounded in the results. In the regressionanalysis, TF has the highest correlation of any variable with FGacross all studies, but the variation in the relationships suggests thatother Factors may influence FC (Figure 1). As single environmental

variables, 131 (models P and C), ERG (models P and C), and DSRare significantly correlated to FG, indicating, as in previous research,that environmental indices can explain variability in observed FC.

Of the fire behavior variables, B! is most strongly associated withPC. The correlation coefficients between individual environmentaland fire behavior variables are generally consistent with how they are

interrelated in NFDRS calculations (Cohen and Deeming 1985).For example, BI builds on ERG by inclusion of wind speed andtherefore should be correlated. The weak relationships between BIin model P and I ,000HR fuels contrast the relationship for 131 inmodel C. The NFDRS model C incorporates substantial amounts

of both 100-hour and 1,000-hour fuels, whereas model P has none.KBDI is a long-term drought index and therefore correlates well

with the 1.000-hour fuels and the fuel model G indices, which areinfluenced by the inclusion of these larger fuels. The 10-hour fuelMG showed a strong negative correlation to KBDI and a positive

correlation to RH.The independent environmental variables RH and 1 ,000HR

were not significantly related (P = 0.05) to FG in any ordinary

least-squares model either alone or when IF was included. Exami-nation of residuals did not suggest alternative model formulationssuch as interactions with TF. This result was not surprising, since

Soul H. J. Aiti FoR. 34(1) 2010 7

Table 1. Fuel, stand, ignition, and environmental characteristics of the three prescribed fire studies at the Savannah River site.

Prescribed fire study

Study attributes Scholl and Waldrop (1999) Urbauski et al. (2007) Sullivan ci al. (2003)

'I Lii of sidyIgnition period (month range)IL'nitlon methodStand types; age

\lost recent prescribed fireI- loading (tons/ac)

IC (tons/ac)01 (model P)

RI (model G)

ITC (model P)

ERC (model G)

1)SR > 0.25 in. (days)OF-JR moisture (%)

I .000HR moisture (%)I H ((4)R1)BI

1995January-February

AerialLoblolly & longleaf pine; 20-50 years;

one hardwood stand4 years; unknown (hardwood stand)

5.9-9.11.4-5.0

17-280978)21-35 (1988)"

1-18(1978)0-20 (1988)"

22-24 (1978)27-31 (1988)"

0-5 (1978)0-6 (1988)"

3-511-1218-5034-4710-18

1996February-MayAerial & hand

Mature loblolly and longleaf pine;one pine pocosi 0

1.5-9 years2.6-12.31.6-8.6

14-38(1978)16-21 (1988)"20-55(197.8)21-30 (1988)"19-29(1978)22-28 (1988)"24-35(1978)

25-33 (1988)"1-48-12

18-1924-5760-253

997lcbruary-March

Hand (head & hacking fires)Old-field longleaf - pine;

40-45 years>7 years

7-9.3

1.0-5.516-260978)16-32 (1988)"12-31 (1978)10-33 (1988)"21-290978)21-36 (1988)"6-25 (1978)5-26 (1988)'

2-98-12

22-2718-486-138

Represenrs the range of values for the 1988 lorniularinuuns of - models I' and C.II JR. I ((-hour fuel moisture: I.0001 IR. I .000-110tii lute) ulisi:iurc: RI. Isuirnirig odes: I )SR. dal I 5115,1 1:155 rIm ill.: (EL. siloryt rIlcascCn u:i;,o,ie ii: IC. (a,) ,n1,urr(. nIb unit arcs; K[ )BI

kecich I(II.lnnIlnnu I/Ii etcics; RI) nun:, lulnilicil :s:H, natal ,i:.iulaInl (Hi bed sLru,rIirc md 1-illiili

Table 2. Pearson's correlation coefficient (r) matrix and probability for r values for fuel and environmental variables across all studies.

1)SR

0.38760.05560.06910.74250.55770.00380.36790.7040.6321)0.00070.24820.2315

EEC r'((1978)

0.62540.00080.30250. 14 160.65150.00040.44990.0240

.RC (C)(1978)

0.47100.01750.05600. 79030.27800.1784O.874_1

0.00010.36450.0732

IOHR

-0.28320.17010.2689(1.1936

-0.23590.2562

-0.57990.0024

-0.58810.0020

-0.68950.0001

-0.46050.0205

1,000HR

-0.2079

0.31850.07620.7170

-0.34490.0913

-0-73190.0001

-0.06880.7437

- (1.66220.0(103

-0.12780.54250.30710.1354

RH

-0.10060.6323(1.04730.8222

-0.19640.3466

-0.28490.1675

-0.62690.0008

-0.22531)2787

-0.29200.15670.66110.0003

-0.00810.9691

K 1)01

(1.13620.4357

-0.35470.08190.01830.93050.59620.00300.2811((.17340.80290.00010.31590.1239

-0.90250.0001

-0.44470.0259

-0.37140.0676

HI (P) RI ((;)TF

(1978) (1978)

IC.. Prom. >

0.5816

0.6124 0.5975

0.0023

0.0011 0.00161 F Prob. >

0.1398 0.13080.5050 0.5332

(I (P) Prob. >

0.62370.0009

Of (G) Prob. >

ERG (P) Prob. >

ERG (C) Prob. >

I )SR Prob. >

Of IR Prob. >

I ,000HR ('rob. >

RH Prob.>

I ,000F1R. 1.000-hour fuel moisture; III. burning index; DSR, days since last rain, >1 us.; F.RC, energy release componienn; EC, fuel consumption per unit area: RI )R I. Kemch-Rvram drought tides;prob., probabiluiv; RH, relative bonn idinv; TF, total available fuel bed structure and loading.

RH was only one of several weather variables effecting MC of fuels,and 1,000-11our fuels represented only one minor component ofsouthern fuel beds, such as fuel model P. Only TF in combinationwith BI (models P and G), ERIC P and G), KBDI, DSR,and IOHR proved to be significant (Table 3). Higher J?, Fvalues,and lower mean square errors were obtained with BI over ERC forboth NFDRS models, probably as a result of incorporation of windspeed. The TF X BI (model P) variable improved the modelR 2-value over TF and BI separately (Table 3, Equation 3 versusEquations 4 and 6), as did TF 3< DSR (Table 3, Equation 2),suggesting that PC may be better understood by considering the

positive feedback between TF and environmental, variables during

prescribed burning. Although DSR was a weak cnvironmcntal vari-able, it improved the R2 for several models, including the KBDIregression model (Table 3, Equation 10).

The effect of study on the intercept and slope of the regressionmodels of TF and BI (model P; Table 3, Equation 3) or the corn-bused variable TP X RI (model P; Table 3, Equation 4) against FCwas not significant (P = 0.05). Very high regression R values wereobserved between TF X BI and FC for two of the individual studies(Figure 2). The R2 values were 0.84 (P = 0.0001)in the third study,conducted by Sullivan et al. (2003) in 1997, and 0.97 for the secondstudy, b y Urbanski et al. (2008) (P = 0.0020). For the Scholl andWaldrop (1999) study, the R 2 was weak (P = 0. 1695), but the data

8 SOU I H. J. AI'l'L. FOR. 34(1) 20 10

JUG

AGO -

FC AG,d,0. 0 JOU, 0 400'A

600

800

330 0cc 4 0 OCY 81C CC C3 '203 430

Total FuI Load (t04A per ,re)

Figure 1. Composite relationships between IF and FC for the 25prescribed fire observations at the Savannah River site. FC, fuelconsumption per unit area; IF, total available fuel bed structureand loading.

range was also more limited than in the other two studies. In con-trast, models involving B! (model C) showed substantial reductionin residual variance by including study dummy variables (Table 3,Equations 5 and 6). The effect of study on the intercept and/or slopeof the regression models of FF and ERG (models P and G), KBDI,and 10HR was also significant and generally improved the overallmodel R 2 (Table 3, Equations 7-10). Some anal yses showed signif-icant effects on model slope, whereas others showed effects on modelintercept, thus making interpretation of the study effects difficult.The effect of stud y on the EC also limits the model utility andgeneralization of the results for estimating PM, 5 emissions. Thelack of significant impact of the study dummy variable on the slopeor intercept of the TE X BI (model P) model allowed the studies tohe pooled and treated as one data set to estimate FC (Fig 3). Theadjusted R for this combined data set is 0.69, which provides areasonable predictive equation for managers.

Finally, we computed the FC using the equation of Hough(1978) for the five SRS fires from the second study conducted dur-ing 1996 in which duff layer MC was measured (Urbanski et al.2008). The average observed FC was 4.1 tons/ac, which compareswell with a predicted average FC of 3.66 tons/ac from Hough'sequation using 'IF and duff MC. The Pearson's correlation coeffi-cient was 0,81 (P = 0,095). Although we did not have duff MC forthe other studies, this simple comparison suggests that the funda-mental relationships proposed by Hough may be valid in the UpperCoastal Plain of South Carolina as well.

DiscussionConsistent with research on FC within other regions (Sandberg

1980, Kauffman and Martin 1989, Brown and Reinhardt 1991Brown et al. 1991, Botelho et al. 1994) and the previous studies of'Hough (1968, 1978), we found TE to be a critical variable control-ling FC ]it study . Fuels inventory or monitoring data, publishedphoto series (Ottmar et al. 2003), and data developed for the FuelCharacteristic Classification S ystem in the southeast (Ottmar et al.2007) call the uncertaint y in TF estimates for landownersand regulatory agencies. The relationships between natural fuelcomponent d ynamics associated with forest type, stand age, stock-ing, site index, and fire histor y may be as important as simple mea-sures of TF to effectively use the full range of silvicultural practices

required to manage hazardous fuels and emissions. Parresol et al.(2006) observed that fuel component loadings derived from site-wide inventory plots at SRS were significant functions of stand age,basal area, and site index, as well as Are history, within forest types.These observations suggest that silvicul tural practices affecting spe-cies composition, stocking. and rotation age are critical in a com-prehensive approach to managing TF.

The combination of NFDRS indices and TF works well in pre-dicting FC. NFDRS calculations are driven by local weather data.Therefore, it is no surprise that variables obtained from the NFDRS,such as BI and ERG, would improve the modeling of PC (Hough1978). The BI (model P) is a very good predictive variable foroccurrence of large wildfires (>10 ac) at SRS (US Forest Service2005). The ERG is also a significant predictive variable for wildfiresat SRS, but it may not have been as effective as B! in reducingresidual variability in the PC prediction because wind speed is notincorporated in the index. ']'he KI3DI is significantly related to FCin this study and generally performs as well as the BT (model C),ERG (model C), and 1 OHR variables, with which it is stronglycorrelated. The generally weaker model significance levels may be aconsequence of the filet that the KBDI is designed to track droughtconditions and is less sensitive to short-term weather changes thatinfluence MC dynamics in litter and duff (Ferguson et al. 2002). Wefound DSR a significant but weak environmental variable for pre-dicting FC. Empirical equations relating DSR directly to duff MCfor the southeast have been developed by Hough (1978). However,the complexity of antecedent moisture conditions and the size of theprior rainfall event may limit the development of a broad-basedenvironmental index for DSR (Ferguson et al. 2002). In contrast tostudies from the western region of the United States (e.g., Sandberg1980, Little et al. 1986, Brown et al. 1991), the calculated MC of1.000-hour fuels was unrelated to PC despite its being correlated toocher environmental variables (e.g., BI, model C) that were them-selves significant in predicting PC. The environmental indices thatoffer the greatest predictive potential for PC during prescribed burn-ing are those that integrate and track the MC dynamics of theavailable fuel bed.

Environmental indices stich as 1 ,000HR and KBDI are not nec-essarily indicative of the MC of the fuel bed components that will beconsumed in a prescribed burn because of their lower sensitivity toshort-term weather fluctuations; in addition, NFDRS indices forfuel models representing dense forests with large dead fttel accumu-lations (BI model G) possess a level of inertia that limits their sen-sitivity to short-term weather changes because of the larger fuelclasses. The 1988 NFDRS formulation, through its inclusion ofKBDI, also exhibits a certain degree of inertia, as shown by theweaker relationships between FC and TP plus either BI or ERGcompared with relationships for the 1978 indices. Other environ-mental variables, such as RH, or indices, such its the 10-hour MC,ma y respond faster than the MC of the available fuel bed. WhetherNPDRS indices can he applied broadly in the South ma y depend onhow the natural range of fLie] loading and fuel arrangement controlsfuel tnoisture dynamics of the available fuel bed (Nelson and Hiers2008).

Despite some success in predicting PC, predicting PM, to showthe potential of reducing particulate emissions is more complex.Reducing the land area treated with prescribed fire is possible ifalternatives exist that can cost-effectivel y achieve hazardous fuel andecological objectives. Although the use Of Silviculture and harvestingpractices can reduce hazardous fuels (Rummer et al. 2002, Waldrop

SOLCIH. J . Airs. Fos. 34(1) 2010 9

RMSE

2.119031.762751.252291.001010.609740.819621.360481.073521.26161.35921

ProbabilityF>

0.00230.00090.00090.000!0.000!0.00(110.00010.0001(1.00010.0003

Equationno.

-7

34-7

6789

10

0.33830 .4 73 50.62590.68740.83440.771420.61210.69390.64030.6309

12

10 •9 1950a 1957

- -L,real 1596;

a 8 --[near (19911

0VEB00a,

FC 119971 = 0 0337TF81 -2 1803R2 = 08412

FC 1956) 0 0199x 0 1175 6

R2=09769

Table 3. Ordinary least squares best fit regression models for FC with associated R2, Probability of F value > and RMSE.

Regression model for FC (tons/ac)

= -0.4903 + 051018FF= -0.6064 + 0.40101 TF + 0.03149 (IT >< DSR)= -3.1979 -!- 0.44374 IF + 0.16001 B] (P)= -0.01516 4- 0.02207 IF x BI (P)

= -2.2357 - 0.21783 IF X Study + 0.70378 IF + 0.05581 RI (G) + 0.05314 131 (C) X Study= 3.12084 + 0.01202 IF X B! (G) X Study - 1.87941 Study= -7.00438 + 0.49116 TF + 0.31312 ERC (P) -i- 0.06188 IF )< Study= -1.3739 + 0.59726 IF + 0.09924 ERC (C) - 0.91037 Study= 3.4804 + 0.72624 TF - 0.45162 IOHR - 0.64606 Study= -2.69319 + 0.72157 TF + 0.01000 KBDI - 0.64633 Study + 0.21826 DSR

I OH R, ! 0-hour 8iei moisture; BE burning index: DSR, days since last rain > Yr in.; ERC, energy release component: FE, fuel consumption per unit area; K[)Rl Knetch-B y ram droiighi index; prob.,probability; RMSE, root mean square error; TI, total available filel bed structure and loading.

-

C 50 100 150 200 250 300 350 400 450Total Fuel Load (tons per acre) x Burning Index

Figure 2. Relationships between FC and TF x BI from three inde-pendent prescribed fire studies at the Savannah River site. The dataand equation from the second study (Urbanski et al. 2008) arelabeled 1996. The data and equation from the third study (Sullivanet al. 2003) are labeled 1997. The data from the first study (Scholland Waldrop 1999) are labeled 1995, but the equation is notshown (P = 0.16). BI, burning index; FC, fuel consumption per unitarea; IF, total available fuel bed structure and loading.

et al. 2004), the existing research does not indicate that ecologicalobjectives for maintaining or restoring pine savanna grasses andforbs are achieved with mechanical or chemical treatments alone(Brockway er al. 2009). There is a range in observed emission factorsfor the southeast (Urbanski et al. 2008), but there is currently nosystematic relationship identified that can be used by managers toreduce emission factors. Firing techniques such as backing fires aresuggested as a means of reducing particulate emissions by increasingcombustion efficiency. However, field studies show that they gen-erally consume more litter and duff but relatively less live fuel(Hough 1968, Botelho et al. 1994, Sullivan et al. 2003). Undersimilar fuel and environmental conditions, the FC is very similar(Hough 1978). Backing fires also have lower lire line intensities thatcan limit smoke plume rise within the mixing layer, and they requirelonger ignition periods that can lead to smoke being trapped duringinversions. The latter condition may increase the 24-hour PM,above regulated levels.

When we combine the FC modeled with variables TF and BI(Figure 3; Table 3, Equation 2) with recently published emissionfactors for PM, 5 from the southeast (Urbanski et al. 2008), and thetarget prescribed fire goal of 22,500 ac/year (US Department ofEnergy 2005), the average annual PM 2 emissions impact from

0 SC lCD 150 200 250 000 013 903 450

Total Fuel Load tons per acre) o Burning index

Figure 3. Composite relationship between FC and TF x BI(model P) for 25 prescribed fire observations at the Savannah Riversite. BI, burning index; FC, fuel consumption per unit area; TF, totalavailable fuel bed structure and loading.

prescribed burning can be evaluated (Table 4). For perspective, themagnitude of emissions From existing plantations, when burned ona cycle of 5+ years, is approximately double the "M,2. 5 emissionsFrom a 50-year old boiler at SRS burning 154,000 tons of coal/yearto generate process steam (US Department of Energy 2008). Since along-term ecological objective for a large portion of the SRS isdevelopment of RCW habitat characterized by frequently burnedopen pine and grass-forh communities (US Fish and Wildlife Ser-vice 2003), the future air quality impacts should reflect these fuelbed conditions. Although these conditions are currently limited atSRS, the live vegetation TF can be approximated from SRS old-fieldstudies (Odum 1960), frequently burned range studies in the South(Gaines et al. 1954, Brockway and Lewis 1997), and dead TF fromthe SRS inventory equations relating dead fuel component loadingto stand age, basal area, site index, and fire frequency (Parresol et al.2006). The total annual PM 25 emissions are drastically reducedunder these ecological conditions as a direct result of changes instand structure, leading to much lower TE (Table 4). Creating andsustaining these fuel bed conditions requires a more frequent pre-scribed fire cycle (McNab et al. 1978, Hough 1978, 1982, Hanulaand Wade 2003, Ghitzensrein et al. 2003); therefore, instead ofburning, for example, a total of 112,500 ac on a nominal 5-year

cycle (22,500 ac/year), we can treat only a total of 67,500 ac on a

B

10 SOUTH. J . ArpL. FOR. 34(1) 2010

Table 4. Estimated annual PM,., emissions from prescribed firesburning at the Savannah River site. Pine plantation total availablefuel (TF) is the average of the 25 prescribed fire observations.Expected long-term ecological restoration fuel loads for pine sa-vannas are derived from the literature as noted.

Emission Annual I otal I'Mfactor" Tb FC5 area burned emissions

Forest condition (lb/tots) (tons/ac) (tons/ac) (ac) (I b)

Pine plantation 27.967.4 3.25 22,500 2.045,330Pine savanna' 26.06 2.05 0.89 22,500 521,852Pine savanna' 26.06 3.25 1.42 22,500 832,617

Urbanski ci al. (2008). The savanna hums included FSI. ICII. 1C12, lc;13, and SC3 fromAppendix Al.

Calculated from Equation 4 iii Table 3; burning index 20.Live fuel load based on Gaines ci al. (1964), Odust ((960), and Brocksvav and Lewis (1997).

Inventory equations of l'arresol cc al. (2006) used no estimate litter, duff, (-hour and 10-hourloading in 60_ye.ir old longleaf pine. site index 70, BA o1`30 and burned every, 3 years.' Live fuel (nail based on Carter and Hughes (1974) and dead fuel is based on Pai rem( ci al.(2006), as in footnote c

FC, fuel consumption per unit area; I'M, , particulate mnaaer; 'IF, total available hid bedstructure and loading

3-year cycle. Hazardous Fuels reduction would require implementa-

tion of other .silvicultural practices, such as biofucl harvests, or stra-tegic placement of fuel treatments across a landscape to limit largewildfires (Finney et al. 2007).

Although the data collected have a limited geographic scope, webelieve they are consistent with other research and support several

important conclusions. Total fuel consumption from prescribedfires can be reliably estimated front of total available fuelloading and environmental indices. We found support for Hough's

(1978) proposal that NFDRS indices may be useful Surrogates fordirect measures of duff MC in the southeast, but additional analysisOf unpublished EC data front prescribed fires (Ottmaret al. 2006, Urbanski et al. 2008) would help validate their broadutility for FC models such as CONSUME (Prichard ct al. 2007).Opportunities to reduce total PM, emissions are limited, but in-creasing fire frequency on a smaller area of the landscape is feasible,and it is compatible with RCW habitat and longleaf savanna resto-ration goal.

Literature CitedOs 1 i i. 'A.. AND R. RAT is s. 2002. Development of emissions trir'entory unee/sod.sJhr

u'i/d/cusul fire. Final Rep. Contract 68-l)-98-046. US Environmental ProtectionAgency, Research Triangle Park, NC. 91 p.

RYFELHO. H., J.A. VFCA, I'. FESN,ANDLs, AND F. Rico. 1994. Prescribed firebehavior and fine fuel consutnprion its northern Portugal and Gahiza Maritimepine stands. P.343-353 in Proc. of 2nd mt. Con/.Forest Fire Research, 21-24/Tot'.1994, Coimbra. Portugal. Viegas. [).X. (ed.). Universidade dc Coimbra,Co,tnbra, Portugal.

BROCK\vAY, D.C., AND C.F. 1.twis. 1997. Long-term effects of dormant seasonprescribed fire oil community diversit y , structure and productivity in alongleaf wiregrass ecosystem. Fan Ecol. Manag. 96:167-1 8.3.

BRUCK\AAY. L).G., K.W. OcTcAIr, B.E. ESIES, AND R.B. Rusi.situ. 2009.Vegetation response to midsrorey nittichting and prescribed burning for wildfirehazard reduction and longlcaf pine (Pinnispaitistrns Mill.) ecosystem restoration.Forestry 82(3):299-.314.

BR0\vN,J.K. 1974. HandhoohJoriiu'encsrying down woody material. Gen. Tech. Rep.INT-16, US. For. Serv. Intermountaits For. Range Exp. Sm., Ogden, UT. 24 p.

Rqowy ,j.K., AND RHNHAIADT. 1991. Predicting aisd managing fuel consumption itsthe interior west. P. 419-429 in l'roc. of 11th C'onf on Fire and forestMereoroloy, April 16-29, 1991, ,"kf,ssou/.a, MT. Society of American Foresterspublication 91-04, Bethesda, MI).

BttowN,J.K., E.D. ReiNH,'\R[il, ANti W.C. Fist os. 1991. Predictingduffand woodyfuel consumptiots in Northern Idaho prescribed fires.. For. 5u'i. 37:1550-1566.

BuRu;A y , R.F. 1988. lire 1988 revisions to the 1978 National Fire Danger RatingSystem. SE-RP-273, US For. Sers'. Southeast. For. Exp. Stn., Asheville, NC. 39 p.

(;ARJER C.W.,.Aiu R.H. HUGIIt;s. 1974. 1.ongleah-.slash pine-wiregrass range.P. 13-16 in Range resources of the South, Lewis. CE., H.D. Grelen, L.D. White,anti C.W. Carter (eds.). Soc. Range Mang. and Georgia Agric. I'xp. Sits. Bull.NS-9, Tifton, GA.

CoheN, ID., AND J.E. DtoiINc. 1985. The National Fire Danger Rating System.'Basic equations. US For. Sen. Pacific Southwest For. and Range Exp. Sm. Gen.Tech. Rep. PSW-82, Riverside, CA. 16 p.

FISRULSON, S.A., J.A. Rc y Hi 01St), S.J. M( CAN, D. Wies;ii I, C. Wiucil, AND

R. 0 1TA.Atl. 2002. Measuring moisture ds'tamics to predict fire severity in

longleafpitse forests. Inc.]. Wile//and Fire 11:267-279.FINNEY, MA., R.C. Stit, C.W. MLHLCH, A.A. Ai;s.u, B. B..soio, AND J.K.Ac,iI.

2007. Simulation of long-term landscape-level fuel treatment effects oilwildfires. lint. J. Wild/and Fire 16:712-727.

GAINES, F.M., P.S. C.ssit'sei I , AND J .J . BR,ASINCI ON. 1954. Forage production onlongleaf pine lands of southern Alabama. Ecology 35:59-62.

GI,ryzi:Ns (tiN, j.S., D.R. STUN(,, AND D.D. W,sot. 2003. Fire frequency efkcts onlongleaf pine (Pinus pa/nazis P. Miller) vegetation in South Carolina andnortheast Florida, USA. PAst. Areas!. 23:22-37.

HANi:u.A, j.L., AND D.D. WADI , . 2003. Inliucinee of long-term dormant-seasotiburning and fire exclusion on ground dss'eIli ng arthtopod populations in longleafpine flatwood ecosystems. For Ecol. Manag. 175:163-184.

H OUGH, W.A. 1968. Fuel consumption and/ire behavior ofhaz.ard reduction burns. USFor. Serv. Southeast. For. Exp. Sin., Asheville, NC. Res. Pap. SE-36. 7 p.

HOUGH, W.A. 1978. Estimating available fuel tvcig/ir ronsunird by prescribed fires intime South. US For. Sets'. Southeast. For. Exp. Sin., Asheville, NC. Res. Pap.SE-187. 12 p.

Hoioii, W.A. 1982. Ph yronsass and nutrients in the understory and forest floor ofslashllongleafptie stands. For. ,S'ci. 28:359-372.

JOHNSTON, J . 2008. Augusta Air Quality (Jpdate (March 27, 2008). GA Environ.Protect. Div., Dept. Nat. Res., Atlanta GA. 10 p.

K.\uiFMAN, j.[).. AND MARTIN, R.E. 1989. Fire behavior, fuel consumption, andforest floor changes following prescribed undersmory fires in Sierra mixed coniferforest. Can./. For. Res. 19:455-462.

KFFTLH,J.J., AND BYRAM, G.M. 1968. A drought indreforforestfire control. US For.Sets'. Southeast. For. Exp. Stn. Res. Pap. SE-38. Asheville, NC. 32 p.

Kit co. iC.. AND J.1. BLAKE. 2005. Ecolog and management oJ'aJhresiee/ landscape:hij^yyears on the Savannah River site. Island Press, Cesvelo, CA. 434 p.

LAwSoN, C. 2008. Ozone and PM2.5 Update for 2007 (March 24. 2008). SouthCarolina Department of Health and Environmental Control, Columbia, S(29 p.

Lrni.t/, SN., R.D. Di tMAtt, AND J.i.. OrhsI.ANN. 1986. Predicting t/uff'ionsurnpeionfrom prescribed burns on csniji'r clearcuis in western Oregon and tvesternWashington. US For. Sers'. Pacific Northwest For, and Range Fxp. Sin. Rem. Pap.PNW-362, Portland, OR.

W.I 1., M.B. EDWARDS. JR., AND W.A. Hou(;t I. 1978. Estimating fuelsveights in slash pine paltneitci stands. For. Sci. 24:345-358.

