OIKOS 72: 349-358.Copenhagen 1995

The effects of long-term manipulation of nutrient supply on competition between the testudinum and wrightii in Bay

JamesW. Fourqurean,George V. N. Powell,W. Judson Kenworthy and JosephC. Zieman

Fourqurean,J. W., Powell, G. V. N., Kenworthy, W. J. and Zieman, J. C. 1995. The effects of long-term manipulation of nutrient supply on competition between the seagrassesThalassia testudinum and in Florida Bay. -Oikos 72: 349-358.

Long tenn (8 yr) continuous fertilization (via application of bird feces) of established seagrassbeds in Florida Bay, FL, USA caused a change in the dominant . Before fertilization, the seagrassbeds were a monocul- lure; after 8 yr of fertilization the seagrassHalodule wrightii made up 97% of the aboveground biomass. Fertilization had a positive effect on the standing crop of T. testudinumfor the first two years of the experiment. The transition from T. testudinum- dominatedto H. wrightii-dominated was dependenton the timing of colonization of the sites by H. wrightii; the decreasein T. testudinum standing crop and density at the fertilized sites occurred only after the colonization of the sites by H. wrightii. There were no trends in the standing crop or density of T. testudinumat control sites,and none of the control sites were colonized by H. wrightii. The effects of fertilization on these seagrassbeds persisted at least 8 yr after the cessationof nutrient addition, suggesting that these systems retain and recycle acquired nutrients efficiently. Results of theseexperiments suggestthat Halodule wrightii, the nonnal early-succes- sional seagrassduring secondary successionin Caribbeanseagrass communities, has a higher nutrient demand than Thalassia testudinum,the nonnallate successionalspe- cies, and that the replacement of H. wrightii by T. testudinum during secondary successionis due to the ability of T. testudinumto draw nutrient availability below the requirements of H. wrightii.

J. W.Fourqurean and J. C.Zieman, DeptofEnvironmental Sciences,Univ. of Virginia, Charlottesville, VA 22903, USA (present address ofJWF: Dept ofBiological Sciences and SoutheastEnvironmental Research Program, Florida International Univ., Miami, FL 33199, USA). -G. v: N. Powell, RARE Center for Tropical Conservation, 1529 Walnut St., Philadelphia, PA 19102, USA. -W. J. Kenworthy, National Marine Fisheries Service, SoutheastFisheries Center, Beaufort, NC 28516-9722, USA.

Species composition of communities is dependent to reach equilibrium following a disturbance, resource- on a number of factors, including prevailing resource based models of community structure predict that the supply rates,the length of time since the last disturbance species composition of the community will be deter- (successionalstate), the historical make-up of the com- mined by the resource supply rates to the community munity, and the likelihood of colonization of the area by (Tilman 1982, 1985, 1988). A resource (sensu Tilman individual species (Harper 1977, Grime 1979, Tilman 1982)is any factor consumedby an organism that causes 1982, 1988). Given sufficient time for plant communities an increase in growth rate or survival. Using this defini-

Accepted 1 September1994 Copyright @ OIKOS 1995 ISSN 0030-1299 Printed in Denmark -all rights reserved

OIKOS 72:3 (1995) 349 Fig. 1. Site map of Florida Bay.

tion, light, nutrients and water are resources,but factors nutrient-limited; fertilization of the Thalassia testudi- suchas temperatureare not..This model explicitly consit:J- num-dominated seagrassmeadows in northeast Florida ers trade-offs in the ability of to compete for Bay increases biomass of T. testudinum (Powell et al. resourcesin two realms: abovegroundand belowground. 1989). The biomass of T. testudinumon a bay-wide scale Due to the limited energy available to a plant, allocation is positively correlated with the concentration of dis- to gathering belowground soil nutrients and water must solved inorganic phosphorus in the sediment porewater necessarilydecrease the relative allocation to gather abo- (Fourqurean et al. 1992a), and the observed increasing vegound resources (e.g. light; for review see Chapin trend in biomass from northeastto southwest in Florida 1980). Bay is a direct result of the increase in P availability The resource-ratio model predicts that availability of along the same transect(Fourqurean et al. 1992b). both soil nutrients and light exert a strong influence on Species composition of seagrassbeds of Florida Bay the species composition of plant communities, and ma- also is correlated with nutrient availability. Zonation of nipulations of the resource supply rates of either the species in relation to point sourcesof nutrients in other- above- or belowground resourcesmay lead to changesin wise oligotrophic northeast Florida Bay suggests that community composition. Addition of nutrients to a com- Halodule wrightii occurrenceis correlated with areas of munity composedof specieswith different aboveground! higher nutrient availability than Thalassia testudinum belowground resource acquisition strategiesshould favor (Powell et al. 1991), and after four years of continuous speciesthat have relatively less of their energy allocated fertilization of once-monospecific T. testudinumbeds, H. to nutrient acquisition (and consequentlymore of their wrightii colonizes the fertilized sites (Powell et al. 1991). energy allocated to light gathering). Nutrient addition Across the entire bay, areas th~t support H. wrightii have experimentsprovide a means for testing ecological theo- higher concentrations of dissolved inorganic P in the ries concerning the effects of nutrient availability on sedimentary porewater than areas that support only T. community composition and successionaldevelopment. testudinum(Fourqurean et al. 1992a). Seagrass meadows of the tropical western Atlantic Why Halodule wrightii should be restricted to areasof provide an accessible system in which to test the re- high nutrient availability is not well understood,but it has source-basedmodel of ecological succession.The small been suggestedthat H. wrightii has a higher demand for number of seagrassspecies, the simplified successional nutrients than Thalassia testudinum (Fourqurean et al. sequence of these seagrass communities, and the 1992a). In most places in Florida Bay, T. testudinum is strengths of environmental gradients in shallow marine the competitive dominant; Zieman et al. (1989) state that systems,all are desirable features for testing model pre- H. wrightii is the dominant seagrassonly in areas phys- dictions on the roles of resource supply rates in structur- ically unsuitable for T. testudinum.In the secondarysuc- ing plant communities. cessionalsequence in seagrassbeds of south Florida and Seagrass communities in Florida Bay (Fig. 1) are the Caribbean, H. wrightii is an early successionalspe- strongly affected by the availability of nutrients. Devel- cies and T. testudinumis a late successionalspecies (see opment of seagrassmeadows in northeastFlorida Bay is Zieman 1982 for review). The mechanismof the replace-

350 OIKOS 72:3 (1995) ment of H. wrightii by T. testudinumduring the succes- Data collection sional processhas been speculatedto be competition for light, with T. testudinum eventually overtopping H. The species composition, leaf biomass and short shoot wrightii due to its larger size. This proposed mechanism (SS) density of the seagrassbed around each control and is unsatisfactory in Florida Bay, since the T. testudinum fertilization treatment stake was measuredwhen the ex- canopy that replaces H. wrightii is often thin and sparse, perimentwas begun,and in late October or early Novem- with small effects on the amountof light that reachesthe ber for the next 8 yr. Short shoot density data were not sedimentsurface (pers. obs.). Further, H. wrightii is more collected in 1988. In 1987, 1989, 1990 and 1991, these shade-tolerantthan T. testudinum (Wiginton and McMil- same variables were measuredat the markers where fer- lan 1979, Iverson and Bittaker 1986), and H. wrightii tilization had been discontiI};uedin 1983. For each sam- exists in many locations as an understory in very dense pling, 4 quadrats (10 cm x 10 cm) were placed within 50 seagrassbeds (pers. obs.). cm of the marker stakes.The number of short shoots of This paperpresents the results of experimental manip- eachseagrass species were counted,and the seagrassleaf ulations of the nutrient supply to shallow water seagrass biomass within the quadratswas harvested.These leaves beds in Florida Bay. These experimentswere designedto were separatedby species,washed in 10% VN HCl to elucidate the mechanismsunderlying the apparent com- remove epiphytes, and dried to a constant weight. Leaf petition between Thalassia testudinum and Halodule biomass,commonly called standing crop in seagrassliter- wrightii. Two types of manipulations were performed: 1) ature, and short shoot (SS) density, were expressedon a continuous fertilization of previously undisturbed, T. tes- m-2 basis. tudinum-dominated seagrassbeds, and 2) cessation of fertilization of H. wrightii-dominated seagrassbeds pre- viously created by 28 months of continuous fertilization. Statisticalanalyses The averagevalue for the four quadrats from each stake were used in statisticalanalyses. Differences in the stand- Methods ing crop and short shoot density of seagrassesat contr@l (CONT) and fertilized (PERT) stakes as a function<>f Experimental design both treatment and time were assessedusing mixed- This study was conducted on Cross Bank, a shallow « 30 model univariate repeatedmeasures analysis of variance cm deep), narrow « 50 m wide) seagrass-coveredcar- (Winer 1971)with treatment (CONT vs PERT) and year bonate mud bank in east-centralFlorida Bay (Fig. 1). In as within-subjects factors. Each pair of CONT and PERT July 1981, location markers were placed at 100-m in- stakes constituted a subject in these analyses. Five of tervals along the center of Cross Bank as part of a sep- these ANOVAs were run: one each for differences in arate study on feeding behavior of wading birds (Powell short shoot densities of Thalassia testudinumand Halo- 1987). The markers were constructed of 1.5 m long, 1.2 dule wrightii, differences in standing crop of both spe- cm diam. PVC pipe with a 5 cm x 10 cm x 10 cm block cies, and differences in total seagrassstanding crop. of wood on top. Once in place, these markers were The discontinuation of fertilization treatment {DISC) heavily usedas roosts by royal terns (Sterna maxima)and was not included in the above analysessince there were double-crestedcormorants (Phalacrocorax auritus). By no measuresmade of the seagrassesaround these stakes November 1983,a patch of unusually denseseagrass was in the years 1983-1986 or 1988. Differences between evident around each location marker. At that time, we years in the standing crop and short shoot density at these began our experiment to test the hypothesis that deposi- stakes were assessedwith repeated measures ANOVA, tion of feces by the roosting seabirdswas responsiblefor with years as the within-subjects factor. the observed increase in seagrass density around the markers. Five of the markers,spaced at 6O0-mintervals, were pushed down into the sedimentso that birds could no longer roost on them, and a pair of new markers were Results placed 5 m on either side of the old marker. One of the new markers was identical to the old marker, while the Some results from the first four years of this experiment other new marker was cut to a point to preventbirds from have beenreported elsewhere (Powell et al. 1989, 1991). roosting. This design provided five sites, each with three Marker stakes in the PERT treatmentgroup hosted roost- treatments: 1) a control treatment, in which the pointed ing birds 84% of the time; 0.68 g of nitrogen and 0.13 g stake provided the sa;ffiehydrological effects as the bird of phosphoruswere deposited daily as bird feces at ea~h roost stake (CONT);"'2) a fertilization treatment, where of these roost markers. From 1983-1987, this nutrient birds could roost on the marker and defecate into the input causedincreases in the leaf biomassand a change in water (PERT); and 3) a treatment in which 28 months of the species composition of the seagrassmeadows in fertilization was discontinued where the markers had 5.4 m x 2.6 m elliptical patches surrounding the roost been pushed into the sediment (DISC). marker stakes,while no trends occurred at the control

OIKOS 72:3 (1995) the sites. Very small patches of Halodule wrightii grew elsewhere on Cross Bank, but none was within 10 m of our sites. In 1991, after 8 yr of fertilization, the seagrass +-'Q) beds at the FERT sites were dominated by H. wrightii, cn while T. testudinumremained the only seagrassspecies presentat the CaNT sites. In 1991,there were 860:t 125 (:i: 1 SE) SS m-2 of T. testudinum at the CaNT stakes, N compared with 40 :t 35 SS m-2 at the FERT sites. H. wrightii densitieswere 6260:t 319 SS m-2 at FERT sites, Q) -+-' and 90 :i: 80 at the controls. (/) Annual trajectories of the standing crop of Thalassia testudinumand Halodule wrightii were site-specific (Fig. 2). At all of the FERT sites, T. testudinumstanding crop f'") increased in the first year. At the sites where H. wrightii Q) becameestablished in 1984 (sites 1,3 and 4), T. testudi- +J num standing crop decreasedthereafter. At sites 2 and 5, "c/) T. testudinumstanding crop continued to increaseuntil H. wrightii became established (1985 and 1987, respec- ~ tively); after H. wrightii became established,T. testudi- num standing crop at these sites decreased.By 1989, H. Q) wrightii was the dominant seagrassat all of the FERT cn sites,while there was no H. wrightii at any of the CaNT sites. There was considerable yearly variation in the standing crop of T. testudinum at the CaNT sites. The total standing crop of seagrasswas affected by L{) fertilization (Fig. 3). When averaged through time, there +"'(l) was a significant difference in the total seagrassstanding (/) crop between CaNT and FERT treatments (ANOVA, Treatment main effect, Fl4 = 10.1, P = 0.03), but when averaged across treatme~ts, there were no significant differences in seagrassstanding crop as a function of time Fig. 2. Trajectories through time of the standing crop (in g(dry) (Time main effect, F8,32= 1.12,P = 0.38). However, there m-2) of Thalassia testudinum (open circles) and Halodule was a significant treatment by time interaction (F 8,32= wrightii (filled circles) at the paired control and fertilized sites. 9.8, P < 0.001), indicating that the pattern through time in the total seagrassstanding crop differed between CaNT and FERT treatments. For the fIrst three years of fertil- markers.The fertilization led to 20-fold increasesin pore- ization, FERT sites had 2 -3 fold higher seagrassstand- water dissolved phosphate concentration, and a 6-fold ing crops than controls, but from 1987 -1991, there was increase in porewaterannnonium. Water column nutrient no difference (Fig. 3). concentrations were not different between FERT and CONT sites, owing to the flow of water past the sites. After 8 yr of fertilization, the seagrassbeds surround- Tolar SeqgrO$s ing FERT stakesdiffered from CONT stakes.While there 200 N was no significant difference in the total standing crop of I E seagrassafter the 8 yr, the species composition and the """ >,1 50 relative contributions of seagrassspecies to the total bio- ... -c mass of the beds was very different. Analysis of the I"'"" r seagrass bed composition around the DISC stakes """,,01 Q.1 00 showed that the differences in the seagrassbeds caused 0 by fertilization persisted until the end of the experiment "-'U in 1991,even in the absenceof continued nutrient inputs. qI c 50 :0 ~ c Fertilized versus control treatments +'0 <1\ 0 The species composition of the seagrassbeds changed 19~3 1984 19~ 198619..7 19881989 1990,199 markedly at the FERT sites when compared to the con- Fig. 3. Total seagrass standing crop at the fertilized (filled trols. At the beginning of the experimentsin 1983,mono- circles) and control (open circles) sites. Values are means j: I specific Thalassia testudinummeadows surrounded all of SE.

352 OIKOS 72:3 (1995)

,+, Thalassia testudinum

1000 w ~ (/) N 800 '-"0'

Q) E QOO N y~ VJ ~ 400 ~ ] 200 W >.. "' i -'- +-' 0, I II 1,.,l.il::;:--=:=- I 0 "'~ ~ "",;> Q;..1 "'~ "'~ () \ .!:: U) ,~o ,~~ \~o \~~ ,~o '\~o\~o \~~ ,~~ VJ C Q) '0 Halodule wrightii Fig. 5. Size of short shoots (g of greenleaves per short shoot) of +-' 9000 Thalassia testudinum at the FERT sites. N.D. indicates no data 0 for 1988. Eachvalue is the mean of the 5 site means,+ I SE, n = ::.c() 7500 1\ 5.

(f) 6000 +-'~ 4500 0 -..C 3000 (f) 1500

o.~ E9 E9 I E9 E9-G I '"'~') ,",~b.'"'~':> '"'~~ ,",~1'"'~~ '"'~~~~C) ~~ \ \"" \"" \"" \"" \"" \"" \"" \ \

.Fertlized 0 Control

Fig. 4. Mean short shoot densities (:f: 1 SE) fQr Thalassia testudinum (top) and Halodule wrightii (bottom) at both control and fertilized sites.

The change from Thalassia testudinum-dorninatedsea- grassbeds to Halodule wrightii-dominated beds occurred gradually from 1983to 1991 (Fig. 4). There were signif- icant differences in the SS density of H. wrightii between CaNT and PERT treatments throughout the experiment Co. b Halodule wrightii (ANaVA, Treatmentmain effect, Fl.4 = 85.7, P = 0.001). 'L 150 Also, averaged across treatments, the density of H. (:) wrightii changed through time (Time main effect, F7,28= 125 23.8, P < 0.001). At PERT sites, H. wrightii density ~ C 100 increasedthrough time, while there was no change at the 0- controls (Treatmentby Time interaction, F7,28= 23.5,P < "~ 75 0.001). A more complicated patternwas evident in the SS ic: density of T. testudinumduring the experiments(Fig. 4). +'0 50 There was an overall effect of the treatment on T. testudi- '(f) num density when averaged through time (ANaVA, 25 Treatmentmain effect, F1,4= 21.0, P = 0.01). Also, there 0 .4 ffi ffi @ @ ffi ffi @ I were significant differences among years averagedacross treatments(Time main effect, F7,28= 2.4, P = 0.05). The \ ~~'J\~~\ ~~ry\~~Q\~~1\ ~~~\~~~\ ~~()\ ~~\ density of T. testudinumat PERT sites exhibited a differ- .Fertilized 0 Controf ent pattern through time than the CaNT sites (Treatment by Time interaction,F7,28= 11.4, P < 0.001). At the PERT Fig. 6. Mean standing crop (:I: I SE) for Thalassia testudinum sites, T. testudinum density remained constant from (top) and Halodule wrightii (bottom) at both control and fertil- 1983-1985, and then decreasedrapidly from 1985-1991. ized sites.

23 OIKOS 72:3 (1995) 353

.~1:1290 T. testudinum at the controls. Standing crop of H. wrightii increased H. wrightii through time when averaged across treatments (Titne 100 main effect, FS.32= 7.0, P < 0.001), and the trend in Q standing crop through time was different for the PERT 0 L 80 sites than the controls (Treatment by Time interaction, U FS.32= 6.8, P < 0.001). CJ' C As with the SS density data (Fig. 4), the pattern in the "0 60 c: standing crop of Thalassia testudinumthrough the exper- Q".. ~ iment was more complicated than for Halodule wrightii ~ 40 (Fig. 6). Averaged across all years of the experiment, S there was a significant difference in the standing crop of 0 I- 10 T. testudinumbetween PERT and CaNT sites (ANOVA, ~ Treatment main effect, FI.4 = 8.2, P = 0.05), and averag- ing both treatments,the change in standing crop of T. testudinumwas significant (Time main effect, FS.32= 2.5, P = 0.03). The patterns in standing crop through time Fig. 7. Average proportion (:I: 1 SE) of the seagrassesThalassia were very different between CaNT and PERT treat- testudinumand Halodule wrightii at the fertilized sites. ments, however (Treatment by Time interaction, FS.32= 21.9, P < 0.001). At PERT sites, standing crop of T. testudinum almost tripled between 1983 and 1984, ~d There was no apparentpattern in the density of T. testudi- then declined steadily between1984 and 1991. There was num at the CaNT sites. no such discernable pattern at the CaNT sites. Changes in SS density do not completely describe the The net result of fertilizing the Thalassia testudinum- change in the speciescomposition of these seagrassbeds dominated seagrass community normally prevalent on since short shoots of Thalassia testudinum are much Cross Bank was a change from a T. testudinum-dom- more massive than those of Halodule wrightii. Averaged inated meadow to one almost completely composed of over all of our measurements,T. testudinumshoots aver- Halodule wrightii (Fig. 7). In 1983, 100% of the total aged 0.133 :t 0.009 g of dry leaves per shoot, compared seagrassstanding crop at the PERT sites was T. testudi- to 0.013 :t 0.003 for H. wrightii. At the FERT sites, the num. After only one year of fertilization, H. wrightii size of T. testudinum shoots was strongly effected by began growing at these sites, and by 1986, the standing nutrient additions (Fig. 5). After one year of fertilization, crop of H. wrightii exceeded T. testudinum. In 1991, H. the averagemass of T. testudinumshoots doubled. As H. wrightii made up 97% of the total seagrassstanding crop. wrightii becameestablished at the FERT sites and began In contrast,H. wrightii nevercomprised more than 2% of replacing T. testudinum,the size of the remaining short the standing crop at the control sites,and in 1991,99% of shootsdecreased. the total seagrassstanding crop at the controls was T. The standing crop of Hqlodule wrightii was strongly testudinum. affected by the addition of nutrients in the form of bird feces to the FERT sites (Fi~. 6). Averaged across all the years of the experiment, there were significant differ- ences between FERT and CaNT treatments (ANaVA, Discontinuedfertilization Treatment main effect, Fl.4 = 28.6, P < 0.01): in all ANOVA indicated no statistically significant changes in instances except for the beginning of the experiment, H. seagrassstanding crop and short shoot density between wrightii standing crop was higher at the FERT sites than years 1987,1989, 1990and 1991 at the DISC sites (Table

Table 1. Characteristics of the seagrassbeds from sites at which fertilization was discontinued in 1983. Values are meanS:l: 1 $E. ANOVA results are for repeated measuresANOVA testing for differences between years.

Year Thalassia testudinum

1987 31.9:1:16.2 61.2:t14.7 93.2:t21.3 156:1:77 5287:1:1432 1989 73.6:1:16.1 47.8:t 5.4 121.4f:14.2 380:1: 99 3310:1: 377 1990 71.9:1:18.3 51.3:t18.3 123.2f: 8.1 440:1:128 4795:1:1483 1991 56.1:1:17.1 34.8:t16.4 9O.8f: 7.9 460:1:124 3115:1: 998 ANOVA Fz,3 1.6 1.3 16.0 0.5 4.0 p 0.4 0.5 0.06 0.7 0.2

354 OIKOS 72:3 (1')95) 1). The total seagrassstanding crop at these sites was mand for P than H. wrightii; this agrees with the esti- about equally divided among Thalassia testudinum(34% mates based on the specific growth rates of the species -58% of total) and Halodule wrightii (40% -66%). (Fourqurean et al. 1992a). Therefore, T. testudinum These proportions are similar to what was present at the should be able to out-competeH. wrightii for nutrients in PERT sites between 1985 and 1986 (Fig. 6). Assuming a nutrient-limited environment, which explains why T. that the 28 months of fertilization from July 1981 to testudinumdominates the nutrient-limited seagrassbeds November 1983 had the same effect on the surrounding of Florida Bay. Similarly, Williams (1987,1990) found seagrassbeds as the fertilization from November 1983 to that T. testudinum could out-compete Syringodium fil- October 1986,the seagrassbeds at the DISC sites did not iforme for nutrients in seagrassbeds in St Croix, U .S.V.I. change appreciably between 1983 and 1991. It is more difficult to explain the dominance of Halo- dule wrightii over Thalassia testudinum under nutrient- enriched conditions than to explain the dominance of T. testudinumover H. wrightii undernutrient-limited condi- Discussion tions. Once the nutrient limitation of H. wrightii was released at the fertilized sites, H. wrightii was able to The development of seagrassbeds in south Florida and colonize the fertilized areas.The decreasein T. testudi- the Caribbeanis often nutrient, and specifically phospho- num at the fertilized sites was directly related to the rus, limited (Short et at. 1985, 1990, Williams 1987, increase in H. wrightii (Fig. 2). There are at least three 1990, Powell et al. 1989, 1991, Fourqureanet at. 1992a, possible reasons for the decline of T. testudinum at the b). Not only is seagrassbiomass controlled by nutrient fertilized sites, including: 1) nutrient enrichment could availability, but there are documenteddifferences in the directly inhibit the growth of T. testudinumand not of H. nutrient availability in seagrassbeds dominated by Tha- wrightii; 2) H. wrightii could have an allelopathic effect lassia testudinum and Halodule wtightii in Florida Bay on T. testudinum,or 3) H. wrightii could, in the absence (Fourqurean et al. 1992a). H. wrightii beds have higher of nutrient limitation, out-compete T. testudinum for concentrationsof dissolved inorganic phosphorus in the some other resource (e.g. light). sediment porewater than T. testudinum beds. This ob- It does not seem likely that nutrient enrichment di- served correlation does not prove that H. wrightii beds rectly suppressedThalassia testudinum. In Florida Bay, are restricted to areas of high P concentration,since the the standing crop of T. testudinumis positively related to elevated P concentrations may be a consequenceof the P availability (Fourqureanet al. 1992a,b), and in the fIrst presence of H. wrightii. The experiments described two years of this experiment T. testudinum was sub- herein, however, demonstrate that H. wrightii colonizes stantially enhancedat the fertilized sites comparedto the T. testudinum-dominatedsea rassbeds and out-competes controls (Figs 4--6). T. testudinumdid not decreaseat any T. testudinum when nutrient availability is increased in of the fertilized sites until after Halodule wrightii became Florida Bay. well established (Fig. 2). Nutrient enrichment has been The dominance of Thalas ia testudinum in the nutri- cited as a factor in the decreasedstanding crop, produc- ent-limited seagrassbeds of orida Bay can be explained tivity and persistenceof some seagrassbeds, however. by the relative nutrient de ds of T. testudinum and Nutrient availability has been correlated with epiphyte Halodule wrightii. Using the elative growth rates and the loads on seagrassleaves (e.g. Sand-Jensen1977, Silber- nutrient contents of the two speciesto estimate relative stein et al. 1986,Tomasko and Lapointe 1991),and shad- nutrient demands,Fourqure et at. (1992a)have shown ing of seagrassesby epiphyteshas beenimplicated as one that H. wrightii has a four-£ d higher demand for phos- of the most deleterious effects of eutrophication of sea- phorus,the limiting nutrient, an T. testudinum.Another grasshabitats (e.g. Bulthuis and Woelkerling 1983, Cam- method for assessingthe r lative nutrient demands is bridge and McComb 1984). Since leaf turn-over is faster estimating the equilibrium r source requirement, or R*, for H. wrightii than T. testudinum(unpubl.) , it is conceiv- for each species (Tilman 198 ). R* can be approximated able that fouling of the longer-lived T. testudinumleaves by the concentration of a f source in the environment could causethe loss of T. testudinumfrom areasof nutri- when population sizes reach uilibrium. Assuming that ent enrichment, but this mechanism was probably not populations reach equilibriu with respect to nutrient operative in theseexperiments. While the epiphyte loads supply when additional nutri nt supply does not cause an of the seagrassesat fertilized and control sites were not increase in population size, * for sedimentnutrients in quantified, there were no visual differences in macro- Florida Bay seagrassbeds c n be estimated by the "sat- phytic or microscopic epiphyte loads at control or fertil- urating" porewater concen ions of nutrients for each ized sites. seagrassspecies. In the case f sedimentaryP supply, the The initial responseof T. testudinumto increasednutri- limiting nutrient for seagras growth in Florida Bay, R * ent supply was an increase in the leaf biomass per short for T. testudinumand H. wr ghtii are approximately 0.2 shoot (Fig. 5). Increasedleafiness is a well-known plant !J.Mand 0.9 !J.MP, respective y (Fourqureanet alI992a). responseto shading, and there is some evidence that T. Since R* for T. testudinumis substantially lower than R* testudinummay respondto decreasedlight availability by for H. wrightii, T. testudinu should have a lower de- increasing shoot size (Dawes and Tomasko 1988). In our

23* OIKOS 72:3 (1995j experiment, however, T. testudinum shoot size decreased fertilization. The loss of nutrients from the system over 8 after 1984, as light availability continued to decrease yr was not great enoughto draw the availability below the concomitantly with increasesin H. wrightii biomass.This threshold value for Halodule wrightii survival. Since sea- suggests that the initial increase in T. testudinum shoot grass beds do lose some nutrients due to diffusion from size was a responseto increasednutrient availability, not the sedimentsand leaf loss, it is inevitable that this will decreasedlight availability. eventually happen,however. Allelopathic effects have been documentedas impor- Successioncan be viewed as the change in species tant controls over interspecies interactions in some ter- composition through time caused by changing resource restrial environments (see Rice 1974). It is unlikely that availabilities due to the impacts of organisms on the Halodule wrightii has any allelopathic effect on Thalas- environment (Pickett 1976,Tilman 1987),and the ability sia testudinum,since there are many locations where H. of an organism to utilize resources may in large part wrightii exists as an understoryin predominantly T. testu- determine its role in succession.In the original formula- dinum seagrassbeds (pers. obs.). tion of the resource-ratio hypothesis, Tilman (1985, The most likely cause of the dominance of Halodule 1988) hypothesized that early successional species wrightii over Thalassia testudinum at the fertilized sites should have lower R* values for nutrients and be more is direct competition for light between the two species. competitive for nutrients than late successionalspecies. Under the fertilized treatment, H. wrightii developed a This observationwas directly at odds with many observa- long and dense canopy, with numerous "aerial runners", tions, however: one of the generalizedpatterns in changes or apical extending 20-30 cm into the water in plant physiological ecology during successionis a column. From distributional evidence(Phillips 1960,Wi- change from high to low resource demands (Bazzaz ginton and McMillan 1979, Iverson and Bittaker 1986)it 1979). In later work, Tilman and Wedin (1991a, b) re- is apparent that T. testudinumhas a higher light require- evaluated this facet of resource ratio theory and found ment than H. wrightii. The densecanopy of H. wrightii at that, contrary to earlier predictions, early successional the fertilized sites probably interrupted sufficient light so grasses have higher R * values than late successional that T. testudinum could no longer maintain a positive grasses.Tilman now considersearly successionalspecies carbon balance,leading to the extirpation of T. testudi- to have an edge over later successionalspecies due to num at those stations. their rapid colonizing rates (Gleeson and Tilman 1990, The resource-ratiotheory of plant community structure Tilman and Wedin 1991a),which agreeswith many other (Tilman 1982, 1988)predicts that experimentally manip- assessmentsof early successionaldynamics (e.g. Platt ulated plots that receive the same supply of limiting 1975, Connell and Slayter 1977, Bazzaz 1979). Early resourcesshould become similar in species compositon, succession in Florida Bay seagrassbeds also fits this or converge on similar communities, through time. The model. Halodule wrightii, the normal early successional convergence of plant communities receiving the same species, has a lower competitive ability for sediment resourcesupplies is dependenton the historical composi- nutrients than Thalassia testudinum, the later succes- tion of the communities, however (Inouye and Tilman sional species,but a potentially much faster colonization 1988). Similarly, the outcome of the manipulations rate than T. testudinum. H. wrightii has a much higher should depend on the availablity of propagules of species rate than T. testudinum (Tom- favored by the manipulation. In the fertilization experi- linson 1974, Fonseca et al. 1987), and sexual reproduc- ments presented here, Halodule wrightii was clearly fa- tion is potentially more important in the spread of H. vored by fertilization. Fertilized sites eventually con- wrightii than T. testudinum(Williams 1990). H. wrightii verged on similar, H. wrightii-dominated seagrassbeds, has a higher flowering rate than T. testudinum(Williams but the trajectories of the individual sites was dependent 1990). H. wrightii seedsform a seedbank in sedimentsof on the colonization of the sites by H. wrightii (Fig. 2). In seagrassbeds (McMillan 1981, 1983), and the seedling the absenceof H. wrightii, Thalassia testudinumbiomass successof T. testudinumis very low (Williams and Adey stayed elevated over control areas, but following the 1983). eventual colonization of the sites by H. wrightii, T. testu- The replacementof Thalassia testudinumby Halodule dinum declined. The stochastic event of H. wrightii colo- wrightii at the fertilized sites was in contrast to the nor- nization therefore controlled the responseof Florida Bay mal successionalsequence in seagrassbeds in southFlor- seagrassbeds to manipulations in resource supply rates. ida and the Caribbean.Under non-enrichedconditions, H. The changescaused by fertilization of seagrassbeds in wrightii is commonly the early successional,colonizing Florida Bay persisted for at least 8 yr after the fertil- species and T. testudinum is the late-successional,"cli- ization was discontinued (Table 1). The primary sourceof max" species (den Hartog 1971). Fertilization of other nutrients for seagrassgrowth in unfertilized seagrassbeds plant communities can also lead to the establishmentof is the remineralization of organic matter in the sediments communities resembling early successionalstates. For (Patriquin 1972, Capone and Taylor 1980, Short 1987). example, in an old-field fertilization experiment Carson The lack of change at the sites where fertilization was and Barrett (1988) found that nutrient enrichment of discontinued suggests that seagrassbeds are very effi- late-successionalfields led to the establishmentof plant cient at retaining and recycling nutrients acquired during communities dominated by summer annuals with high

356 OIKOS 72:3 (1995) photosynthetic rates and high relative growth rates. The dependenton the duration of the fertilization of the sea- resultant communities resembled earlier stages of sec- grass beds. Increased nutrient availability causeda dou- ondary succession. Similarly, McLendon and Redente bling of the Thalassia testudinumleaf biomass over con- (1991)noted that fertilization of a sagebrushsteppe com- trols for the first two years of this experiment; and were it munity allowed early-successionalannuals to persist as to have ended at that time we would have concluded that community dominants,and they concluded that the dom- increased nutrient supply to Florida Bay seagrasses inance of a site by annuals during the early stages of would causean increasein the biomassof the late succes- secondarysuccession is related to high nutrient availabil- sional seagrassT. testudinum.The true outcome of sucha ity. Successionin plant communities is dependenton the change in nutrient supply rates was dependent on the temporal pattern in the availability of nutrients. Increased colonization of these fertilized, and therefore newly suit- .. nutrient availability may alter the trajectory of succession able, areas by the early successionalseagrass Halodule . by allowing plant speciesadapted to high nutrient envi- wrightii. This time-dependentresult underscoresthe im- ~ ronmentsto displace establishedspecies (Grime 1979). In portance of designing field experiments of the proper areas of high nutrient availability in Florida Bay, H. duration to capture the dynamics of the system being wrightii will be the dominant late successionalspecies, studied. while T. testudinum will be the dominant late-succes- sional species in areasof low nutrient availability. Acknowledgements-This research was funded in part by the We propose that the change in speciesdominance from John D. and Catherine T. MacArthur Foundation,and through a Thalassia testudinum to Halodule wrightii was caused cooperative agreementbetween the U.S. National Park Service and the Univ. of Virginia (CA-5280-O-9009). We thank H. directly by a change in the supply rates of light and soil McCurdy, L. Lagera, R. Zieman, K. Halama and R. Bjork for nutrients. This shifted the outcome of competition be- help with the field work. We thank the administration and tweenthe two seagrassspecies. We have interpretedthese rangers of Everglades National Park for their tolerance and results to be consistentwith the resourceratio hypothesis logistical assistancewith this long-term study within the park. J. P. Grime, J. Dooley, R. Price, D. Tomasko and R. Chambers (Tilman 1982, 1985, 1987, 1988), which asserts that provided comments that greatly improved this paper. competition is the main determinantof vegetationdom- inance in both nutrient-rich and nutrient-poor environ- menUs.In contrast,the triangular model of plant strategies (Grime 1974, 1977, 1979)holds that the relative impor- tance of competition decreasesin nutrient-poor environ- References ments,where the ability of plant speciesto weatherstress Bazzaz,F. A. 1979. The physiological ecology of plant succes- becomesthe primary determinant of vegetation. Experi- sion. -Annu. Rev. Eco1. Syst. 10: 351-371. ments have shown that some aspects of competition do Bulthuis, D. A. and Woelker1ing,W. J. 1983. Effects of in situ nitrogen and phosphorusenrichment of the seagrassHetero- indeed decrease in intensity with an increase in stress zostera tasmanica (Martens ex Aschers.) den Hartog in (Carl(lpbelland Grime 1992), and a simulation model of Western Port, Victoria, Australia. -J. Exp. Mar. BioI. Ecol. successionbased on the triangular model has been shown 53: 193-207. to produce results consistent with successionaltheory Cambridge, M. L. and McComb, A. J. 1984. The loss of sea- grassesin Cockburn Sound, Western Australia. I. The time (Colasantiand Grime 1993). An alternative interpretation course and magnitude of seagrass decline in relation to of the results of these experiments, consistent with the industrial development. -Aquat. Bot. 20: 229-243. triangular model, ascribesthe dominanceof T. testudinum Campbell, B. D. and Grime, J. P. 1992. An experimental test of overH. wrightii in oligotrophic Florida Bay to a conse- plant strategy theory. -Ecology 73: 15-29. Capone,D. G. and Taylor, B. F. 1980. Microbial nitrogen cy- quence of the low nutrient demand of T. testudinum cling in a seagrasscommuity. -In: Kennedy, V. S. (ed.), causedby slow growth and efficient retention of retained Estuarine perspectives. Academic Press, New York, pp. nutrients compared to H. wrightii. In other words, T. 153-162. testudinumis more tolerant of the stressof low nutrient Carson,W. P. and Barrett, G. W. 1988. Successionin old-field plant communities: effects of contrasting types of nutriejlt availability than H. wrightii. The triangular model works enrichment. -Ecology 69: 984-994. well if we consider only one kind of stress,in this case Chapin, F. S. III. 1980. The mineral nutrition of wild plants. - nutri~nt limitation; but becomescumbersome once light Annu. Rev. Eco1. Syst. 11: 233-260. "'. stressis also considered,because light stressand nutrient Colasanti,R. L. and Grime, J. P.1993. Resourcedynamics ~d vegetation processes:a deterministic model using two-dl- ..~ stres~ do not act in concert in aquatic environments. mensionsal cellular automata. -Funct. Ecol. 7: 169-176. Often, light availability and nutrient availability are in- Connell, J. H. and Slayter, R.O. 1977. Mechanisms of succes- versely correlated. For nutrients, T. testudinumis clearly sion in natural communities and their role in, communiw the bettercompetitor; but it is also clear that H. wrightii is stability and organization. -Am. Nat. 111: 1119-1144. Dawes, C. J. and Tomasko, D. A. 1988. Depth distribution of a better competitor for light. In this case, the tradeoffs Thalassia testudinum in two meadows on the west coast of made between accumulationof light and accumulationof Florida; a difference in effect of light availability. -Mar. nutrients, as described by the resource ratio hypothesis, Ecol.9: 123-130. deteIJminesthe behavior of T. testudinumand H. wrightii den Hartog, C. 1971. The dynamic aspect in the ecology ~f seagrasscommunities. -Thalassia Jugosl. 7: 101-112. in Flbrida Bay. Fonseca,M. S., Thayer, G. W. and Kenworthy, W. J. 1987. The Interpretation of the results of this experiment was use of ecological data in the implementation and manage-

OIKO~ 72:3 (1995) 317

. ment of seagrassrestorations. -In: Durako. M. J., Phillips, tropical estuary: observationaland experimentalevidence. - R. C. and Lewis, R. R. III (eds), Proceedingsof the sympo- Estuarine Coastal Shelf Sci. 32: 567-579. sium on subtropical seagrassesof the southeasternUnited Rice, E. L. 1974. Allelopathy. -Academic Press, New YorK. States.Florida Marine ResearchPublications No. 42., Flor- Sand-Jensen, K. 1977. