Estuaries and Coasts Vol. 30, No. 3, p. 469M-81 June 2007
Threshold Effects of Coastal Urbanization on Phragmites australis (Common Reed) Abundance and Foliar Nitrogen Chesapeake Bay
in
RYAN S. KING L*, WILLIAM V . D E L U C A 2 ' t , DENNIS. F. W H I G H A M 2, a n d PETER P. MARRAZ~
1 CenterJbr Rese~woir and Aquatic ,Systems Research, Department of Biology, Baytor University, One Bear Place #97388, Waco, Texas 76798-7388
2 Smithsonian Environmental Research Center, Box 28, Edgewater, Maryland 21037
ABSTRACT: T h e invasion o f N o r t h American tidal m a r s h e s by Phragmites anstralis, or c o m m o n r e e d , is a large-scale ecological p r o b l e m that h a s b e e n primarily s t u d i e d at small spatial scales. Previous local-scale studies have p r o v i d e d evidence that the e x p a n s i o n of Phragmites is facilitated by disturbance a n d i n c r e a s e d n i t r o g e n (N) associated with agricultural a n d u r b a n - s u b u r b a n (developed) l a n d u s e s along wetland-upland borders. We tested t h e generality of previous findings across a larger spatial scale a n d wider range o f e n v i r o n m e n t a l conditions in C h e s a p e a k e Bay, t h e largest estuarine ecosystem in t h e USA. We s a m p l e d 90 tidal wetlands n e s t e d within 30 distinct s u b e s t u a r i n e w a t e r s h e d s a n d e x a m i n e d t h e relationship b e t w e e n l a n d u s e a n d Phragrnites a b u n d a n c e a n d foliar N, an indicator o f nitrogen availability. We e s t i m a t e d l a n d u s e adjacent to wetland b o r d e r s a n d within s u b e s t u a r y w a t e r s h e d s a n d e x p l o r e d the i m p o r t a n c e of spatial proximity by weighting l a n d u s e by its distance f r o m t h e wetland b o r d e r or s u b e s t u a r y shoreline, respectively. Regression tree a n d c h a n g e p o i n t analyses revealed that Phragmites a b u n d a n c e sharply i n c r e a s e d in almost every wetland w h e r e d e v e l o p m e n t adjacent to b o r d e r s e x c e e d e d 15%.
W h e r e d e v e l o p m e n t was < 1 5 % b u t natural l a n d cover at t h e n e a r the s-ubestuary s h o r e l i n e was low ( < - - 3 5 % ) , Phragmites was a b u n d a n t , suggesting that wetlands in highly m o d i f i e d w a t e r s h e d s also were susceptible to invasion, regardless of l a n d u s e adjacent to wetlands. Phragmites foliar N was markedly elevated in w a t e r s h e d s with > 1 4 - 2 2 % shoreline d e v e l o p m e n t , t h e s a m e level of d e v e l o p m e n t that c o r r e s p o n d e d to high levels of invasion. O u r results suggest that d e v e l o p m e n t n e a r wetlands is at least partially responsible for patterns of invasion across Chesapeake Bay. Larger-scale p h e n o m e n a , s u c h as n i t r o g e n pollution at t h e watershed-subestuary scale, also m a y be facilitating invasion. Urbanization n e a r coastlines appears to play an i m p o r t a n t role in the invasion success o f Phragmites in coastal wetlands o f C h e s a p e a k e Bay a n d probably m u c h of eastern N o r t h America.
Intro duction
A critical issue in ecology is w h e t h e r processes identified f r o m small-scale, controlled mechanistic experiments can generate patterns across m u c h larger spatial scales (e.g., W o o t t o n 2001; Steele and F o r r e s t e r 2005). T h e rapid invasion o f N o r t h A m e r i c a n tidal marshes by Phragmites australis
(Cav.) T r i n ex Steud (hereafter, Phragmites), or c o m m o n reed, is an excellent example o f a large- scale ecological p r o b l e m that has b e e n studied primarily at small spatial scales. Although Phragmites is native to N o r t h America, a nonnative genotype is believed to be primarily responsible tor its rapid expansion (Saltonstall 2002; Vasquez et al. 2005).
This large, clonal grass was historically c o n f i n e d to
* C o r r e s p o n d i n g author; tele: 254/710-2150; fax: 254/710- 2969; e-maih [email protected]
+ C u r r e n t address: H o l d s w o r t h N a t u r a l Resources Center, D e p a r t m e n t of Natural Resources Conservation, University o f Massachusetts, A m h e r s t , Massachusetts 01003.
C u r r e n t address: Srnithsonian Migratory Bird Center, Nation- al Zoological Park, 3001 C o n n e c t i c u t A v e n u e NW, Washington, D.C. 20008.
h i g h marsh fringes with relatively low salinity (Hellings and Gallagher 1992), b u t now aggressively- invades lower elevations once t h o u g h t to be too physiologically stresstul to support dense popula- tions (Amsberry et al. 2000; Burdick and Konisky 2003). Phragmites has b e c o m e a d o m i n a n t species across a range o f wetland habitats, resulting in the displacement o f native m a c r o p h y t e communities ( C h a m b e r s et al. 1999; Meyerson et al. 2000;
M i n c h i n t o n et al. 2006), d e g r a d a t i o n of habitat tbr wildlife (Benoit and Askins 1999; Weinstein and Balletto 1999), a n d alteration of ecosystem pro- cesses (Windham and E h r e n f e l d 2003; W i n d h a m and Meyerson 2003).
A series of mechanistic studies c o n d u c t e d primar- ily in small p o c k e t marshes at a local scale in Narragansett Bay-, R h o d e Island, have p r o v i d e d compelling evidence that the expansion o f Phrag- mites is tacilitated primarily by disturbance and n i t r o g e n (N) e n r i c h m e n t associated with agricultur- al and urban-suburban (developed) land uses along wetland-terrestrial borders. In these salt marshes, w h i c h s p a n n e d a n a r r o w r a n g e o f salinities
9 2007 Estuarine Research Federation 469
470 R . S . King et al,
(25-30%0), the extent of d e v e l o p m e n t along wet- land-terrestrial b o r d e r s e x p l a i n e d over 90% of the variation in ~he a b u n d a n c e o~:
PAra~g~m-i~c,s
a n d was a t t r i b u t e d m the c o m b i n e d e f f e c t of physical d i s t u r b a u c e (Minchinton a n d Bertness 2003) mid e u t r o p h i c a t i o n (Bertness et at.2002)
related to s h o r e l i n e d e v e l o p r a e n t . Salt m a r s h e s with h i g h levels o f adjacent d e v e l o p e d land were shown to h a v e e l e v a t e d N availability in c o m p a r i s o n to m a r s h e s with u n d e v e l o p e d shorelines, p r e s u m a b l y d u e to r e m o v a l o f vegetative buffiers b e t w e e n wetland b o r d e r s a n d u p l a n d source areas (Bertness et al. 2002; Sillimm~ a n d Bertness 2004). In these s a m e marshes, M i n c h i n t o n a n d Bertness (2003) showed t h a t physical d i s t u r b a n c e o f the vegetation matrix a n d N e n r i c h m e n t often associated with shoreline d e v e l o p m e n t tacilitaIed initial establish- m e n t a n d growth ofPbczg,n~ites.
Adjacent develop- m e n t was shown to reduce wetland soil salinity (Silliman a n d g e r t n e s s 2004), which m a y also s u p p o r t invasion by r e d u c i n g physiological suess (Burdick et at. ~001).Phragmites
invades m a r s h e s e x p o s e d to thll-strength sea water via clonal in- tegration (Amsberry et al. 2000), so the ultimate i m p o r t a n c e o f salinity to the invasion success is s o m e w h a t equivocal.Tidal wetlands o i o t h e r eastern USA estuaries, particularly C h e s a p e a k e Bay, have also e x p e r i e n c e d a p p a r e n t increases in the a b u n d a n c e of
Phragmites
in r e c e n t decades (reviewed by Rice et al. 2000). In contrast to the intensive efforts in R h o d e Island, the e x t e n t o f
Phragmites
invasion of C h e s a p e a k e Bay tidal wetlands a n d its linkages to a n t h r o p o g e n i c factors have not b e e n e x p l o r e d in detail. T h e C h e s a p e a k e Bay watershed is rapidly urbanizing a n d o n e o f the t3stest growing coastal regions in N o r t h A m e r i c a ( C h e s a p e a k e Bay P r o g r a m 2006) a n d r e p r e s e n t s a n a r e a at increasing risk t o r coastal w e t l a n d d e g r a d a t i o n . C h e s a p e a k e Bay- has also e x p e r i e n c e d significant cultural e u t r o p h i c a t i o n in t h e p a s t two c e n t u r i e s ( B o e s c h et al. 2001), primarily caused by p o i n t a n d n o n p o i n t source N inputs associated with agricultural a n d u r b a n (de- veloped) lands ( J o r d a n et al. 1997, 2003). Given the mechanistic relationships b e t w e e nPhragmites
a n d d e v e l o p m e n t r e p o r t e d elsewhere, the increase in a n t h r o p o g e n i c N a n d shoreline disturbances caused by agricuhural a n d m-ban-suburban devel- o p e d lands may be at Ieast partially responsible for the e x p a n s i o n o f
.Pgrag,#~ites
in C h e s a p e a k e Bay.Because spatial .patterns of land use s m r o u n d i n g C h e s a p e a k e Bay tidal wetlands vary markedly at local a n d subwatershed scales ( D e L u c a et at. 2004;
King et al. 2005a),
Phragmites
a b u n d a n c e may also exhibit spatial variability- in its distribution that c o r r e s p o n d s to p r o x i m a l l a n d use at o n e or multiple scales.In this study, we t e s t e d w h e t h e r
Phrag,'t,,~tcs
a b u n d a n c e a n d tbliar N are linked to ~he a m o u n t o f d e v e l o p e d lmrds across an extensive geographic region ( C h e s a p e a k e Bay) in an effbrt to scale up f r o m previous small-scale mechanistic studies to a target ecosystem with m o r e variable e n v i r o n m e n - tal conditions. We hypothesized that the distribu- tion a n d a b u n d a n c e o f
Phragmites
would be linked to a n t h r o p o g e n i c l a n d use t h r o u g h pathways at b o t h local (e.g., physical d i s t u r b a n c e a n d surtace water r u n o f f ti-om local l a n d use will cause eleva- tions in available N a n d decreases in salinity at the u p l a n d - w e t l a n d b o r d e r ) a n d watershed scales (e.g., smdhce water r u n o f f a n d p o i n t source discharges D o m watershed-wide l a n d uses including agriculture mad d e v e l o p m e n t will increase available N at. the seaward-wetland b o r d e r ) . To test this h}~othesis, we relatedPh.ragm.i~es
distribution a n d a b u n d a n c e d a t a collected f l o m 90 tidal wetlands s p a n n i n g over 950 km of C h e s a p e a k e Bay to digital land cover d a t a s u m m a r i z e d at b o t h local a n d watershed scales. We also e x p l o r e d the potential linkage b e t w e e n land use a n d increased N availability at the watershed scale usingPhragmites
l e a f tissue N f l o m subestuary s h o r e l i n e s as an i n d i c a t o r o f e n r i c h m e n t . We p r e d i c t e d that increasing a m o u n t s o f urban-sub- u r b a n d e v e l o p m e n t a n d agricultural lands adjacent to study wetlands or within their p r o x i m a l water- sheds would b e related to an increasing f r e q u e n c y o f o c c u r r e n c e a n d a b u n d a n c e ofPhrag, mites
a n d increasing foliar N concentrations.M e t h o d s
,STUDY AREA AND SAMPLING
This study was c o n d u c t e d in 30 estuarine tributar- ies, or subestuaries, o f C h e s a p e a k e Bay (Fig. 1).
Subestuaries were selected f r o m ml initial p o p u l a - tion o f 60 b a s e d o n several criteria. Each selected subestuary h a d a distinct e m b a y m e n t that was well s e p a r a t e d f r o m the m a i n s t e m or m a j o r tributaries (e.g., P o t o m a c River) of C h e s a p e a k e Bay in the states o f Maryland a n d Virginia. Selected subestu- m'ies also s p a n n e d a north-to-south salinity g r a d i e n t r a n g i n g f r o m oligohaline to u p p e r m e s o h a l i n e . N o n t i d a l or tidal f r e s h w a t e r systems were n o t i n c l u d e d in this study-. Subestuaries were c h o s e n so that l a n d use or l a n d cover a m o n g watersheds s p a n n e d a gradient s p r e d o m i n a t e I y • to highly agricultmal or u r b a n - s u b u r b a n d e v e l o p e d ( h e r e ~ t e r , developed). T h e most h e a v i y d e v e l o p e d watersheds were primarily located in the u p p e r a n d middle bay o n the western shore in the Baltimore- Washington D.C. m e t r o p o l i t a n area a n d in a local- ized area of the lower bay in the Nortblk, Virginia, m e t r o p o l i t a n area, a l t h o u g h localized s h o r e l i n e d e v e l o p m e n t was widely distributed axa,ong m o s t
Urbanization Thresholds and Phragmitos 471
/ - \ _ Wate, .ed
.. ---~ scale
\
I "~-~--~-~
"\.. /
"..\ . . . .<.
L_
Sub
Local
Study scale
w e t l a nd
U-~.,7 ~- ~ .~ ~o,~
Fig. 1. The laxger map is of Chesapeake Bay, USA, illustrating the locations of the 30 subestuaxies and their watersheds. Points correspond to the locations of the 90 study wedands. The Watershed scale illustration is of inverse distance weighted {IDW) developed land in relation to the subesmaxy {blue)-tldal wetland {magenta) border within a contributing watershed. Weights assigned to each developed land pixel axe represented by a gradient fi-om Mack {neax) to light gray {fax) and correspond to their inverse distance {m) to the subestoaxy-tidal wetland border. White corresponds to nondeveloped land covers or developed land that is sufficiently fax that it receives minimal weight. In this example, unweighted % developed land in the watershed was 66.5%, whereas the distance-weighted % developed land was 46.8%. The Local scale illustration is of IDW developed land cover within 500 m of an individual tidal wetland. In this example, developed land occupied 34% of the 30-m buffer adjacent to the wetland border, whereas 74% of the axea within 600 m of the wetland, weighted by its distance to the border, was developed. This 9.3 ha wetland was heavily invaded by
Phragrnites
{abundance index - 7, the maximum value possiMe).s u b e s t u a r i e s . A g r i c u l t u r a l l a n d u s e was also w i d e l y d i s t r i b u t e d b u t m o s t i n t e n s i v e i n w a t e r s h e d s o n t h e e a s t e r n s h o r e o f t h e b a y i n M a r y l a n d . G i v e n t h e s e e x i s t i n g s p a t i a l p a t t e r n s o f l a n d u s e , we c h o s e s u b e s t u a r i e s t h a t r e s u l t e d i n t h e g r e a t e s t s p a t i a l d i s t r i b u t i o n o f t h e p r e d o m i n a n t l a n d u s e classes a c r o s s t h e r e g i o n t o m i n i m i z e c o n f o u n d i n g e f f e c t s o f s p a t i a l p h e n o m e n a u n r e l a t e d t o l a n d c o v e r ( K i n g e t al. 2 0 0 4 b , 2 0 0 5 b ) . G r e a t e r d e t a i l s a b o u t w a t e r - s h e d s a n d t h e i r s u b e s t u a r i e s c a n b e f o u n d i n K i n g e t al. ( 2 0 0 4 a ) , D e L u c a e t al. ( 2 0 0 4 ) , a n d K i n g e t al.
( 2 0 0 ~ a ) .
Phragmites
a b u n d a n c e was e s t i m a t e d at t h r e e b r a c k i s h t i d a l w e t l a n d s w i t h i n e a c h s u b e s t u a r y (Fig. l ) . T h r e e w e t l a n d s w e r e r a n d o m l y s e l e c t e d f r o m t h e l a r g e r p o p u l a t i o n o f sites, b u t s t r a t i f i e d b y w e t l a n d size. O n e w e t l a n d f r o m e a c h o f t h r e e size classes was s a m p l e d : < 2 ha, 2 - 7 h a , a n d > 7 ha.T h e s e n u m e r i c a l c r i t e r i a w e r e b a s e d o n t h e distri- b u t i o n o f size classes i n s u b e s t u a r i e s w i t h t h e f e w e s t n u m b e r o f t i d a l w e t l a n d s a n d e m p l o y e d as p a r t o f a l a r g e r s t u d y o f t h e s e s a m e w e t l a n d s o n m a r s h b i r d c o m m u n i t i e s o f C h e s a p e a k e Bay ( D e L u c a e t al.
2 0 0 4 ) , w h e r e w e t l a n d size was s h o w n t o a f f e c t b i r d a s s e m b l a g e s . T h e s e size c r i t e r i a e n s u r e d t h a t a w i d e r a n g e o f w e t l a n d s i z e s w e r e s a m p l e d e q u a l l y t h r o u g h o u t t h e s t u d y a r e a . A w i d e r a n g e o f w e t l a n d sizes a l s o a l l o w e d u s t o e v a l u a t e w h e t h e r size i n f l u e n c e d w e t l a n d s u s c e p t i b i l i t y t o
Phragmites
in-v a s i o n . G r e a t e r d e s c r i p t i o n o f w e t l a n d c h a r a c t e r i s - tics a n d site s e l e c t i o n c r i t e r i a a r e p r o v i d e d i n D e L u c a e t al. ( 2 0 0 4 ) .
W e e s t i m a t e d a b u n d a n c e o f
Phragrrdtes
at e a c h o f t h e 90 s t u d y w e t l a n d s u s i n g 1 0 0 - m d i a m e t e r s t u d y p l o t s . T h e n u m b e r o f p l o t s u s e d i n e a c h w e t l a n d v a r i e d b y w e t l a n d size: 1 ( < 2 h a ) , 2 ( 2 - 7 h a ) , a n d 3 ( > 7 h a ) . P l o t s w e r e l o c a t e d r a n d o m l y i n t h e i n t e r i o r o f t h e m a r s h w h e r e p l o t b o u n d a r i e s d i d n o t c r o s s t h e t e r r e s t r i a l o r o p e n w a t e r b o r d e r . W i t h i n e a c h p l o t , f o u r 1 2 - m d i a m e t e r v e g e t a t i o n s a m p l i n g c i r c l e s w e r e e s t a b l i s h e d , w i t h t h e f i r s t c i r c l e c e n - t e r e d i n t h e m i d d l e o f t h e p l o t , a n d t h e r e m a i n i n g t h r e e c i r c l e s l o c a t e d 3~ m f r o m t h e c e n t e r at 0 ~ 120 ~ , a n d 240 ~ , r e s p e c t i v e l y . E a c h v e g e t a t i o n c i r c l e was f u r t h e r d i v i d e d i n t o f o u r e q u a l q u a d r a n t s . W i t h i n e a c h q u a d r a n t ,Phragmites
a b u n d a n c e was472 R.S. King et al.
estimated using a m o d i f i e d B r a u n - g l a n q u e t cover scale (hereafter, a b u n d a n c e index) r a n g i n g t r o m 0 (absent) to 7 ( > 75% cover; Philips 1959; Leps a n d H a d i n c o v a 1992),
Phragmi~s
a b m a d a n c e index values were averaged a m o n g the four q~adrants o f e a c h circle, a n d t h e n averaged again a m o n g the f o u r circles, resulting in a m e a n a b u n d a n c e index value t b r the plot, I n t e r m e d i a t e a n d large wetlands h a d > 1 plot, so m e a n a b u n d a n c e index values a m o n g plots were averaged at those sites to yieId ma overallP]zrag,m.ites
a b u n d a n c e ~ndex value p e r wetland. We recognize flais s a m p l i n g m e t h o d may- have u n d e r e s t i m a t e d a b u n d a n c e osPhragmites
p a t c h e s a l o n g the u p l a n d or s e a w ~ ' d b o r d e r s . O u r goal was to characterize the e x t e n t of invasion o f the majority o f the m a r s h surface (not just the b o r d e r ) a n d relate this to digital [and cover data.
B e c a u s e o f t h e lag t i m e a m o n g d i s t u r b a n c e , establishment, a n d e x p a n s i o n of
Phragmites
(Ams- b e r r y et al. 2000; Philipp a n d Field 2005), s a m p l i n g the i n t e r i o r would yield a m o r e r e p r e s e n t a t i v e estimate of the e x t e n t o f im,asion ~ d its p o t e n - tial linkage to b o t h local ( u p l a n d wetland b o r d e r ) a n d watershed (seaward-wetland b o r d e r ) land use drivers.Because of the significant spatial e x t e n t o f the study area a n d additional goals o f the overall estuarine indicator project, n o t all wetlands c o u l d b e s a m p l e d d u r i n g t h e s a m e year. S a m p l i n g o c c u r r e d d u r i n g the s u m m e r s (Jnne-A.ugust) o f
2002
a n d 2003. N i n e t e e n subestuaries (57 wetlands) were s a m p l e d in2002,
whereas 21 subesmaries (63 wetlands) were s a m p l e d in 2003, T e n subestuaries were s a m p l e d in b o t h years for sm-tace water salinity, b u tPhragmites
distribution a n d a b u n d a n c e was n o t sufficiently variable at the scale o~ o~r wetland m e a s u r e m e n t s to warrant r e s a m p l i n g each wetland f o r i n t e r a n n u a l changes in abundmace ( D e L u c a p e r s o n a l observation). T h e intraann~al a n d inter- annual sequence in which snbestu~'ies were sam- p i e d was d e t e r m i n e d randomly.We e x p e c t e d that geographical differences in the d i s t r i b u t i o n o~
PlzrLz~mites
c o u l d be c a u s e d by salinity, contagious dispersal, or o t h e r ~actors un- related to land use, so s m l a c e water salinity a n d g e o g r a p h i c a l coordinates ( n o r t h i n g a n d casting, m) were m e a s u r e d a n d i n c l u d e d as predictors in data analyses (see n e x t section). G e o g r a p h i c a l location (Universal Trans-Mercator, Zone 18, in meters) o f e a c h wetland was r e c o r d e d using a global position- ing system. Surtace water salinity was s a m p l e d at two locations in the o p e n water at a d e p t h o f 10 cln i m m e d i a t e l y adjacent to the seaward b o r d e r o f each study wetland. I n t e r a n n u a l variation in salinity m a d e c o m p a r i s o n s diftlcult a n l o n g wetlands sam- p l e d in d i f f e r e n t years. We e x p e r i e n c e d very- different levels o f freshwater discharge into Chesa-p e a k e Bay d u r i n g
2002
(year 1) a n d 2003 (year 2).T h e year 2002 was a b n o r m a l l y dry-, particularly d u r i n g s u m m e r (ca. 95 cm a n n u a l rainfall, mostly in late winter a n d fall), whereas 2003 was the wettest year o n r e c o r d for the entire region (ca. 160 cm).
We standardized salinity- b e t w e e n years using mea- s u r e m e n t s f r o m 30 wetlands in which salinity was s a m p l e d in b o t h
2002
a n d 2003. T h e s e 30 wetlands s p a n n e d m o s t o f the bay's north-to-south salinity g r a d i e n t a n d iikely ca.t]t.ured c o a r s e spatial differ- ences in salinity between years. Salinity observations f r o m the 30 wetlands yielded file ~i~llowing re- gression equation: 2003 salinity (%0) - 2002 salinity (%0) • 0.582 - 0.37; r ~ - 0,75, p <- 0.0001. Using this equation, we estimated salinity at o t h e r wet- lands t o r the year n o t s a m p l e d (2002 or 2003) using the salinity- observed d u r i n g the actual vear o f sampling(2002
or 2003). Predicted salinity values were averaged with observed m e a s u r e m e n t s t b r all wetlands to yield a m e a n salinity- value for two consecutive s u m m e r s , standardizing salinity across all wetlands.A different sampling design was necessary- to estimate
Phragmites
tbliar N c o n c e n t r a t i o n s b e c a u s e m a n y study w e t l a n d s d i d n o t h o s tPhragmites.
A l t h o u g h previous studies have u s e d a b o v e g r o u n d N of a p h y t o m e t e r c o m m o n to all wetlands as an indicator of site-specific N availability (e.g.,
S partina Mter~iflora;
Bertness et al. 2002; Silliman a n d Bertness 2004), no such p l a n t was c o m m o n to all wetlands across the large spatial e x t e n t o f o u r study.It was n o t possible to standardize toliar N or standing stock a b o v e g r o u n d N at e a c h individual study wetland in the m a n n e r d o n e in these previous studies. Almost all o f the subestuaries (29 of 30) h a d stands os
Phragmites
at o t h e r locations along their shoreline (all os which in the general vicinity o f the study wetlands), so we c o m p a r e d m e a n ti~liar N levels a m o n g subestnaries as an indicator o f N availability a~ the watershed scale.We s a m p l e d
Pli.raA~m,ites
leaves f~om u p to 6 s h o r e l i n e s e g m e n t s in each os 29 subestuaries.S h o r e l i n e s e g m e n t s were 100 m in length a n d selected r a n d o m l y within each s u b e s m a r v using m e t h o d s described in King et al. (2005a). S a m p l e d plants were located o n the seaward b o r d e r o f the stand a n d were in contact with surface water o f the subestuary during high tide. We targeted seaward b o l d e r i n g plants r a t h e r t h a n interior or u p l a n d b o r d e r plants b e c a u s e seaward plants were m o r e likely to integrate watershed-scale n u t r i e n t availabil- ity ( a n d this c o m p o n e n t os t h e analysis was watershed-scale only). We collected o n e r e c e n d v e m e r g e d leaf located n e a r the tip o f each of 10 distinct shoots (i.e., 10 leaves p e r s a m p l e ) . In situations where
Phragmites
was n o t p r e s e n t b u t in the general vicinity- of the s e g m e n t , we s a m p l e dUrbanization Thresholds and Phragmites 473
these adjacent stands instead. Sampled leaves were placed o n ice until return to the laboratory.
Leaf composite samples were dried for 48 h at 60~C, g r o u n d , and analyzed for total carbon (C) a n d N using a Perkin-Elmer 2400 CHN Aa~alyzer.
Total N was almost periectly correlated to the C:N ratio (r e - 0.98), so we only u s e d total N (%) in t h r t h e r analyses. We averaged % total N a m o n g c o m p o s i t e sm~ples p e r subestuary to g e n e r a t e a spatially integrated estimate for each subestuary (Weis et al. 2003).
Foliar N samples were collected in 2002 a n d 2003 following the same sampling schedule described for
Phragmites
a b u n d a n c e . Nine subestu~'ies, rather than 10, were sampled in b o t h },ears because one o f the 10 subestuaries did not havePhragmites.
Because interannual variation in precipitation a n d r u n o f f may have i n f l u e n c e d N availability to
Phragmites,
we statistically analyzed foliar N data f r o m each year separately (see Statistical Analyses).We recognized that % N alone may n o t fully characterize N availability, as N e n r i c h m e n t can lead to b o t h increased foliar N a n d greater standing biomass without a difference in tissue c o n c e n u a t i o n (Windham aa'~d Meverson 2003). If an increase in folim N was f o u n d to be related to increasing percentages of p r o b a b l e N source areas (agricultur- al a n d developed lands), this would s u p p o r t the hypothesis that land use may be linked to greater N availability to
Phragmites.
We also r e c o g n i z e d that the presence o fPhragmites
in a particular location could, in itself, suggest local N e n r i c h m e n t if N is i n d e e d a tacilitator in its expansion, so sampling particular stands ofPhragmites
might be a biased indicator o f N availability. This bias should be equivalent for all stands a m o n g subestuaries, so a positive correlation s h o u l d again suggest an association between land use and N availability toPhragmites.
A l t h o u g h leaf tissue % N was n o t a direct m e a s u r e o f availability, we c o n t e n d its use as a relative indicator o f N availability was sufficiently- justified.GEOGRAPHIC ANALYSES
We used the A R C / I N F O 9.1 Geographic Infor- m a t i o n System (GIS; ESRI, Redlands, California) for geographic analyses. Watershed boundaries a r o u n d each subestuary were delineated m~aually from 1:24,000 digital elevation models expressed as a 30-m raster (USGS National Elevation Data Set;
www.usgs.gov). L a n d use data were extracted fi-om the National L a n d Cover Database (NLCD), a raster data set developed fi-om 30-m Landsat thematic- m a p p e r images taken d u r i n g 1992 (USEPA 2000).
We also c o n s i d e r e d two m o r e recent land use or land cover data sets derived f r o m satellite images t a k e n in 1999-2000: N L C D 2001 a n d RESAC
(2003); b o t h were experiencing problems along t h e estuarine-terrestrial b o r d e r a n d were b e i n g m o d i f i e d by their respective authors at the dine of analysis. T h e NLCD 1992 closely m a t c h e d o u r g r o u n d - t r u t h e d field observations o f land cover along the estuarine shoreline (i.e., developed pixels almost always c o r r e s p o n d e d to real d e v e l o p m e n t in the field), We expected a b u n d a n c e o f
Yhragmites
in 2 0 0 2 - 2 0 0 3 to be m o r e strongly c o r r e l a t e d to patterns o f land use several years earlier because o f t h e time lag t h a t w o u l d b e r e q u i r e d f o r establishment a n d expansion caused by land use (Amsberry et al. 2000; Philipp a n d Field 2005),We considered f o r e types o f land use or land cover as correlates of
5~/~ra.g~#tes:
developed land, which was the sum o f NLCD 1992 low intensity a n d high-intensity residential a n d c o m m e r c i a l classes;agricultural land, defined as the sum of pasture, hay-, cropland, a n d recreational grasses; cropland, which was e x a m i n e d separately- f r o m c o m b i n e d agricultural classes because it is the primary non- p o i n t source o f nitrate in the Chesapeake Bay watershed (jordan et at. 1997); a n d forest-wetland cover, which was the sum of all tbrest a n d wetland classes resoh,ed in the NLCD 1992 (i,e., all natural land cover classes c o m b i n e d ) .
We expected that land use close to shorelines may have a greater unit effect o n
Phrag, m#es
a b u n d a n c e a n d foliar N than land use l a t h e r away. To a c c o u n t f o r land use p r o x i m i t y , we first d e f i n e d a n d delineated shorelines in the NLCD 1992 as the b o r d e r between terrestrial land a n d h e r b a c e o u s tidal wetlands or the o p e n water estuary (Fig. 1). We subsequently summarized land use at both whole- watershed (contributing area to the subesmary) a n d local (area immediately a d j a c e n t to ~he study wetland) scales. At the watershed scale, we estimated p e r c e n t o f the area o c c u p i e d by certain land uses in three different ways: percentage o f the watershed area, excluding the subestuary a n d tidal wetlands;p e r c e n t a g e of a 30-m buffer along the shoreline (i.e., only land use immediately adjacent to the shoreline per results of Bertness et at. [2002] a n d Silliman a n d Bertness [2004]); a n d p e r c e n t a g e of the watershed area weighted by its inverse distance
(IDW) to the shoreline (Comeleo et al. 1996;
S o r a n n o et aL 1996; Fdng et aL 2004a, 2005b).
At the local scale, we c o n s t r m n e d our estimates of land use p e r c e n t cover to a m a x i m u m distance of 500 m o f the wetland b o r d e r (Fig. 1), This scale was c h o s e n to a v o i d spatial o v e r l a p in l a n d use percentages a m o n g some wetlands a n d because a previous study f o u n d that land use within 500 m o f these same wetlands was the strongest correlate of bird communities when c o m p a r e d to o t h e r scales (DeLuca et al, 2004). Study-wetland boundaries were delineated as contiguous e m e r g e n t wetland
4 7 4 R. S King et al.
pixels resolved in the NLCD 1992 fi)r each o f the 90 sites. We summarized local land use adjacent to the study wetlm~ds in each of three ways: p e r c e n t a g e c o n t r i b u t i o n o f a particular class to the total land area within 500 m o f file wettand border, percent- age o f land immediately adjacent to the wetland b o r d e r (one-pixel width or 30-in b u f f e r ) , a n d p e r c e n t a g e a r e a within 500 m o f the w e t l a n d border, weighted by its IDW to the border. We also estimated the area (ha) o f each study wetland, as delineated in the GIS, to e ~ l u a t e whether wetland size h a d all influence o n
Phra, gmites
invasion (see next section).We c o n s i d e r e d all of these land use metrics as predictors of
Phragmi~s
variables. T h e IDW metrics were calculated because they implicitly p r e s u m e that discharges from more distant land uses may- be a t t e n u a t e d by a variety o f processes along transport pathways Be%re r e a c h i n g the wetland or subestuary a n d should receive less emphasis, But should n o t be excluded altogether (as is d o n e in butter analysis).Fixed distance BuKers assume that all pixels within a certain distance of a %ature o f interest have ml equivalen~ effect on a response variable, which is also unlikely (King et al. 2005b). We predicted that the IDW m e n i c s would more e~tectively describe land cover ettects on
PMagmites
than simple whole watershed or fixed distance buffer percentages.To calculate IDW p e r c e n t land use, every pixel of each of the t b u r classes was assigned a distance (meters) to the shoreline using simple Euclidean distance. Pixels were weighted by the inverse o f their distance to the shoreline (1/distance) a n d s u m m e d for a distance-weighted pk<el c o u n t t o r the entire watershed. T h e process was repeated for all pixels in the watershed. T h e sum of distance-weighted, land- use pixels was divided by the sum o f distance- weighted, total land in the watershed to yield distance-weighted, percentage land use (e.g., IDW p e r c e n t a g e developed land is illustrated in Fig. 1).
STATISTICAL ANAL'~ SES
Preliminary examination of the
Phragmitz'~s
abun- dance and tbliar N data revealed relationships to predictor variables that were strongly nonlinear, heteroscedastic, and involved higher o r d e r interac- tions. We used a non.parm~etric alternative to multiple ,egression, regression tree (Ck&RT) analysis (Breiman et al. 1984; De'Ath a n d Fabricius 2000;U r b a n 2002), to predict
Phragm~ites
distribution, a b u n d a n c e , or toliar N according to land use, salinity-, wetland size, mid geographical c o o r d i n a t e variables.CART explains variatioa of a single response variable using o n e or m o r e predictor variables.
Response variables can either Be categorical (classi- fication tree) or numerical (regression tree). CART
works by recursively partitioning data into two mutually exchlsive groups by selecting a predictor variable that Best explains variation in the response variable (i.e., greatest reduction in deviance). T h e process is r e p e a t e d until the tree can no longer be grown Based o n a set o f stopping rules a n d cross validation o f the model. T h e m e t h o d works partic- ularly well when critical levels o f a predictor result in a nonlinear, threshold c h a n g e in the m e a n a n d variance of the response. T h e objective of the m e t h o d is to explain as m u c h variation (r ~) in the response variable as possible while minimizing the size o f the tree. This is analogous to incorporating explanatory- variables into multiple regression.
We built a CART m o d e l to explain variance in the
Phragmites
a b u n d a n c e index using the 90 study wetlands as observations. Separate tree models were also developed for thePfiragmites
tbliar N data t~om2002
(18 snbesmaries) and 9003 (20 subestuaries).Models were regression trees because the response variables were continuous. We required that termi- nal nodes (leaves) and 5 r a n c h e s have no • than 5 and 10 observations, respectively.
We cross validated each tree model to d e t e r m i n e the most appropriate size o f the tree (i.e., n u m b e r o f explanato,y variables included in the model).
Cross validation was c o n d u c t e d by r a n d o m l y parti- tioning the data (wetlands stratified by subestuary to a c c o u n t for spatial nestedness in a b u n d a n c e analy- sis, following De'Ath a n d Fabricius [2000]) into 10 groups of equal or similar size a n d creating a cross validation regression tree with only nine o f the 10 groups. This cross validation tree was subsequently u s e d to predict response variable data t}om each of the stations r e m a i n i n g in the tenth group. T h e process was r e p e a t e d 10 times so that each o f the 10 groups of sites was used as the cross validation g r o u p once. We retained predictors that resulted in an average overall model r 2 s~4thin o n e standard e r r o r (SE) of the m i n i m u m model r ~ a m o n g all possible trees (1 SE rule; Breiman et al. 1984;
De'Ath a n d Fabricius 2000).
We also recognized that although C&RT will choose the single best predictor at any given level of a tree, other predictors may expIain simiiar amounts o~ variance in the response. This was a p a r t i c u l a r c o n c e r n in o u r use o f m u l t i p l e measures o f land use as predictors, which were necessarily correlated (King et al. 2005b). For every split in the tree, we also r e p o r t e d up to the top three alternative p r e d i c t o r variables that were d e e m e d statistically significant (%2 1 ds C&RT analyses were c o n d u c t e d using the RPART library (Therneau a n d Atkinson u n p u b l i s h e d data) in S-P/us 6.1 (Insightful Corp., Seattle, Washington).
We t u r t h e r validated the predictors that explained t h e most variance in
Ph~agmites
regression treeUrbanization Thresholds and Phragmitss 4 7 5
Phragmites
A b u n d a n c e I n d e x M o d e l r 2 = 7 3 . 4 %
M e a n = 0.7 S D -- 1.2 n = 7 0
- < 1 5 . 3 % .~ p a r t i a l r 2 = 4 8 . 4 % ~. > 1 5 , 3 %
. . . O . . . . ~5~00 . . . .
7 o ( .... o
g s
_ o
/
8 5 o o
" p -< 0.0001 o o o
<~3 o o o
b
O O 0I 9 O
0 10 20 30 40 50 60 70 80
% IDW D e v e l o p e d L a n d (Local) p a r t i a l R = 12.5%
-<34.4% 4 ~, > 3 4 . 4 %
O
'1-0=O
0 8 ~ - ,~+.
06~
o 0.4 ~_ g
" 8
0 2 g .~- (l,0 9(1
M e a n = 4.3 S D = 2.3 n = 2 0 p a r t i a l r ~ = 12.5%
-<4108993 .~ ~- > 4 1 0 8 9 9 3
7 9
x
_~6 o
~s
] 4 3 -< 0.0001
.13
~ 2
b o
o o o
o o
o
o o o
o o o o
o n O G O ~ o
~ o
- 1 . 0 0 3 0 . 8 ~- E
o . 6 # - D.4 ~_
nO
- 0 . 2 g
. <
0.0
__s61
# 5 i
m I
5 4
I
. . . .oo~
. . . O O
O ~ O O
O
p = 0 . 0 2 8 9 o
o o
_i
- 1 . 0 0
" ::1
-O.8N
<
9 f o
0.6 ~"
22r
04 e.
"0
0 , 2 g
. ,~-
- 0 0
25 35 45 55 65 75 4100000 4200000 4300000
% I D W Forested L a n d ( W a t e r s h e d ) V 415ooo0 4250000 4350000
M e a n = 2.4 M e a n = 0.3 M e a n = 1.5 Northing (m) M e a n = 5.2
S D = 1 . 7 S D = 0 , 5 S D = 1 . 5 S D = 1 , 8
n = 14 n = 5 6 n = 5 n = 15
Fig. 2. Resuks fl~om regression tree a n d c h a n g e p o i n t analysis o f Phragraite~ a b u n d a n c e (n = 90 wetlauds). Variance in Ph~r~grnites
a b u n d a n c e was best explained by [DW % developed land adjacent to the wetland bot'dec, IDW % furest-wetlaad (for-we0 cover in the watershed, a n d geographical location a l o n g fl~e norfl>to-south axis o f C h e s a p e a k e Bay (notching, m). Scatterplots illustrate the response of
Ph~regmites a b u n d a n c e to e a c h predictor in the tree. T h e ~olid vet'tical line w h h i n eacta plot identis ~he value o f e a c h predic~er that best explained variance in Ps T h r e s h o l d values of predictors are shown to the left a n d right of each split in fl~e tree. Variance explained (r e) is shown above each split. Means, smndar'd deviations (SD), and tmrnber of wetlands are r e p e r t e d for each spli[~ X,~r~thin each scatterplot, dotted lines r e p r e s e n t the cumulaffve pr'obabilit T (y-axis, right side) of a threshold respcmse in Ptna~,~iees a b c m d a n c e with increasing vaIues of each predictor, as estimated u s i n g c h a n g e p o i n t anahsis, p xvtlues c o r r e s p o n d to the likeiihood that t h e r e was n o c h a n g e p o i n t in the data 9
models using n o n p a r a m e t r i c c h a n g e p o i n t analysis (IgJng and Richardson 2003; Qian et al. 2003).
C h a n g e p o i n t analysis estimates the numerical value o f predictor, x, resulting in a threshold in the response variable, y. Because c h a u g e p o i n t analysis uses the same test statistic as tree regression to identis nonlinear breaks or splits in a response variable (deviance reduction), the two m e t h o d s are similar. Unlike tree regression, the c h a n g e p o i n t m e t h o d employs a bootstrapping (resampling) tech- nique to estimate a percentile confidence interval a r o u n d the observed threshold. We c o n d u c t e d c h a n g e p o i n t analysis for each variable included in the fiaal tree models. We overlaid the cumulative
distribution o f the percentile confidence limits o n each predictor iD the tree models as a measure of the cumulative probability o f a threshold (DeLnca et al.
2004; King et al. 20051); Fig. 2), We also estimated the probability that the observed variance explained by the c h a n g e p o i n t was n o t different from zero (de- viance reduction = 0), a test that further validated the inclusion o f variables in tree models.
R e s u l L s
aPHR:&G'u ABUNDANCE
PMagmites was present in 52 os the 90 study wetlands (57.8%), \.~-u'iance in Phragmites a b u n d a n c e
476 R . S . King et al.
a m o n g wetlands was best explained by local-scale development, iorest-wetland cover near the subes- m a r y shoreline, a n d geographical location along the north-to-south axis of Chesapeake Bay (regression tree m o d e l r e - 73.4%; Fig. 2). T h e most influential factor in the regression tree was local IDW % developed land (partial r e - 48.4%; Fig. 2; also see Fig. 1). W h e r e local IDW developed land was >
15.3%, Phragudtes
a b u n d a n c e increased dramatically a n d was relatively high at all b u t a iew wetlands. T h e range of d e v e l o p e d land within the percentile c o n f i d e n c e limits was small, with the lower 5% limit c o r r e s p o n d i n g to 10% developed land, with 95%certainty that the threshold was no higher than 20%
(Fig. 2). Percent developed land within 500 m o f the wetland b o r d e r was the best alternative pre- dictor to local IDW % developed land (r 2 - 46.4%),
followed by % IDW s cover at the
watershed scale (r e - 36.4%), a n d % IDW de- veloped land at the watershed scale (r ~ - 35.7%).
Where local d e v e l o p m e n t was -< 15.3%, wetlands located in subestuaries with < 34.4% IDW ibrest- wetland cover at the watershed scale also t e n d e d to have a h i g h e r a b u n d a n c e o f
PMagmites
(partial r 2 = 12.5%; Fig. 2, les branch). This threshold was most p r o b a b l e b e t w e e n 34% a n d 40% I D W s wetland. Wetlands with relatively low local develop- m e n t a n d m o d e r a t e a m o u n t s of natural land cover near the shoreline in their proximal watershed also h a d the lowest cover osPMagmites.
T h e only significant alternative p r e d i c t o r was % iorest-wet- land in the watershed (r e - 10.4%).Wetlands with relatively high adjacent develop- m e n t a n d located in the middle a n d n o r t h e r n regions of Chesapeake Bay- h a d m o r e
PMagmites
than similar wetlands in the s o u t h e r n part o f the bay (northing, partial r 2 - 12.5%; Fig. 2, right b r a n c h ) . Wetlands in the s o u t h e r n bay with locally high d e v e l o p m e n t h a d low-to-moderate a b u n d a n c e o f
Phragmites,
whereas middle a n d u p p e r bay- sites subject to the same levels o f local d e v e l o p m e n t h a d more. Salinity (r 2 - 7.7%) was the best alternative p r e d i c t o r to n o r t h i n g , a l t h o u g h n o n e o f the alternate predictors at this n o d e o f the tree were significant.PHRAGMITES FOLIAR NITROGEN
N concentrations in
Pkragudtes
leaves were related to watershed land use d u r i n g b o t h years (Fig. 3). In2002,
N increased markedly- i}om an average o f 2.3% to 2.9% o n c e IDW developed land e x c e e d e d 14.3% (r 2 - 65.3%). The 95% c o n f i d e n c e limits for t h e t h r e s h o l d r a n g e d f i o m 7% to 30% I D W developed land, with high certainty- ( > 80%) that the threshold was at or below the observed re- g r e s s i o n tree split o f 14.3%. P e r c e n t N was consistently elevated a m o n g subestuaries with rela--<14,3%
= 65.3%
7
, "' Elizabeth
'10=0"0178 e (Hioghe;~iaryt)y"
3.50 1 3,25 1 3001 2.751 2.5o 1 2,251 2.001
M e a n - 2.3 SD - 0.2 r ~ - 1 0
<22A%
: o /
t
O / o
I
o !O
i
% ~ : . , 2 0 0 2 { D r y Y e a r )
0 15 30 45 60 75 90
% I DW Developed Land (Walershed}
72 = 51 5%
>22.4%
1 0 o
c :=]
o6=-#
04 ~ 02 ~ S
O0
Meat] = 2 9 SD = 0 . 2 ,r1 = ;5
3.50 3.25
/
3,00
U_
.a 2.50 2 25 2.00 Mean = 2.6
S D = 0 3 n ' = 1 3
,.,.j'-'"e ... 1,0 @
r ~' 3
0 . 8 ~"
0 ',
{.> 1
,' ~ 0.6
O 0~
8 ~ / " #=0,0237 0.4 ~
,/" 0 2 ~
,," 2003 (Wet Year) ,~
Y " 0 0
0 15 30 45 60 75 90
% IDW Developed Land (Watershed) M e a n = 3,1 SD = 0.2 1 1 = 7
Fig. 3. Results from regression tree and changepoint analysis o f Phragraites foliar nitrogen (N, %) composite samples collected from 29 subestuaries during 2002 (n 18) and 2003 (n 20).
Bold symbols correspond to subestnaries sampled in b o t h years.
Symbol size is proportional to IDW % agriculture in the watershed. See Fig. 2 for other details.
tively high levels o f d e v e l o p m e n t near the shoreline, including the Elizabeth River, a southern-bay sub- estuary with tile highest average salinity. During this dry year, % N in subestuaries with p r e d o m i n a n t l y agricultural watersheds (Fig. 3; IDW % agriculture p r o p o r t i o n a l to increasing size of symbols) were n o t consistently elevated over torested systems, a n d were always m u c h lower than developed watersheds. T h e only alternative p r e d i c t o r in the regression tree that was significant was % developed land (unweighted by distance) in the watershed @2 _ 41.2%).
In 2003, a wet year, IDW % developed land in the watershed was again the best p r e d i c t o r of ibliar N (r ~ - 51.5%; Fig. 3). Above
22.4%
IDW developed land, foliar N averaged 3.1% c o m p a r e d to 2.6% in less developed watersheds. Subestuaries sampled in b o t h years (bold symbols in Fig. 4) showed similar patterns to o t h e r subestuaries sampled in the same year, d e m o n s t r a t i n g these trends were n o t limited just to those nine systems. Unlike2002,
several3.5Q
3.25
>-
r eq 2.75 7
2.50
2.25 J
2.00
Ag ricultu ral watersheds ." 0 " %
1
,' / / / / b ' . O O,,' /
l/ O / z//z
e
z / /
~ / ' 1 : 1 l i n e
/ z ,i
2.00 2.25 2.50 2 7 5 3.00 3.25 3,50
Leaf-tissue % N, 2002 {Dry Year) Fig. 4. Relationship between Ph~agmi~e's t'ot[ac t][trogen (N, %) c o m p o s i t e samples collected f r o m t h e s a m e tocations in 9 s u h e s m a r i e s in 9002 (a dry year) a n d 2003 (a record wet year).
Symbol size c o r r e s p o n d s to I D W % agriculmce in the watershed.
watersheds below the d e v e l o p e d land threshold b u t with relatively high I D W 2 agficulture t r e n d e d u p w a r d in 2 N in c o m p a r i s o n to watersheds with the m o s t r e m a i n i n g tbrest-wetland cover n e a r the shoreline (Fig. 4; increasing size o f s?mbols pro- p o r t i o n a l to IDW 2 agriculture) a n d w h e n com- p a r e d to the previous, m u c h dryer year. Despite this a p p a r e n t i n t e r a n n u a l effect o f agriculture o n 2 N, cross validation (1 SE rule) excluded % agriculture fl-om the tree m o d e l as a second predictor, likely d u e to a small s a m p l e size below the d e v e l o p e d - l a n d threshold. T h e best alternative p r e d i c t o r to IDW % d e v e l o p e d l a n d as the p r i m a r y p r e d i c t o r was IDW };:b forest-wetland in the watershed (r ~ = 50.5%). It was a strong alternative variable because o~ the t r e n d o f increasing % N in b o t h d e v e l o p e d a n d agricultural systems in 2003, b o t h of which were negativeIy correlated with % fbrest-wetland.
D~smtssion
P]~mitgs
a b u n d a n c e exhibited a m~nlinear re- sponse to d e v e l o p m e n t that was consistent with an ecological t h r e s h o l d (Muradian 2001; King a n d Richardson 2003; H u g g e t t 9005), b e y o n d which m o s t wetlands were predictably invaded byPhrag- mites.
C h a n g e p o i n t analysis revealed thatPhra~dtes
m a y proliferate in wetlands with as little as 10 % local d e v e l o p e d land, whereas t h e r e was a high probabil- ity- o f a n o n l i n e a r c h a n g e in
Phra~mztes
a b u n d a n c e b e y o n d 2 0 2 . This finding is consistent with results r e p o r t e d by Bertness et el. (2002) a n d Silliman a n d Bertness (2004), who c o n c l u d e d that h u m a n de- v e l o p m e n t o f adjacent u p l a n d s was the criticalUrbanization Thresholds and Phragmites 477
factor driving the successiul e x p a n s i o n of
Phragmites
in N a r r a g a n s e t t Bay- salt marshes. T h e i r results also suggested a similar t h r e s h o l d level of local de- v e l o p m e n t : salt m a r s h e s with < 1 5 - 2 5 2 o f the b o r d e r d e v e l o p e d consistently h a d low p e r c e n t a g e s o f their b o r d e r s d o m i n a t e d by-
Phrag'mites,
whereas wetlands above this t h r e s h o l d were heavily invaded.Many o f the m e c h a n i s m s driving
Phrag?nites
invasion at a m u c h smaller spatiaI e x t e n t in Narragansett Bay may also be o p e r a t i n g at a. m u c h larger scale across C h e s a p e a k e Bay. It is intuitive thatP]zraA~,fftes
is likely to c o m p l e t e l y invade wetlands o n c e they' have b e e n disturbed a n d sufficiently e n r i c h e d by b o r d e >ing d e v e l o p m e n t , r a t h e r t h a n
only
invading wet- lands in p r o p o r t i o n to b o r d e r i n g d e v e l o p m e n t (e.g., 50% b o r d e r i n g d e v e l o p m e n t resulting in 5 0 2Phragmites
cover). O u r findings suggest that relative- ly low cover o f adjacent d e v e l o p m e n t m a y be all that is r e q u i r e d to create an ideal e n v i r o n m e n t for e s t a b l i s h m e n t a n d d o m i n a t i o n of C h e s a p e a k e Bay tidal wetlands byPhragmites.
O n e distinction b e t w e e n o u r results a n d those of Bertness et al. (2002) a n d Silliman a n d Bertness (2004) was that we c o n s i d e r e d agricultural land uses separately f r o m u r b a n - s u b u r b a n d e v e l o p m e n t in o u r analysis. We e x p e c t e d b o t h d e v e l o p e d a n d agricultural lands would b e related to
Phragmites
a b u n d a n c e , b u t s e p a r a t i n g these two l a n d use classes w o u l d allow us to distinguish w h e t h e r
Phragmites
r e s p o n d e d difterently to e a c h land use.Agricultural cover, b o t h in terms of c r o p l a n d or all agricultural classes c o m b i n e d , was n o t as clearly linked to
Phragmites
a b u n d a n c e as d e v e l o p e d land, at least at a local scale. Several wetlands with relatively high cover o f adjacent agriculture h a d littlePhragmites,
whereas virtually all wetlands with similar c o v e r o f a d j a c e n t d e v e l o p m e n t h o s t e d a b u n d a n t standsof Phragmites.
This may have reflected wariability in the status or type of local- scale agriculture a m o n g wetlands (active row c r o p versus • tields, which would not be distin- guished accurately in the NLCD data set), or that u r b a n land use is simply a m o r e p r e d i c t a b l e stressor.Broader-scale a g r i c u l t m e may still have b e e n an h n p o r t a n t driver o f
Phrag~,dt~;s
invasion ill several of the wetlands we sampled. W h e r e local-scale de- v e l o p m e n t was low ( < 1 5 2 ) b u t watershed-scale forest-wetland cover n e a r shorelines was also rela- tively low ( <352,
agricultural a n d d e v e l o p e d land cover were high), wetlands t e n d e d to have high a b u n d a n c e o fPhragmites.
Several wetlands located in highly agricultural watersheds on the eastern shore o f Maryland were heavily- invaded byPhrag~ites
( a l t h o u g h these wetlands d i d n o t consistently have cultivated wetland borders). This result implies that even in wetlands where local-scale d e v e l o p m e n t is relatively low,
Phragmites
invasion is still p r o b a b l e if478 R.S. King e l al.
subesmaries are located in highly agricultural or developed watersheds, particularly if these land uses occur near the subestuary shoreline. T h e watershed- scale resuhs linking increasing urbanization near shorelines to elevated fi)liar N in
Phragmites
provides tilrther s u p p o r t that watershed-scale n u t r i e n t pollu- tion may also facilitate invasion b e y o n d the effects o f local d e v e l o p m e n t alone. T h e distribution o f b o t h local a n d watershed land use is likely an i m p o r t a n t consideration in predicting areas suscep- tible to the invasion o fPhragmites.
Latitude or o t h e r geographical covariates also may have played a secondary- or tertiary- role in regulating the d e g r e e o f invasion by
Phragmites
in these tidal wetlands. We observed a relationship between latitude (northing) andPhragmites
abun- dance in wetlands with > 15% IDW local de- velopment. Wetlands in the s o u t h e r n third o f the bay ~ t h relatively high amounts of d e v e l o p m e n t n e a r their borders t e n d e d to have lower a m o u n t s o fPhragmite~s
cover than their n o r t h e r n counterpa,ts, although all three o f the wetlands sampled in the highly u r b a n i z e d Elizabeth River were moderately- invaded byPhn~gmztes,
despite b e i n g the farthest south a n d having the highest salinities o f the study wetlands. Most o f t h e r e m a i n i n g wetlands we sampled in the s o u t h e r n bay were not yet invaded byPhragmites.
This may reflect several factors, some o f which have b e e n previously p r o p o s e d (Saltonstall 2002; Silliman and Bertness 2004). At a regional scale, the lower ba) is less developed than the middle and u p p e r bay and hosts a large p r o p o r t i o n o f the remaining {0rest cover in the region. This spatial pattern ihrdaer supports a linkage between d e v e l o p m e n t a n dPhragmi~s
because low regional d e v e l o p m e n t may be limitingPhragmites
invasion by r e d u c i n g physical disturbance and n u t r i e n t inputs to wetlands (which also would r e d u c e the probabil- ity o f spreading propagules to n e i g h b o r i n g wetlands via contagious dispersal). Others have r e p o r t e d a t r e n d o f decreasing p r e v a l e n c e o fPhragmites
moving southward along the Atlantic coast. Salt- onstall (2002) suggested that the n o r t h e a s t e r n USA was the initial invasion p o i n t for the aggressive, nonnative genotype that is believed to be responsi- ble for most o f the invasions in Atlantic tidal wetlands, Because
Phragr, zites
e x p a n s i o n rate is r e p o r t e d to be about 1-2 m ~- ~ (Amsberry et al.2000; Philipp and FieId 20051, it would require u p to several decades fbr
P/,rag~ites
to completely invade a marsh o f m o d e r a t e size.PA'nzge~ait,'~s
may be spreading from north-to-south and simply has not had sufficient time to invade wetlands in the s o u t h e r n bay- as completely as it has in the north, H i g h e r salinities in the s o u t h e r n bay- may- also be limiting dispersal a n d germination o f propagules (Bart and H a r t m a n 9003), although this seems lesslikely given that others have d e m o n s t r a t e d the ability o f
P/~.rc~g~7,it,~s,
especially the nonnative geno- type M, to b e c o m e established and invade salt marshes e x p o s e d to full strengfll seawater (At-us- berry et al. 2000; M i n c h i n t o n a n d gertness 2003;Vasquez et al. 2005). We believe that both the first and second explanations are two of the most likely causes for the regional p a t t e r n observed in o u r study, a l t h o u g h these a n d o t h e r potential explana- tions deserve closer scrutiny.
N e n r i c h m e n t may- p r o v i d e o n e i m p o r t a n t , mechanistic explanation for the local a n d water- shed-scale linkages between
Phragmites
expansion m~d land use in the subestuaries we studied. N e n r i c h m e n t is an i m p o r t a n t driver o f invasion because it reduces b e l o w g r o u n d competition t o r nutrients and increases the importance o f above- g r o u n d c o m p e t i t i o n Ibr lighL a shff~ that dramati- c~ly ~:avors toweringPk"rag~,zites
shoots over m u c h s h o r t e r native species (Minchinton .and gertness 2003). Although we acknowledge the association in o u r study is correlative, the fact that N e n r i c h m e n t has b e e n shown to b e critically i m p o r t a n t in the expansion ofPhragmites
in n o r t h e a s t e r n USA salt marshes a n d elsewhere (e.g., Rickey at~d 'Anderson 200~i) based on mechanistic, e x p e r i m e n t a l research provides strong evidence that the positive correla- tions observed in this study were not spurious.Nitrogen concentrations in
Phragmites
leaf samples were h i g h e r in subestuaries with > 14-22% IDW d e v e l o p m e n t in their watersheds, approximately the same threshold level of local development that c o r r e s p o n d e d to high a b u n d a n c e ofPhrage~ites.
This result was consistent between two consecutive years o f smnpling despite wildly different interannual precipitation, freshwater a n d n u t r i e n t runoff; and salinities. This p a t t e r n also was consistent across the north-to-south salinity- gradient of the bay. T h e highest foliar N was d e t e c t e d in b o t h n o r t h e r n , low salinity watersheds in the Baltimor~Washington, D.C. m e t r o p o l i t a n areas as well as the Elizabeth River (Norfolk m e t r o p o l i t a n area), a highly u r b a n watershed in the s o u t h e r n bay with the highest salinity o f all watersheds studied, Given that the spatial e x t e n t o f this o b s e r v e d p h e n o m e n o n s p a n n e d the entire length of the bay, it is highly likely that elevated N inPhrag?nites
was related to developed land uses in these watersheds rather than some o t h e r spatial covariate (e.g., s~inity or N discharges f i o m the S u s q u e h a n n a River in the u p p e r bay).We hypothesized that land use would be linked to increased N availability- to
Phragmites,
but did not a.nticipate that foliar N would be m u c h m o r e clearly related to developed than agricultural land. Fresh- water streams in agricultural watersheds in the Coastal Plain of Chesapeake Bay- maitttain h i g h e rnitrate-N concentrations than those of developed watersheds ( J o r d a n et al. 1997; King et al. 2005b) a n d agricultural watersheds discharge m o r e N t}om sin-face-water r u n o f f into Chesapeake Bay subestu- aries than watersheds in o t h e r types o f land use ( J o r d a n et al. 2003; Weller et al. 2003). U r b a n areas contribute large quantities o f nutrients via point- source discharges from wastewater t r e a t m e n t plants, so m u c h of the n u t r i e n t pollution associated with u r b a n land avoids stremn networks altogether a n d is discharged directly into estuarine ecosystems (Cas- tro et al. 2003). Point source discharges represent a constant level o f input to systems m~d are relatively invariant to hydrologic ~lux ( a l t h o u g h can be diluted by high ~ieshwa~er flux into the estuary).
AgricuhuraI sources o f nutrients are chiet~y non- point, so loadings exhibit dynamic variability in c o n c e r t with watershed hydrological flux. In this study, we observed no a p p a r