Characterization of Atmospheric Ammonia Emissions from a Commercial Chicken House on the Delmarva Peninsula R O N ALD L. S I E F E R T , *,† JO S E P H R . S C U D LAR K, ‡ AM E LI A G . P O T T E R , § KI R S T E N A. S I M O N S E N , † AN D KAR E N B . S AVI D G E ‡ Ch esapeak e Biological Laboratory, Un iversity of Marylan d Cen ter for En viron m en tal Scien ce, 1 W illiam s Street, Solom on s, Marylan d 20688, Gradu ate College of Marin e Stu dies, Un iversity of Delaw are, Lew es, Delaw are 19958, an d Departm en t of Natu ral Scien ces, Un iversity of Marylan d Eastern Sh ore, Prin cess An n e, Marylan d 21853 A three-dimensional sampling grid using passive collectors w as used to characterize the dow nw ind gas-phase ammonia plumes originating from a commercial chicken house on the Delmarva Peninsula in the Chesapeake Bay w atershed. Inverse Gaussian plume modeling w as used to determine the source strength of the chicken house and the corresponding chicken emission factors. A total of seven field deployments w ere performed during tw o different flocks w ith a sampling duration ranging from 6 to 12.6 h. The deployments occurred during w eeks 3, 4, and 5 of the 6-w eek chicken grow -out period in the months of M ayJuly 2002. The ammonia emission factors ranged from 0.27 to 2.17 g of NH3-N bird -1 day -1 w ith a mean of 1.18 g of NH3-N bird -1 day -1. W eighted emissions factors that accounted for the nonlinear increase in ammonia emissions over the 6-w eek grow -out period w ere also calculated and ranged from 0.14 to 1.65 g of NH3-N bird -1 day -1 w ith a mean of 0.74 g of NH3-N bird -1 day -1. These w eighted emission values w ould correspond to an annual release of approximately 18 × 106 kg of NH3-N to the atmosphere from broiler production on the Delmarva Peninsula. This assumes that the emission factors in this study are representative for the entire year w ith varying meteorological conditions and are representative of all chicken husbandry practices. The Delmarva Peninsula could represent a significant source of nutrient nitrogen to the Chesapeake Bay and Delaw are Bay w atersheds through atmospheric deposition w hen considering the size of this annual release rate, the relative short atmospheric lifetime of ammonia due to deposition, and the proximity of the Delmarva Peninsula to the Chesapeake and Delaw are Bays. Introduction Atm osp h eric dep osition of n itrogen (N) sp ecies (am m on ia an d n itrate) is estim ated to con tribu te 27% of th e total † ‡ § n u trien t N load to th e Ch esap eake Bay (1). Both wet an d dry dep osition con tribu te to th is atm osp h eric dep osition of n itrogen sp ecies. Castro an d Driscoll (1) u sed th e best available wet an d dry dep osition data from th e Nation al Atm osp h eric Dep osition Program (NADP), th e Clean Air Status an d Tren ds Network (CASTNET), an d the Atm osp heric In tegrated Mon itorin g Network (AIRMoN). However with in these data sets, dry deposition data for particulate am m on ium is scarce, an d dry dep osition data for gas-p h ase am m on ia does n ot exist sin ce gas-p h ase am m on ia is n ot m easu red. Am m on ia is a m u lti-p h asic atm osp h eric sp ecies th at can occu r in th e gas p h ase (NH 3), in th e p articu late p h ase (NH 4+), an d in th e aqu eou s p h ase (p redom in an tly as NH 4+). NH x is defin ed as th e su m of am m on ia (NH 3) an d am m on iu m (NH 4+). There are large differen ces between the dry deposition rates of gas-p hase NH 3 an d p articulate-p hase NH 4+ (2). Th ese differen ces can lead to h igh dep osition rates of gas-p h ase NH 3 n ear sou rces of gas-p h ase NH 3 (2). Th e con version of gas-p h ase am m on ia to p articu late-p h ase NH 4+ is dep en den t on th e am bien t con cen tration s of gas-p h ase an d aerosol species (e.g., acids) an d can im pact the tran sport of am m on ia. Becau se of th is m u lti-p h asic ch em istry, NH x h as a relatively sh ort atm osp h eric lifetim e (h ou rs to days) th at is sh orter th an th e atm osp h eric lifetim e of n itrate or su lfate (3). Stu dies in Western Eu rop e h ave sh own th at th e tran sp ort an d dep osition is a com p lex p rocess n ear agricu ltu ral sou rces (4, 5). Cu rren t estim ates of NH x sou rces an d dep osition in th e Ch esap eake Bay airsh ed are n ot well con strain ed. A recen t in ven tory for th e Ch esap eake Bay watersh ed in dicates th at agricu ltu ral livestock con tribu te 81% of th e an n u al NH x atm osp heric burden (6). However, such estim ates are derived from an im al-sp ecific em ission factors (7), wh ich are often based on m an y assu m p tion s. Valu es for em ission factors vary widely dep en din g on th e sp ecific an im al h u sban dry an d m an u re disp ersal p ractices u tilized. An n u al broiler p rodu ction on th e Delm arva Pen in su la, cu rren tly at 6 × 10 8 birds yr -1, h as in creased m ore th an 20fold over th e p ast two decades an d h as been cited for th e 60% in crease in NH 4+ wet dep osition observed n earby at Lewes, DE, du rin g th is tim e p eriod (8). A p rim ary con cern wh en raisin g ch icken s in a con fin ed sp ace is th e rem oval of elevated NH 3 in side th e h ou se by ven tilation sin ce NH 3 h as adverse h ealth effects on th e ch icken s. Oth er fu n ction s of a ven tilation system in clu de th e rem oval of h eat, th e rem oval of m oistu re (th is m ay also decrease NH 3 volatilization by keep in g th e litter dry), an d th e p rovision of oxygen . Sidewall ven tilation an d tu n n el ven tilation are two com m on m ethods of ven tilation . Direct NH 3 em ission s from ven tilatin g ch icken h ou ses con tribu te to NH x em ission s in th e Ch esap eake Bay airsh ed. Previou s stu dies h ave u sed differen t m eth ods for m easu rin g NH 3 em ission s from agricu ltu ral bu ildin gs. Dem m ers et al. (9) u sed carbon m on oxide as a tracer to determ in e th e airflow rate th rou gh n atu rally ven tilated livestock bu ildin g wh ile m easu rin g NH 3 con cen tration s to determ in e th e NH 3em ission factors for th e livestock. Fowler et al. (5) m easu red NH 3 con cen tration s n ear a broiler h ou se su rrou n ded by woodlan ds an d calcu lated th at on ly 3.2% of th e NH 3 em itted from th e p ou ltry u n it was dep osited with in 230 m of th e h ou se an d th at 10% was dep osited with in 1000 m . Asm an (4) p redicted som ewh at greater dep osition over com p arable distan ces, wh ich is in flu en ced by factors su ch as sou rce h eigh t, win d sp eed, atm osp h eric stability, su rface resistan ce, su rface rou gh n ess, an d p lan t NH 3 com p en sation p oin t. FIGURE 1. M ap of farm w ith chicken houses and locations of sampling tow ers. Inset A provides more details about the arrangement of sampling tow ers and also show s the cross-sectional view of the sampling tow ers looking dow n the centerline of the chicken house. Th e cu rren t estim ates of th e atm osp h eric dep osition of am m on ia/ am m on iu m to th e Ch esap eake Bay an d Delaware Bay are h igh ly u n certain du e to ou r lim ited u n derstan din g of th e sou rce stren gth s, atm osp h eric ch em istry, an d dep osition al p rocesses for NH x n ear stron g sou rces. Th is stu dy in vestigated NH 3 em ission s from a side-wall ven tilated com m ercial ch icken h ou se n ear Prin cess An n e, MD, on th e Delm arva Pen in su la. Ath ree-dim en sion al sam p lin g grid was u sed to m easu re th e gas-p h ase NH 3 disp ersion down win d of th e 11 500 cap acity ch icken h ou se. In verse m odelin g of th e NH 3 p lu m e was p erform ed to determ in e th e am m on ia sou rce stren gth from th e ch icken h ou se. Methods Am m onia Plum e Sam pling Downwind of the Chicken House. An array of 10 m h igh sam p lin g towers was p laced down win d of ch icken h ou se A to form a th ree-dim en sion al sam p lin g grid (Figu re 1). Th e arran gem en t of th e towers was op tim ized for characterizin g the NH 3 p lum e from the chicken h ou se du rin g SSW win ds wh ere th e flow of air was p arallel to th e len gth of th e ch icken h ou se. Th is arran gem en t was ch osen in order to m in im ize th e overlap of NH 3 p lu m es from the adjacen t houses an d also to m in im ize airflow disturban ces du e to th e h ou ses by m in im izin g th e cross-section al p rofile of th e h ou ses. Th e area h as a flat top ograp h y, an d th e u p win d an d down win d areas h ave m in im al trees an d bu ildin gs. Th e field wh ere th e sam p lin g was con du cted was n o-till p lan ted with soybean s, wh ich in May were ju st em ergin g an d in Ju ly were at a h eigh t of abou t 30 cm A su rface rou gh n ess len gth of 0.1 m was u sed for both th e May an d Ju ly stu dy p eriods. Passive gas-p h ase NH 3 sam p lers were located at h eigh ts of 1, 5, an d 10 m on each sam p lin g tower excep t at th e research trailer location wh ere sam p les were collected at on ly 1 an d 5 m. Th e tower arran gem en t an d location h eigh ts of sam p lers were op tim ized by u sin g a Gau ssian p lu m e m odel to p redict down win d NH 3 con cen tration s (based on the n um ber of birds in th e h ou se an d p u blish ed em ission factors). Th e sam p lin g grid was design ed to be close en ou gh to th e sou rce wh ere su fficien tly h igh NH 3 con cen tration s wou ld allow a dep loym en t tim e of abou t 6 h for th e p assive sam p lers. It was also advan tageou s to locate th e sam p lin g towers n ear th e h ou se to m in im ize the in fluen ce of the other n earby chicken houses. Hou se B (Figu re 1) was exp ected to in flu en ce con cen tration s in th e sam p lin g grid becau se of its p roxim ity to h ou se A. However h ou se C an d h ou se D (a tu n n el-ven tilated h ou se with th e exh au st fan on th e n orth en d of th e h ou se) were far en ou gh away from th e sam p lin g grid th at th ey were n ot exp ected to in flu en ce th e con cen tration s in th e sam p lin g grid. Th e sam p lin g grid also n eeded to be located far en ou gh from th e sou rce so th at airflow distu rban ces du e to th e crosssection al p rofile of th e ch icken h ou se did n ot sign ifican tly im p act th e Gau ssian p lu m e p rofile (see in set A in Figu re 1). An addition al ration ale for arran gin g th e sam p lin g grid at som e distan ce from th e ch icken h ou se is becau se it m akes the in fluen ce of the source height on the con cen tration profile less im p ortan t. Th e m an u re an d com p ost sh ed (Figu re 1) also could in fluen ce the airflow; however, the n orth an d south en ds of th e bu ildin g were op en . Th erefore it h ad a m in im al cross-section du rin g SSW win ds. Th e m an u re an d com p ost sh ed cou ld also in flu en ce NH 3 con cen tration s for th e n orth west sam p lin g tower if th e m an u re an d com p ost h ad source stren gths com parable to the chicken houses; however, n o litter was bein g stored in th e sh ed du rin g th is stu dy. Th e accuracy of th e vertical an d horizon tal disp ersion coefficien ts u sed in th e Gau ssian p lu m e m odel also decreases n ear th e p lu m e sou rce (with in rou gh ly th e first 50 m ); th erefore, it was better to sam p le fu rth er away from th e h ou se. Overall, th e sam p lin g design in volved com p rom ises between th ese com p etin g con sideration s. Meteorological Measurem ents. Meteorological m easurem en ts were p erform ed at th e research trailer location . Am ast with an an em om eter, win d van e, an d tem p eratu re an d relative h u m idity sen sors was m ou n ted above th e research trailer. Th e overall h eigh t of th e win d sp eed an d win d direction sen sors was ap p roxim ately 6 m (abou t 3 m above th e roof of th e trailer on th e u p win d side of th e trailer). Th e win d sp eed sen sor h ad an accu racy of (0.5 m s -1, an d th e win d direction sen sor h ad an accu racy of (5° (Win d Sen try m odel 03001 an em om eter an d van e, Cam p bell Scien tific, In c.). Th e tem p eratu re sen sor h ad an accu racy of (0.5 °C, an d th e relative h u m idity sen sor h ad an accu racy of (3% (m odel CS500 with a m odel 41301 6-p late radiation sh ield, Cam p bell Scien tific, In c.). Data acqu isition was p erform ed u sin g a PC with a data acqu isition card an d LabView software (Nation al In stru m en ts, In c.). NH3 Measurem ents. Ogawa passive sam plers (Ogawa USA In c., Pom p an o Beach , FL) were u sed to determ in e th e tim eaveraged gas-p h ase NH 3 con cen tration s. Com p ared with tradition al m eth ods of NH 3 sam p lin g (su ch as den u ders), th e p assive sam p lers are ideally su ited for th is stu dy sin ce th ey are relatively in exp en sive, robu st, can be dep loyed u n atten ded for exten ded p eriods, an d do n ot requ ire electricity. Th e sam p lers were dep loyed u sin g th e rain sh elter an d clip s p rovided by th e m an u factu rer (see ogawau sa.com ). Th e rain sh elter is requ ired, even in th e absen ce of rain , to m in im ize th e effects of tu rbu len ce on th e diffu sive sam p lers as well as con tam in ation from litter du st th at is p revalen t arou n d th e ch icken h ou ses. After dep loym en t, th e exp osed sam p lers were sealed in side airtigh t p lastic vials, an d th e vials were tran sp orted to th e laboratory in an airtigh t p lastic con tain er an d frozen u n til an alysis. Filters from both en ds of th e p assive sam p ler were com bin ed for an alysis in order to im p rove sen sitivity, particularly durin g short deploym en ts. Addition al sam p lin g details can be fou n d in Roadm an et al. (10). Usin g strin gen t clean in g an d h an dlin g p rotocols, th e field blan ks were 236 ( 112 n g of NH 3 (n ) 32). For a typ ical 8-h dep loym en t, th is tran slates to an effective con cen tration of 15.8 µg of NH 3 m -3, wh ich rep resen ts less th an 9% of th e average observed con cen tration . On th e basis of trip licate dep loym en ts at on e location du rin g each set of m easu rem en ts, rep rodu cibility averaged 7.6% at con cen tration s ran gin g from 46 to 505 µg of NH 3 m -3. Gaussian Plum e Model. Lagran gian -based Gau ssian p lu m e m odels (GPMs) are com m on ly u sed to describe th e disp ersion of a sp ecies em itted from a p oin t sou rce (11). GPMs assu m e h om ogen eou s h orizon tal an d vertical tu rbu len ce an d th erefore are lim ited to sp ecific idealized con dition s. For exam p le, vertical tu rbu len ce is n ot h om ogen eou s, an d th e effect of th is n on -h om ogen eou s vertical tu rbu len ce is m ore p ron ou n ced for sou rces n ear th e grou n d du e to th e vertical win d velocity stru ctu re. However, th e m odels are com m on ly u sed becau se of th eir sim p licity, an d in th is stu dy we h ave ch osen a GPM to an alyze con cen tration s p rofiles of NH 3 p lu m es em itted from com m ercial ch icken h ou ses. Th e error in trodu ced by n on -h om ogen eou s vertical tu rbu len ce is also less p ron ou n ced in th is stu dy sin ce th e sam p lin g array is located relatively close to th e sou rce. An iterative com p u ter p rogram based on th e GPM was u sed to determ in e th e sou rce stren gth of NH 3 from th e ch icken h ou se for each of th e field dep loym en ts. Th e GPM is referred to as an in verse GPM in th is stu dy sin ce th e GPM is bein g u sed to calcu late th e sou rce stren gth based on th e observed con cen tration s in th e p lu m e, as op p osed to th e u su al u se of a GPM to m odel th e p lu m e con cen tration s for a given sou rce stren gth . Th e p rogram was written u sin g LabView software (Nation al In stru m en ts In c., www.n i.com ). Th e ch icken h ou se was ap p roxim ated as a lin e sou rce in th e m odel by u sin g th e p rin cip le of su p erp osition to sim u late th e lin e sou rce by in corp oratin g m u ltip le p oin t sou rces alon g th e len gth of th e ch icken h ou se (12). Th e m odel was ru n u sin g a h orizon tal grid resolu tion of 1 m ; th erefore, th e distan ce between th e p oin t sou rces was 1 m . Th e len gth of th e lin e sou rce was allowed to vary in th e m odel, bu t th e en d of th e lin e sou rce was always fixed at th e n orth en d of th e ch icken h ou se. Du rin g th e May/ Ju n e stu dy th e exh au st fan s were located alon g th e len gth of th e east side of th e ch icken h ou se an d th erefore th e lin e sou rce was p laced alon g th e east side of th e h ou se. Du rin g th e Ju ly stu dy, th e fan s were arran ged alon g th e cen terlin e of th e h ou se an d were directed to blow th e air toward th e n orth en d of th e bu ildin g wh ere th e n orth en d doors were left op en (see Figu re 1, in set A); th erefore, th e lin e sou rce was p laced alon g th e cen terlin e of th e h ou se. Th e effective release h eigh t of th e NH 3 was also allowed to vary in th e m odel. Th e GPM also in clu ded ch icken h ou se B sin ce th is h ou se cou ld also in flu en ce th e NH 3 con cen tration s on th e west side of th e sam p lin g tower array. Again th e p rin cip le of su p erp osition was u sed to in clu de th is secon d ch icken h ou se in th e m odel. Both ch icken h ou ses A an d B were assum ed to have iden tical em ission characteristics in th e m odel (e.g., sou rce h eigh t, sou rce len gth ). Th is assu m p tion was u sed sin ce both ch icken h ou ses were iden tical in size an d were op erated iden tically (e.g., flocks arrived on the sam e date, sim ilar feedin g, sim ilar ven tilation ). Th e followin g GPM form u la was u sed to gen erate th e con cen tration p rofile for each p oin t sou rce: 〈c(x,y,z)〉 ) ( )[ ( ) (z - h )2 y2 q exp exp + 2 2πuj σyσz 2σy 2σz 2 ( exp - (z + h )2 2σz 2 )] (1) wh ere c is th e con cen tration (g m -3), q is th e sou rce stren gth (g s -1), x is th e distan ce in x direction (m ), y is th e distan ce in y direction (m ), z is th e distan ce in vertical direction (m ), σy is th e h orizon tal disp ersion coefficien t (m ), σz is th e vertical disp ersion coefficien t (m ), an d h is th e h eigh t of release (m ). Th is form u la is for total reflection at z ) 0 (su rface) an d u ses th e slen der p lu m e ap p roxim ation (11). Th e total reflection at th e su rface assu m es th at th ere is n o sign ifican t dep osition of gas-p h ase NH 3 in th e sam p lin g area. Un der n eu tral con dition s an d with a source height of 3 m , Asm an (4) showed th at ju st over 20% of th e NH 3 was dep osited with in 200 m of the source. These con dition s would represen t a con servative m axim u m sin ce th e field sam p lin g in th is stu dy was closer to th e sou rce (with in 120 m of th e sou rce). An d Fowler et al. (5) observed on ly 3.2% of th e total NH 3 bein g dep osited with in 230 m of a ch icken farm su rrou n ded by woodlan ds. Th e h orizon tal an d vertical disp ersion coefficien ts u sed in th e Gau ssian p lu m e m odel were from Pasqu ill-Gifford correlation s for differen t Pasqu ill stability classes (see ref 11). Th ese disp ersion coefficien ts are a fu n ction of th e down win d distan ce, x. Th e con cen tration p rofile gen erated by eq 1 is for on ly on e poin t source. The m odel in cluded m ultiple con cen tration p rofiles th at were gen erated u sin g eq 1 an d th en su p erim - TABLE 1. Summary of Conditions for Each Field Deployment (FD) FDa start date description study period (h) I II III IVd V VI VII M ay 31 June 5 June 5 July 22 July 23 July 27 July 27 afternoon daylight overnight overnight daylight afternoon overnight 6.9 9.9 12.6 10.0 6.0 6.9 12.0 temperature RH w ind speed flock age (d) mean ( SDc (°C) mean ( SDc (%) mean ( SDc (m s-1) w ind vector direction (deg) speed (m s-1) skye Pasquill stability date 33 37 37 29 30 34 34 28.5 ( 1.1 29.4 ( 1.4 24.8 ( 0.7 28.7 ( 1.9 31.5 ( 0.8 28.1 ( 1.0 23.6 ( 1.4 58 ( 6 62 ( 5 80 ( 3 62 ( 11 54 ( 4 76 ( 5 96 ( 4 4.3 ( 1.2 3.4 ( 1.1 3.5 ( 1.3 2.6d ( 1.7 5.6 ( 1.4 2.6 ( 1.4 0.1 ( 0.3 193 180 186 187 193 184 206 4.11 3.02 3.32 2.42 5.35 2.42 0.09 CL CL CL CL, M OC OC, M OC, M , F B, C B, C D C C B NA f a FD, field deploym ent. b The w ind vector is the average w ind vector over the duration of the deploym ent. Wind direction is relative to axis of the chicken house. For exam ple, for a w ind direction of 180°, the research trailer w ould be directly dow nw ind of the chicken house. This corresponds to a true w ind direction of 195° (SSW) since the chicken house m akes an angle of 195 relative to true north. c SD, standard deviation. d There is m issing data due to a pow er failure during deploym ent V. Therefore, the m eans and standard deviations are not for the entire deploym ent period. Weather data from Princess Anne indicated that the m ean w ind speed during the deploym ent w as 3.5 m s-1. f There is not a stability class defined for nighttim e periods w ith w ind speeds less than 2 m s-1. p osed to m odel th e ch icken h ou se as a lin e sou rce an d to also add th e secon d ch icken h ou se (h ou se B in Figu re 1). A cost fu n ction (CF) was defin ed to qu an tify h ow well th e m odeled con cen tration s com p ared to th e m easu red NH 3 con cen tration s: CF ) ( )∑ 1 j j (cm easu red,i - cm odeled,i)2 (2σm easu rem en t,i)2 (2) wh ere 2σm easu rem en t ) 2 × stan dard deviation based on th e accu racy of th e atm osp h eric NH 3 m easu rem en ts an d j is th e n u m ber of m easu rem en ts. Th e p rogram was written to sequ en tially vary each of th e p aram eters (win d sp eed, win d direction , sou rce len gth , sou rce h eigh t, an d sou rce stren gth ) un til the program con verged on a set of values that m in im ized th e CF. Confidence Lim its on Source Strength. Two differen t ap p roach es were u sed to estim ate th e con fiden ce lim its of th e sou rce stren gth from th e in verse GPM. Th e first was to u se th e CF to determ in e h ow well th e m odel con cen tration s com p ared to th e actu al m easu rem en t, an d th e secon d was to u se a Mon te Carlo sim u lation to qu an tify th e variability of th e in verse m odel in determ in in g th e NH 3 sou rce stren gth . Th e valu e of th e CF is u sefu l in qu an tifyin g h ow well th e m odel com p ares to th e m easu rem en ts. Th e GPM accu rately sim u lates th e down win d disp ersion of NH 3 from th e ch icken h ou se, with in th e accu racy of th e m easu rem en ts, wh en th e cost fu n ction is less th an or equ al to 1. Th is in fers th at, on average, th e m odel con cen tration s are with in th e accu racy error of th e NH 3 m easu rem en ts. Valu es of th e CF th at are greater th an 1 in fer th at th e m odel con cen tration s are, on average, outside the accuracy error of the NH 3 m easurem en ts. For exam p le, a valu e of th e CF equ al to 10 (u sin g an estim ated error for th e m easu rem en ts of (10%) wou ld corresp on d to th e average differen ce between th e m odeled an d m easu red NH 3 con cen tration s of 32%. Th e m odel is an idealized sim u lation of th e actu al con dition s, an d valu es of th e CF th at are greater th an 1 m ay be du e to several reason s related to factors n ot accou n ted for by th e GPM. Airflow distu rban ces du e to stru ctu res are n ot accou n ted for in th e m odel alth ou gh th e field site was ch osen to m in im ize th ese effects. Th e m odel also sim u lates th e ch icken h ou se as a lin e sou rce; h owever, th e ch icken h ou se is actu ally 9 m wide, an d ven tilation is on th e sides of th e bu ildin g. Th e m odel also assu m es th at m eteorological con dition s do n ot vary over th e dep loym en t p eriod; h owever, variation s in win d sp eed an d win d direction do occu r du rin g th e sam p lin g p eriod even th ou gh th e field sam p lin g strategy attem p ted to ch oose sam p lin g p eriods where these dyn am ics were m in im ized. Th e m odel also assu m es th at th e h orizon tal an d vertical disp ersion coefficien ts are accu rate for th is stu dy site. Overall, all of th ese factors m ay con tribu te to h igh er valu es of th e CF. However, it is difficu lt to ascertain a con fiden ce lim it on th e sou rce stren gth determ in ed by th e m odel by u sin g th e CF. A h igh CF m ay n ot n ecessarily m ean th at th e sou rce stren gth h as h igh u n certain ty bu t th at th e m odel assu m p tion s are n ot accu rate. AMon te Carlo sim u lation was also p erform ed to estim ate th e con fiden ce lim its on th e sou rce stren gth . Sim u lated data sets were gen erated for win d sp eed, win d direction , sou rce h eigh t, an d sou rce len gth . Th e rejection m eth od was u sed to gen erate n orm al (Gau ssian ) distribu tion s of win d sp eed an d win d direction (13). Th e m an u factu rer’s stated accu racy for th e sen sors was u sed for th e stan dard deviation of th e win d sp eed (accu racy ) (0.5 m s -1) an d win d direction (accu racy ) (5°) for th e sim u lated data sets. Th e reason for usin g th e m an u factu rer’s stated accu racy is because th e GPM is based on th e assu m p tion th at th e win d vectors h ave a Gaussian distribution . Th is Gau ssian distribu tion is th e result of h om ogen eou s tu rbu len ce an d is th e reason for th e Gau ssian -sh ap ed p lu m e. Th e sou rce release h eigh t was a ran dom h eigh t between 0 an d 4 m (abou t th e h eigh t of th e ch icken h ou se). Th e sou rce len gth was a ran dom len gth between 11 an d 64 m (th e len gth of th e ch icken h ou se) at 1-m in tervals. Mon te Carlo sim u lation s were origin ally p erform ed with ran dom sou rce len gth s between 0 an d 64 m ; however, the results of these Mon te Carlo sim ulation s resulted in n on -n orm al distribu tion s. Th ese Mon te Carlo sim u lation s were p erform ed for each field dep loym en t an d rep resen t th e sen sitivity of th e sou rce stren gth valu es with resp ect to th e accu racy of th e m odel in p u ts. Results and Discussion Field Deploym ents. Table 1 lists th e seven field dep loym en ts (FDs) th at were p erform ed for th is stu dy. Th ese FDs occu rred du rin g two differen t ch icken flocks. Th e first th ree FDs were du rin g th e first flock, an d th e last fou r FDs were du rin g th e secon d flock. All of th ese FDs h ad du ration s between 6 an d 12.6 h . Th e tim e to eith er dep loy or retrieve th e p assive sam p lers was on average 40 m in . Th e order in wh ich th e p assive sam p lers were dep loyed an d retrieved was th e sam e for each FD. Chicken House Operation and Field Conditions. Chicken s were grown ou t for a 6-week p eriod, an d Table 1 in clu des the age of the flock for each field dep loym en t. Chicken houses A an d B both h ad ap p roxim ately 11 500 ch icks p laced in th e h ou se at th e sam e tim e, an d ap p roxim ately 11 150 were h arvested (assu m in g a 3% m ortality rate). Both ch icken h ou ses A an d B h ad side-wall ven tilation . FIGURE 2. M eteorological conditions for each field sampling period (I-VII). One hour intervals are used for the inside tick marks on the x -axis. Meteorology. Figu re 2 sh ows th e tim e-series m easu rem en ts for win d sp eed, win d direction , tem p eratu re, an d relative h u m idity for each FD. Table 1 lists th e m ean an d averages for th e m eteorological m easu rem en ts an d also in clu des th e sky con dition s du rin g th e dep loym en t p eriods (i.e., clear, m ist, overcast, or fog) an d th e estim ated Pasqu ill stability class (11). Th e win d direction in Figu re 2 an d Table 1 is relative to ch icken h ou se A (Figu re 1), wh ere a win d direction of 180° is p arallel to th e h ou se (i.e., a SSW win d direction ). Overall, th e m eteorological con dition s (win d sp eed, win d direction , tem p eratu re, relative h u m idity) were relatively con stan t du rin g each of th e FDs. Th ere is m issin g m eteorological data for p art of FD IV becau se of a p ower failu re du rin g th e dep loym en t p eriod; h owever, th e p ower failu re did n ot effect th e p assive m easu rem en ts. FDs I-VI h ad win d sp eeds th at varied from 2.6 to 5.6 m s -1, an d FD VII h ad m u ch ligh ter average win ds (0.09 m s -1) sin ce it took p lace overn igh t u n der extrem ely calm con dition s alth ou gh th e win d vector du rin g th is tim e p eriod was still “in sector”. FDs I, II, V, an d VI took p lace du rin g dayligh t h ou rs; FDs III, IV, an d VII occurred durin g n ighttim e hours. The con cen tration p rofiles m easu red in th e FDs were u se to determ in e th e NH 3 sou rce stren gth u sin g in verse m odelin g. Table 2 h as win d frequ en cy an d direction distribu tion s for location s with lon g-term m eteorological observation s an d p roxim ity to Prin cess An n e, MD, to determ in e th e typ ical win d velocities an d win d direction s for th e stu dy site in Prin cess An n e, MD. Th ese location s in clu ded Wilm in gton , DE (WDE: located 170 km N of site); Baltim ore, MD (BMD: located 150 km NW of site); an d Norfolk, VA (NVA: located 180 km SSW of site). Su rface m eteorological d ata (www.webm et.com ) for th e years 1961-1990 were u sed to gen erate frequ en cy distribu tion s u sin g WRPLOT View software from Lakes En viron m en tal Software (www. weblakes.com ). Table 2 sh ows th e frequ en cy distribu tion s for th ese location s an d for th e tim e p eriods relevan t to th is stu dy. Th e win d sp eed frequ en cy distribu tion for th e FDs in th is study is also shown in Table 2, an d the distribution is relatively close to the win d sp eed distribution s for the n earby location s, alth ough this study had a relatively sm all sam p lin g (i.e., seven FDs). Th e tem p eratu res du rin g th is stu dy ran ged from 23.6 to 31.5 °C. Th is ran ge of tem p eratu res is n ot very large bu t is con sisten t with th e lon g-term tem p eratu re observation s for th e stu dy p eriod in n earby Salisbu ry, MD (20 km N of site). The lon g-term observation s in Salisbury, MD, had m ean tem p eratu res, m ean h igh tem p eratu res, an d m ean low tem p eratu res of 21, 27, an d 14 °C for May 31-Ju n e 5 an d 26, 31, an d 20 °C for Ju ly 22-27. Th e m ean yearly tem p eratu re is 13 °C, ran gin g from m ean daily tem p eratu res of 2 °C in TABLE 2. Frequency Distributions for Wind Speed and Wind Direction for Nearby Loacations w ith Long-Term Meteorological Observations a w ind speed (m s-1) <0.51 (%) 0.51-1.8 (%) 1.8-3.34 (%) Wilm ington, DE Baltim ore, M D Norfolk, VA 5 3 2 3 6 3 35 37 21 Wilm ington, DE Baltim ore, M D Norfolk, VA 8 4 3 4 6 6 Wilm ington, DE Baltim ore, M D Norfolk, VA 6 4 4 4 6 4 Wilm ington, DE Baltim ore, M D Norfolk, VA 14 9 9 7 11 8 field deploym ents 14 0 a 3.34-5.4 (%) w ind direction >11.06 (%) SSE (%) S (%) SSW (%) SW (%) 5 1 1 3 0 0 0 5.7 5.6 5.2 8.6 7.2 10.4 6.0 6.5 11.2 8.5 8.7 13.9 9.3 7.9 7.0 36 41 27 Time Period: July 22-27 37 15 0 38 11 0 41 22 1 0 0 0 6.8 7.5 5.4 10.2 9.5 10.0 6.3 6.3 10.1 8.2 7.2 11.9 6.8 8.4 7.5 38 40 24 Time Period: M ay 31-July 27 36 16 1 37 12 1 40 25 2 0 0 0 6.5 6.8 5.4 10.0 8.4 10.6 6.3 7.1 12.0 8.6 8.4 14.1 8.5 8.7 6.8 M ay 31-July 27 (midnight to 6:00 am) 22 4 0 0 22 3 0 0 35 16 1 0 2.9 3.0 4.1 8.3 6.5 12.9 5.9 5.9 17.0 6.7 8.4 18.9 7.2 10.0 6.9 Time Period: 38 37 41 Time Period: 54 54 31 29 5.4-8.49 (%) 43 8.49-11.06 (%) M ay 31-June 18 15 30 14 0 W SW (%) 0 The data sets used for each of the locations w ere from 1961 to 1990. Jan u ary an d 25 °C in Ju ly. Overall, th e tem p eratu res du rin g th e stu dy p eriod were con sisten t with su m m er tem p eratu res. Overall, th e FDs in th is stu dy h ad win d sp eeds an d tem p eratures that were represen tative of sum m er m eteorological con dition s in th is region . Concentration Profiles. Tables with th e NH 3 con cen tration data for all of th e FDs are in clu ded in th e Su p p ortin g In form ation (see Su p p ortin g In form ation Tables 1-8). Th e m ean down win d con cen tration s in th e p lu m es ran ged from 44.3 to 262 µg m -3 with con cen tration s from 22.1 to 1740 µg m -3 for all of th e FDs. Th ese con cen tration s are h igh relative to am bien t NH x m easu rem en ts in ru ral an d u rban areas arou n d th e Ch esap eake Bay th at are typ ically on th e order of 1-3 µg m -3 (14). Th e con cen tration s in th e p lu m e are relatively low as com p ared to recom m en ded exp osu re levels by th e Nation al In stitu te of Occu p ation al Safety an d Health (NIOSH) th at recom m en d th at workp lace air sh ou ld n ot exceed 25 p p m NH 3 (15 625 µg m -3) over a 10-h workday or 40-h work week. Th e u p win d NH 3 con cen tration s ran ged from 0.2 to 59.8 µg m -3 with a m ean of 16.7 µg m -3 for all th e FDs (see Su p p ortin g In form ation Table 1). Th e m axim u m valu e of 59.8 µg m -3 was for FD VII du rin g calm con dition s. Th e u p win d con cen tration was su btracted from all of th e down win d p lu m e con cen tration s for each FD sin ce th e GPM was for con cen tration s above the backgroun d con cen tration s. Th e ratio of th e m axim u m con cen tration to th e m in im u m con cen tration s in th e down win d sam p lin g grid for each FD ran ged from 18 to 60 except for FD VII, which had sign ifican tly less variability in the con cen tration s due to the m eteorological con dition s (see Su p p ortin g In form ation Table 1). Th e vertical NH 3 con cen tration p rofiles sh owed valu es th at were typ ically h igh er n earer th e su rface for all of th e FDs excep t for FD VII, wh ich h ad lower con cen tration s n ear th e su rface. Th e con cen tration s p rofiles for FDs I-VII were con sisten t with a p lu m e origin atin g from th e ch icken h ou se, an d in verse GPMs were p erform ed for th ese FDs to determ in e th e sou rce stren gth s. Inverse Gaussian Plum e Modeling. Table 3 lists th e in verse GPM resu lts for each th e field dep loym en ts. Th e GPM varied win d sp eed, win d direction , h ou se len gth , sou rce h eigh t, an d sou rce stren gth u n til it m in im ized th e cost fu n ction . Th e m odel was also ru n u sin g differen t Pasqu ill stability classes, an d th e m odel ru n with th e lowest CF is in boldface typ e in Table 3. Figu re 3 sh ows an exam p le of th e con cen tration den sity p lots p rodu ced by th e GPM for each of th e sam p lin g h eigh ts alon g with th e corresp on din g observed gas-p h ase NH 3 con cen tration s for FD IV. Th e m easu red con cen tration s in Figu re 3 h ave th e u p win d NH 3 con cen tration su btracted sin ce th e p lu m e m odel is for NH 3 con cen tration s above th e am bien t (i.e., u p win d) levels. FD IV was ch osen as an exam p le sin ce it h ad a CF th at was n ear th e m edian of all th e dep loym en ts. Both ch icken h ou ses A an d B are in clu ded in Figu re 3 sin ce th ey are both in clu ded in th e m odel calcu lation s. Figu re 3 clearly sh ows th e stron g vertical an d h orizon tal gradien ts of NH 3 con cen tration s in th e p lu m e an d is con sisten t with th e m easu red NH 3 con cen tration s for th is dep loym en t. For exam p le, th e m ast n earest th e en d of th e ch icken h ou se h ad m easu red NH 3 con cen tration s (above backgrou n d) ran gin g from 772 µg m -3 at 1 m to 55 µg m -3 at 10 m . Su p p ortin g In form ation Table 5 lists th e associated CF for each of th e sam p lin g location s for FD IV (th e average of th is CF, 3.6, is given in Table 3). The m odel con sisten tly con verged on values of win d speed an d win d direction th at were con sisten t with th e m ean win d sp eed an d win d direction m easu rem en ts con siderin g th e accu racy of th ese sen sors (com p arin g Table 3 resu lts in boldface typ e with Table 1 m easu rem en ts). Th e m odel also con verged on sou rce release h eigh ts between 0.9 an d 3.4 m th at are con sisten t with th e size of th e ch icken h ou se for FDs I-VI. FD VII h ad a sign ifican tly h igh er sou rce release h eigh t of 14.5 m . FD VII also h ad a vertical gradien t with lower con cen tration s n ear th e su rface th at was op p osite of th e vertical gradien ts observed for FDs I-VI. Th is h igh sou rce release h eigh t for FD VII is p robably do to th e bu oyan cy of th e ven tilated air du rin g th e relatively calm an d cooler overn igh t con dition s du rin g FD VII. Th e ven tilated air h ad an estim ated tem p eratu re of 30 °C as com p ared to th e am bien t tem p erature of 24 °C an d would n ot have un dergon e as rap id of dilu tion as th e p reviou s FDs becau se of th e relatively calm overn igh t con dition s. Th e sou rce h eigh t of 14.3 m wou ld rep resen t an effective sou rce h eigh t. Th e effective sou rce h eigh t is well above th e h igh est sam p lin g h eigh t (10 m ). Th is bu oyan cy wou ld also be a sou rce of tu rbu len ce th at cou ld in flu en ce th e disp ersion of NH 3 th at is n ot accou n ted for by th e disp ersion p aram eterization in th e m odel. All of th ese factors p robably con tribu te to th e poor ability of the GPM to m odel the observation s as in dicated by th e relatively h igh CF (27.0) for FD VII. Th e resu lts are still TABLE 3. Inverse Gaussian Plume Model Results To Determine the Ammonia Source Strength for Field Deployments I -VI a stability classc w ind speed (m s-1) w ind directiond (deg) house start (m) house length (m) I I I I A B C D 6.4 3.9 3.9 3.6 205 187 183 184 55 36 0 1 9 28 64 63 II II II A B C 3.0 2.8 2.7 185 176 180 20 25 0 III III III III A B C D 3.8 3.4 3.4 3.3 201 179 180 180 IV IV IV A B C 2.9 2.8 2.5 V V V A B C VI VI VI VII VII VII VII FDb height of release (m) cost function source strength (g s-1) 0.6 3.0 2.1 3.2 27.5 13.7 14.0 26.7 0.18 0.13 0.15 0.13 44 39 64 0.4 1.7 2.2 29.7 1.8 9.0 0.07 0.11 0.09 51 33 0 1 13 31 64 63 0.7 2.0 1.4 3.3 28.3 6.2 6.0 22.4 0.09 0.14 0.15 0.12 181 182 187 8 3 0 56 61 64 0.4 0.9 3.2 22.4 3.6 13.4 0.24 0.25 0.15 6.2 6.0 6.0 186 183 184 23 32 0 41 32 64 0.6 2.7 2.8 25.9 2.9 8.8 0.22 0.28 0.25 A B C 2.2 2.0 2.1 171 177 180 14 16 0 50 48 64 0.3 3.4 3.7 25.7 17.4 18.1 0.16 0.11 0.11 A B C D 0.3 0.4 0.4 0.4 201 209 205 204 61 45 13 0 3 19 51 64 16.0 10.9 9.7 7.7 39.9 28.5 27.0 31.3 0.028 0.034 0.030 0.031 a The m odel w as run using different Pasquill atm ospheric stabiltity classes for each run. The results in bold represent the m odel param eters that best fit the observations. b FD, field deploym ent. c The letters represent various Pasquill stability classes: A, extrem ely unstable; B, m oderately unstable; C, slightly unstable; D, neutral. d Wind direction is relative to axis of the chicken house. For exam ple, for a w ind direction of 180°, the research trailer w ould be directly dow nw ind of the chicken house. p resen ted in Tables 3 an d 4; h owever, th e con fiden ce in th e sou rce stren gth valu e for th is case is m u ch lower th an th e oth er cases an d difficu lt to qu an tify. Th e len gth of th e lin e sou rce was allowed to vary in th e m odel; h owever, th e en d sou rce of th e lin e sou rce was always fixed at th e n orth en d of th e ch icken h ou se. Th e len gth of th e ch icken h ouse was obviously kn own ; however, the reason for allowin g th e sou rce len gth to vary was becau se th e ven tilation of th e side-wall ven tilated h ou se was n ot well con strain ed. The m odel con verged on source len gths between 28 an d 64 m . Th e GPM was also ru n for several atm osp h eric stability classes (see Table 3) to in vestigate if th e m odel wou ld con verge to th e estim ated atm osp heric stability class for each case (see Table 1) an d to also in vestigate th e sen sitivity of th e resu ltin g sou rce stren gth s for differen t stability classes (an d th erefore variou s disp ersion coefficien ts). Th e Pasqu ill stability class was estim ated for each h ou r du rin g th e sam plin g periods usin g an estim ate for solar in solation (based on latitu de, tim e of day, an d clou d cover) an d th e win d sp eed (11). Th e GPM ru n s with th e lowest CF agreed with th e estim ated Pasqu ill stability classes for FDs I, II, an d VI. Th e GPM resu lts for FD V con verged on a Pasqu ill stability class of B alth ou gh th e p redicted stability class was C. Th e GPM for FDs III an d IV con verged on stability classes on e class higher (i.e., less stable) than the estim ated classes. This higher stability class m ay reflect added tu rbu len ce (e.g., from th e fan s, win d direction ch an ges, tu rbu len ce cau sed by th e h ou se’s p rofile), wh ich wou ld in crease th e disp ersion coefficien ts th at is n ot accou n ted for by th e m eteorology. Th ere was n o defin ition of a stability class for th e m eteorological con dition s observed for FD VII. Th e sen sitivity of th e stability class on th e resu ltin g sou rce stren gth is sh own in Table 3 by com p arin g th e sou rce stren gth s for each case. Th e average ch an ge in th e sou rce stren gth is 20% wh en ch an gin g th e stability class to th e n ext h igh er or lower stability class, an d th e direction of th at ch an ge was both p ositive an d n egative. Th e ch oice of stability class h as a sign ifican t effect on th e resultin g source stren gth although a m ore rigorous sen sitivity an alysis is n ot p ossible sin ce th e stability class is n ot a con tin u ou s fu n ction . Th e u se of an oth er stability p aram eterization su ch as th e Mon in -Obu kh ov len gth wou ld be a m ore u sefu l p aram eterization to in vestigate th e sen sitivity of th e atm osp h eric stability on th e resu ltin g sou rce stren gth sin ce th is is a con tin u ou s p aram eter. For th is stu dy, h owever, th e added in sigh t of u sin g th is p aram eterization wou ld n ot be very ben eficial sin ce m ore detailed atm osp h eric stability m easu rem en ts wou ld n eed to h ave been m ade du rin g th e stu dy p eriods. Th e added u n certain ties associated with th e stu dy site (e.g., h ou se ven tilation variability, tu rbu len ce aroun d the buildin g) an d m ethodology (e.g., sam plin g several h ou rs, accu racy of p assive collectors) m ay also h ave lim ited th e u sefu ln ess of a m ore detailed atm osp h eric stability param eterization . Overall, the ability of the m odel to con verge on valu es th at are con sisten t with th e m eteorological m easu rem en ts gives som e qu alitative con fiden ce in th e sou rce stren gth s gen erated by th e in verse GPM. Th e con cen tration p rofiles gen erated by th e GPM were u sed to test th e assu m p tion in th e GPM th at n o sign ifican t dep osition occu rs to th e su rface with in th e m odelin g (an d sam p lin g) dom ain . On e m eth od for estim atin g th e dry dep osition to th e su rface is to m u ltip ly th e dry dep osition velocity, by th e area an d th e con cen tration at a referen ce h eigh t over th e m odel dom ain , an d th en sp readin g th is NH 3 deficit over th e en tire p lu m e (ref 4 an d referen ces cited th erein ). Th e dep osition of NH 3 was calcu lated for all of th e GPM resu lts by su m m in g th e dep osition for each 1 m × 1 m grid p oin t in th e m odel dom ain . Th e total dry dep osition al flu x for th e sam p lin g dom ain was in vestigated u sin g a ran ge of dep osition velocities an d u sin g th e NH 3 con cen tration at 1 m . For a dep osition al velocity (Vd ) of 2 cm s -1, th e estim ated total dry dep osition al flu x to th e su rface ou tside of th e h ou se TABLE 4. Summary of Chicken House Source Strengths and Emission Factors Determined for Each of the Field Deployments w ith the Corresponding Confidence Limts Determined from Monte Carlo Simulations FDa source strengthb,d (g of NH3-N s-1) emission factorc ,d (g of NH3-N bird-1 day-1) w eighted emission factorc ,d,e (g of NH3-N bird-1 day-1) I II III IV V VI VII 0.13 ( 0.009 0.11 ( 0.014 0.15 ( 0.023 0.25 ( 0.033 0.28 ( 0.026 0.11 ( 0.021 0.035 ( NA f 1.01 ( 0.07 0.85 ( 0.11 1.16 ( 0.18 1.94 ( 0.25 2.17 ( 0.20 0.85 ( 0.16 0.27 ( NA f 0.56 ( 0.04 0.32 ( 0.04 0.44 ( 0.07 1.65 ( 0.22 1.64 ( 0.15 0.43 ( 0.08 0.14 ( NA f a FD, field deploym ent. b Source strength is for a 11 500 capacity chicken house. c The em ission factor is calculated using a flock size of 11 155 birds. (This assum es a 3% m ortality rate and an original flock size of 11 500 birds.) d Confidence lim its show n are ( one standard deviation determ ined from the M onte Carlo sim ulations. e The w eighted em ission factor accounts for the nonlinear increase in am m onia em issions over the grow -out period of the flock and w ould represent the average em ission factor over the entire 6-w eek grow -out period. f The results of the M onte Carlo sim ulations for FD VII resulted in a non-norm al distribution of source strengths due to the low w ind speeds. FIGURE 3. Ammonia concentration density plots generated by the Gaussian plume model for field deployment IV at the three height levels w here the passive samplers w ere located. The tw o chicken houses are show n along w ith the location of the sampling tow ers and the corresponding measurements of gas-phase NH3 concentrations (in µg m-3). ran ged from 2.9% to 8.7% of th e en tire sou rce stren gth for all of con cen tration p rofiles p rodu ced by th e GPM. Th is estim ate is p robably a m axim u m sin ce th e dep osition al flu x n ear th e su rface wou ld resu lt in decreasin g NH 3 con cen tration s n ear th e su rface an d as a con sequ en ce a decrease in th e dep osition al flu x. Th e dep osition velocity for th is calcu lation is also p robably h igh sin ce th e dep osition velocity is gen erally lower th an 2 cm s -1. Th e in clu sion of dry dep osition to th e GPM wou ld lead to h igh er NH 3 em ission rates from th e ch icken h ou se sin ce th e dry dep osition down win d of th e h ou se wou ld resu lt in an NH 3 sin k an d th erefore th ere wou ld n eed to be a h igh er em ission rate from th e ch icken h ou se to offset th is NH 3 sin k for a given NH 3 p lu m e. Th e residu als (m odeled - observed con cen tration s) also did n ot sh ow an y system atic tren d of th e m odeled con cen tration s at th e su rface bein g lower th at th e oth er h eigh ts, wh ich wou ld be exp ected if th ere was sign ifican t dry dep osition (th e Su p p ortin g In form ation tables sh ow th e m odeled an d observed con cen tration s for each sam p lin g location an d case). Overall, th e assu m p tion th at dry dep osition within the m odel dom ain is n ot sign ifican t is a reason able assu m p tion for th e cases in th is stu dy. It is difficult to determ in e relation ships between the source stren gth valu es for th e variou s dep loym en ts an d th e gen eral con dition s observed (e.g., m eteorological con dition s, chicken farm op eration ) du e to th e lim ited observation s. Th e resu lts seem to sh ow th at th e sou rce stren gth s were h igh er du rin g Ju ly th an late May/ early Ju n e (om ittin g th e an om alou s FD VII). Th is m ay be du e to differen t ven tilation p ractices du rin g th ese differen t tim es alth ou gh th e tem p eratu res were very sim ilar. Table 4 su m m arizes th e sou rce stren gth s for each of th e FDs an d also lists th e con fiden ce lim its for th ese sou rce stren gth values based on Mon te Carlo sim ulation s. The source len gth s u sed in th e Mon te Carlo sim u lation s were a ran dom n u m ber between 11 an d 64 m , sin ce in clu din g sou rce len gth s that were less than 11 m resulted in a n on -n orm al distribution of sou rce stren gth s for th e Mon te Carlo sim u lation s. A sh ort sou rce len gth (e10 m ) resu lted in a stron g p oin t sou rce at th e n orth en d of th e h ou se th at cau sed stron g NH 3 gradien ts n ear th e sam p lin g location s; th erefore, th e in verse GPM con verged on sou rce stren gth s with an om aly h igh cost fu n ction s as com p ared to Mon te Carlo sim u lation s with sou rce len gth s between 11 an d 64 m . Overall th ese sou rce stren gth s rep resen t atm osp h eric NH 3 em ission rates from com m ercial side-wall ven tilated ch icken farm s on th e Delm arva Pen in su la, u sin g m easu rem en t of th e actu al p lu m e con cen tration p rofile down win d of th e ch icken h ou se. Th e sam p lin g m eth odology an d u se of an in verse GPM sh ow th e ability of th is com bin ation to ch aracterize sou rces of NH 3 from ch icken h ou ses. Am m onia Em ission Factors. Com p arison s can be m ade u sin g th e sou rce stren gth s determ in ed in th is stu dy to p reviou sly p u blish ed em ission factors for ch icken s. Table 4 con verts th e NH 3 sou rce stren gth s to NH 3 em ission factors u sin g a flock size of 11 155 birds for th e ch icken h ou se u sed in th is stu dy (th is assu m es a 3% m ortality rate). Table 4 also calcu lates a weigh ted em ission factor u sin g em ission rate versu s tim e data observed for a com m ercial ch icken h ou se in th e U.K. (15). Dem m ers et al. (15) observed a stron g n on lin ear in crease in NH 3 em ission s as th e birds age in creased. Th e p lot from Dem m ers et al. (15) of cu m u lative em ission s versu s tim e was u sed to determ in e th e fraction of NH 3 em itted each day with resp ect to th e total em ission s an d th en u sed to weigh t th e em ission factors in Table 4. For exam p le, it is estim ated th at 2.8% of th e total em ission s are released over th e cou rse of 1 day wh en th e birds are 29 days TABLE 5. Literature Values of NH3 Emission Factors for Broilers emission factor (g of NH3-N bird-1 day-1) ref 0.72 0.46 0.35 0.27 0.44a 0.80b 16 7 17 18 10 10 a Integrated over the entire grow -out period. out period. b Last 2 w eeks of grow - old. On e day rep resen ts 1/ 42th (2.38%) of th e en tire growou t p eriod, so for FD V in Table 4, th e em ission factor of 1.94 g of NH 3-N bird -1 day-1 was m u ltip lied by th e ratio (0.0238/ 0.0280) to estim ate the average em ission factor over the en tire 6-week grow-ou t p eriod. Dem m ers et al. (15) also observed NH 3 released from th e ch icken h ou se after th e ch icken s h ad been rem oved bu t before th e litter was rem oved. Th e weigh ted em ission factors calcu lated in Table 4 do n ot take in to accou n t an y NH 3 released from th e litter after th e ch icken s are rem oved from th e h ou se. Table 5 lists NH 3 em ission factors from several differen t literatu re sou rces. Overall, th e literatu re valu es in Table 5 are on average lower than the NH 3 em ission factors estim ated by th is stu dy wh ich h ad a m ean of 1.18 g of NH 3-N bird -1 day-1. Th e m ean valu e of 1.18 g of NH 3-N bird -1 day-1 corresp on ds to 50 g of NH 3-N bird -1 released to th e atm osp h ere du rin g a 6-week grow-ou t p eriod. Usin g th e weigh ted em ission factors in Table 4 resu lts in a m ean of 0.74 g of NH 3-N bird -1 day-1 th at corresp on ds to 31 g of NH 3-N bird -1 released to th e atm osp h ere du rin g a 6-week grow-out period. Therefore the values in this study were about 1.5 tim es greater th an th ese literatu re valu es wh en u sin g th e weigh ted em ission factors. Th e differen ces between th e valu es from th is stu dy an d th e valu es in Table 5 m ay be du e to several reason s. First, m an y of th e valu es from Table 5 are from stu dies in Eu rop e wh ere differen t an im al h u sban dry m eth ods are ap p lied to th e raisin g of broilers. Also, becau se NH 3 em ission rates are tem p eratu re-dep en den t, th e valu es determ in ed in th is stu dy durin g sprin g an d sum m er (m ean tem perature ran ge of 23.631.5 °C) m igh t be exp ected to be greater th an th e Eu rop ean stu dies, wh ere cooler tem p eratu res typ ically p revail. Th e tem p eratu re an d win d sp eed are also exp ected to be im p ortan t m eteorological factors th at determ in e th e NH 3 em ission rates from ch icken h ou ses. Th ere are m an y oth er factors (e.g., age of flock, litter m oistu re, diet) th at are also im p ortan t bu t are n ot directly related to th e m eteorology. Tem p eratu re is a factor th at is im p ortan t to th e ven tilation rate of th e h ou se. Th ere is a m in im al ven tilation rate to keep NH 3 con cen tration s with in th e h ou se below a level th at is h arm fu l to th e birds, an d as ou tside tem p eratu re in creases, so does th e ven tilation rate to keep th e birds cool (m isters also m ay be u sed as th e tem p eratu re in creases). Th e release of NH 3 with in th e h ou se (in clu din g th e litter) m ay also in crease with in creasin g tem p eratu res. Win d sp eed m ay also effect NH 3 em ission rates sin ce win ds m ay in flu en ce th e ven tilation rates of th e ch icken h ou se esp ecially for sidewall ven tilated h ou ses wh ere win ds directly con tribu te to th e ven tilation of th e h ou se. Th e effect of win ds are p robably less im p ortan t for tu n n el-ven tilated h ou ses wh ere th e fan s dom in ate th e ven tilation rate. Th e valu es from th is stu dy m ay also be h igh er becau se th e birds in th is stu dy were n ear th e en d of th eir 6-week grow-ou t p eriod for all of th e FDs, alth ou gh we attem p ted to correct for th is by u sin g a weigh ted em ission factor (see above). Roadm an et al. (10) in vestigated a tu n n el-ven tilated h ou se on th e Delm arva Pen in su la an d fou n d an average NH 3 em ission rate of 0.8 g of NH 3 bird -1 day-1 over th e last 2 weeks of th e grow-ou t p eriod. Th is was abou t twice th e rate in tegrated over th e en tire 6-week grow-ou t p eriod an d in closer agreem en t to th e rates observed in th is stu dy. Biases in th is stu dy m ay also be p resen t sin ce FDs to determ in e NH 3 em ission rates were on ly p erform ed wh en th e win d direction was ap p roxim ately from th e SSW. Th e win d sp eed frequ en cy distribu tion s an d tem p eratu res were con sisten t with su m m er con dition s for th e seven stu dy p eriods so th e biases sh ou ld be m in im al for th ese factors. However waitin g for win ds from a certain sector m ay in flu en ce th e resu lts sin ce th ese win ds m ay be th e resu lt of certain m eteorological system s th at also brin g distin ct air m asses. For exam p le, NH 3 em itted from th e h ou se m ay u n dergo gas to p article con version if th e am bien t air m ass h as acidic gas-p h ase an d aerosol-p h ase sp ecies p resen t. However, th e con cen tration s of NH 3 close to th e h ouse wh ere th e sam p lin g array is located are extrem ely h igh relative to typ ical atm osp h eric con cen tration s of NH 3 an d oth er gasp hase an d p articulate-p hase sp ecies that m ay react with NH 3. Overall, th ese in teraction s are p robably n ot im p ortan t with in th e p lu m e wh ere th e sam p lin g array was located bu t will becom e m ore im p ortan t as th e NH 3 is disp ersed. Th e atm osp h eric dep osition an d lifetim e of NH 3 will be effected if NH 3 u n dergoes gas to p article reaction s. Broiler Production on the Delm arva Peninsula and Im plications for the Chesapeake and Delaware Bays. Th e source stren gths determ in ed by this study can be extrapolated to the en tire Delm arva Pen in sula. Ap p roxim ately 587 000 000 broilers are p rodu ced an n u ally on Pen in su la (Delm arva Pou ltry In du stry, In c., www.dp ich icken .org). Mu ltip lyin g th e an n u al broiler p rodu ction by th e weigh ted em ission factor (31 g of NH 3-N bird -1) yields an estim ate of 18.2 × 10 6 kg of NH 3-N yr -1 em itted to th e atm osp h ere from ch icken h ou ses on the Delm arva Pen in sula. When com p ared with the curren t estim ates of total atm osp h eric N dep osition (n itrate + NH x ) to th e Delaware (34 × 106 kg of N yr -1) an d Ch esap eake Bay (177 × 106 kg of N yr -1) watersh eds (1) an d assu m in g a sizable fraction of th e local em ission s are dep osited locally, NH 3 em ission s from p ou ltry op eration s on ju st th e Delm arva Pen in sula would represen t a sign ifican t addition al con tribution . Un fortu n ately, th e cu rren t atm osp h eric dep osition n etworks (NADP, CASTNET, AirMON) do n ot m easu re gasp h ase NH 3 con cen tration s or dep osition , th us it is im p ossible to gau ge th e relative con tribu tion . Acknow ledgments Fu n din g for th is p roject was p rovided by th e U.S. EPA, Ch esap eake Bay Office. We th an k Joel E. Baker for h is com m en ts an d h elp with th is stu dy. We th an k th e staff at th e Un iversity of Marylan d Eastern Sh ore an d esp ecially Jim Jardin e, wh o assisted u s with logistics at th e farm . We also ap p reciate th e th ou gh tfu l com m en ts an d su ggestion s p rovided by th e reviewers. Supporting Information Available Eigh t tables. Th is m aterial is available free of ch arge via th e In tern et at h ttp :/ / p u bs.acs.org. Literature Cited (1) Castro, M. S.; Driscoll, C. T. Atm osp h eric n itrogen dep osition to estu aries in th e m id-Atlan tic an d n orth eastern Un ited States. En viron . Sci. Tech n ol. 2002, 36 (15), 3242-3249. (2) Sin gles, R.; Su tton , M. A.; Weston , K. J. A m u lti-layer m odel to describe th e atm osp h eric tran sp ort an d dep osition of am m on ia in Great Britain . Atm os. En viron . 1998, 32 (3), 393-399. (3) Galp eriin , M. V.; Sofiev, M. A. Th e lon g-ran ge tran sp ort of am m on ia an d am m on iu m in th e North ern Hem isp h ere. Atm os. En viron . 1998, 32, 373-380. (4) Asm an , W. A. H. Factors in flu en cin g local dry dep osition of gases with sp ecial referen ce to am m on ia. Atm os. En viron . 1998, 32, 415-421. (5) Fowler, D.; et al. Th e m ass bu dget of atm osp h eric am m on ia in woodlan d with in g 1 km of livestock bu ildin gs. En viron . Pollu t. 1998, 102, 343-348. (6) Ch im ka, C. T.; Galloway, J. N.; Cosby, B. J. Am m on ia an d th e Ch esapeak e Bay Airsh ed; Scien tific an d Tech n ical Advisory Com m ittee of th e Ch esap eake Bay Program : 1997. (7) Battye, R.; et al. Developm en t an d Selection of Am m on ia Em ission Factors; Office of Research an d Developm en t, U.S. EPA: Washin gton , DC, 1994. (8) Scu dlark, J. R.; Ch u rch , T. M. A Com preh en sive Re-Evalu ation of th e In pu t of Atm osph eric Nitrogen to th e Reh oboth an d In dian River Estu aries; Delaware Cen ter for th e In lan d Bays: 1999; p 36. (9) Dem m ers, T. G. M.; et al. First exp erien ces with m eth ods to m easu re am m on ia em ission s from n atu rally ven tilated cattle bu ildin gs in th e U.K. Atm os. En viron . 1998, 32 (3), 285-293. (10) Roadm an , M. J.; et al. Validation of Ogawa p assive sam p lers for th e determ in ation of gaseou s am m on ia con cen tration s in agricu ltu ral settin gs. Atm os. En viron . 2003, 32, 2317-2325. (11) Sein feld, J. H.; Pan dis, S. N. Atm osph eric Ch em istry an d Ph ysics: From Air Pollu tion to Clim ate Ch an ge; Joh n Wiley an d Son s: New York, 1998. (12) Fisch er, H. B.; et al. Mixin g in In lan d an d Coastal W aters; Academ ic Press: Orlan do, FL, 1979. (13) Press, W. H.; et al. Nu m erical recipes in Fortran 77: Th e art of scien tific com pu tin g, 2n d ed.; Cam bridge Un iversity Press: New York, 1992. (14) Larsen , R. K.; Stein bach er, J. C.; Baker, J. E. Am m on ia exch an ge between th e atm osp h ere an d th e su rface waters at two location s in th e Ch esap eake Bay. En viron . Sci. Tech n ol. 2001, 35 (24), 4731-4738. (15) Dem m ers, T. G. M.; et al. Am m on ia em ission s from two m ech an ically ven tilated UKlivestock bu ildin gs. Atm os. En viron . 1999, 33 (2), 217-227. (16) Misselbrook, T. H.; et al. Am m on ia em ission factors for UK agricu ltu re. Atm os. En viron . 2000, 34 (6), 871-880. (17) EEA. Join t EMEP/ CORINAIR Atm osph eric Em ission In ven tory Gu idebook , 3rd ed.; Eu rop ean En viron m en t Agen cy: Cop en h agen , 2001. (18) U.S. EPA. Review of em ission factors an d m eth odologies to estim ate am m on ia em ission s from an im al w aste h an dlin g; U.S. EPA: Research Trian gle Park,NC, 2002. ES0345874