NrIsON. R.M., Asi) K.J. Hit-its. 2008. The influence of fttelbed properties onmoisture (Irving rates and ii melags of longleaf pine litter. Can. J. For. Res.38:2394-2404.

Oousi, F.P. 1960. Organic production and turnover in old field succession. Lcolagr.4 1:34-49.

01 - i MAR, R. AND R. VIHNANtK. 1997. Study of emissions o,f ' air pollutants frombiomass fires in the United .States: A progress report on .f'ael characterization- Year 2.Unpublished report. US For. Scrv. Pacific Northwest Res, Sm., Seattle, WA. 26 p.

O y i y sis, RI)., R.E. VIHNANtK, AND J.W. MA] 111 N. 2003. Stereo photo series forqteanttJyinq natural/belt. Volume Via: sane/ lit/I. sand pine scrub, and hso'clu'oaelswith white pine types inn the Southeast United States with supplemental sites forvolume VI. PMS 838. National Wildfire Coordinating Group, NationalInteragency Fire Center, Boise, It). 78 p.

C) IlMAR, RI)., S.J. PRILFIARD, R.E. VII INANEK, AND D.V. SANDISFItc. 2006.Modification and validation afflict consumption nindetu Jar s/i rub anadforesied landsin i/se .Southwesi, Pacific Northwest, Miduoest, Southeast, and Alaska. Final ReportJoint Fire Sciences Program Project 98-1-9-06. US For. Serv. Pacific NorthwestRes. Stn., Pacific Wildland Fire Sci. I_ahoratory, Seattle, WA. 97 p.

0i'ist.sis, RI.)., D.V. SANImERL, C.L. RILL,SRD, Asi) S.J. PRICHARI). 2007. Anoverview of the Fuel Characteristic Classificatiott System: Quattrih'ing,classifying, and creating fuelbeds for resource planning. U(iit. J. Far. Res.37:2383-2393.

PARKI.]cM.J.,..\ND R.P. ADDIS. 1993. Meteorological Monitoring Program.WSRC-'l R-93-0 106. Westinghouse Savannah River Co., Savannah RiverTechnology Center, Aiken, SC. 87 p.

PAiskesol , BR.. I). Stat,., AND R. OrrSi.sR. 2006. Creating a fuels baseline andestablishing fire frequency relationships to develop a landscape managementstrategy at the Savannah Riversite. P. 351-366 i n Proc. I-ne/s 'v/ann gemneni-Howto Ivleasure ,Success, Andrews, P.L., and B.W. Bret (cotsips.). US For. Serv. RockyMtn. Res. Sits. Gen. Tech. Rep. RMRS-P-41, Fort Collins, CO.

PETERSON, J. L. 1987. Analysis and reduction aJ'che errors ofpreduccing prescribed burnemissions. M.S. thesis, University of Washington, Seattle, WA. 70 p.

SOLID. J . AIn't.. FOR. 34(1) 2010 11

Psi lisso, S. R.D. Di i \ l iS. ,\\I CR. A\ Iv l0\. 200. (o\c(:.1//:i() 15.

guide. US For. Serv. Pacific Northwest Res. Mn., Pacific \X'jldjand Fire Sd.Laboratory, Seattle, WA. 238 p.

RFINH,;RIsr, E. 2003. Using FOFEM 5.0 to estimate tree mortality, f/sml consumption,smoke production and soil heating from wildiand fire. In Proc. of2nd InternationalWild/and lire Ecology and hire Management Congress, 16-20 November 2003,Orlando, FL. American Meteorological Society, Boston, MA. 6 p.

RcINciARor, E.D..J.K. llRo\SN, W.C. FISCHvR, AND R.T. GRAHAM. 1991. Woodyfuel and duff consumption b y prescribed fire in northern Idaho mixed coniferlogging slash. US For. Serv. Intermountain Res. Stn. Res. flap. INT-443, Ogden,UT. 22 p.

RI11NFIARD1, ED., R.F. KEANF, AND J.K. BROWN. 1997. First Order Fire hjJicesModel.' FOFE'vI 4.0, user's guide. L'S For. Serv. Intermountain Res. Stn. Gen.Tech. Rep. INT-344, Ogden, UT. 65 p.

RtJMMER, B., K. OuTcAIT, ANt) D. BROCKwAY. 2002. Mechanical mid-storyreduction treatments for fuel management [Abstract]. P. 76 in Proc. of Anon.Meet. Southern Weed Sci. Soc., Champaign, IL.

SANDBERc, D.V. 1980. Duff reduction by prescri bed underburning in Douglas-fir. USFor. Sers'. Pacific Northwest For, and Range Exp. Stn. Res. Pap. PNW-272,Portland, OR. 18 p.

S.sNDBF.IsG, D.V., R.D. OTFMAR, J.L. PFli,RSoN, AND J . Cosu-.. 2002. Wild/and-fireon ecosystems: Efl'ects offire on air. US For. Sets'. Rocky Mtn. Res. Stn. Gen. Tech.Rep. RMRS-GTR-42 Vol. 5, Ogden, UT. 79 p.

SAS INS ...... .uTii. 2003. SAS/SiTusersguide. version 6 Vol. 2(4th ed.). SAS InstituteInc., Cary, NC. 1848 p.

Sci cii.. E. R. 1996. Fuel classification f/sr prescribed burning of longleaf and loblollypine plantations in the Upper coastal Plain. MS thesis, Clemson University,Clemson, SC. 62 p.

Sc,soi i. ER .,s.scn 1.A. WAtoRol'. 1999. Photos fir escimatingjdel loadings beforeand after prescribed burning in the upper coastal plain ofthe southeast. US For. Serv.Southern Res. Stis. Gen. Tech. Rep. SRS-26, Asheville, NC. 29 p.

SNYDER,J.R., M.S. Ross, S. Koi'i CR. AND j. S.SH. 2005. Developing ecological criteriafor prescribed fire in stint/s Florida Pint Rock/and Ecosystems. SoutheastEnvironmental Research Center. Florida International Univ., and US GeologicalSurvey, Miami, FL. Open File Report OF 2006-1062. 114 p.

'si: I H (.iisoi Ni Us vu i : 10 Cs s\liiI'Hs s',. 21)0 ..s,ss/,'s ss5s;inyessissit yss0/clziie.sfsi

vegetative debris burn/sic' f/)r Jossscry. sgrscislcusc ,sssd wi/ti/i/c ' ussposu's 6th

printing). South Carolina Forestry Commission, Columbia, SC.

SPARKS,J.C., R.E. MAsIERs, D.M. FACt r., AND G.A. Bi:Kr.NI-ori:R. 2002. Season ofburn influences fire behavior and fuel consumption in restored shortleaf

pine-grassland communities. Restoration Ecol. 10:714-722.

SL'l I IVAN, B.T., C.J. Frrric, W. OTR0SINA, M.J. DA[ USKY, ANDCX'. BERISF0RD.

2003. Association between severity of prescribed burns and subsequent activityof conifer-infesting beetles in stands of longleaf pine. For. Fail. Manag.185:327-340.

WAIDRol', TA., D.W. GlAss, S. Riotour. AND V.B. Siiirisuiosr. 2004. Anevaluation off/se/-reduction treatments across a landscape gradient in piedmontforests: Preliminary results of the National Fire and Fire hurro gate Study. US For.

Serv. Southern Res. Sm. Gen. Tech. Rep. SRS-71. Asheville, NC. S4-59.URRANSKI, S.P., W.M. Hso, AND S.P BAKER. 2008. Chemical composition of

wildland fire emissions. P. 79-107 in Wild/anti fires and air pollution,Bvtnerowicz, A., M. Arbaugh, C. Andersen, and A. Riebau, (eds.). Elscvicr.Amsterdam, The Netherlands.

US FORJsu SFRVICE. 1989. A guide far prescri bed fire southern forests. Tech. Publ.R8-TP-1 1, Southern Region, Atlanta, GA.

US F0R1si Sr.Rvicr. 2005. Savannah River wildfire preparedness level staffing guide.US For. Sees'., Savannah River, New Ellenton, SC. 5 p.

US DF.i'sRroioT or f0sriw;s . 20(15. Natural resource management plan for theSavannah River site. New Ellenton, SC. 48 p.

US DrPAR! Mt NT OF ENFRC;Y. 2008. Environmental assessment fir biomasscogeneration and heating fbcilicies at the Savannah River site. DOE/FA-1605.

Savannah River Operations Office, Savannah River Site, Aiken, SC. 78 p.US ENVIRONMEN si. PR0Tsc i 10N AGENcY. 2007. (Jean air fine particle

implementattousftna/rule. Part II, Federal Re ister. 40 C'FR 51. US Environ nierital

Protection Agency, Washington, DC.US Fist] AND W11 .1)1_11T Setivici . 2003. Recovery plan for the red-cockadedivousdpecker

(Picoides borealis) (2nd revision). US Fish and Wildlife Scrs'icc, Atlanta, GA.296 p.

12 Souri- t.J. API'L. FOR. 34(1) 2010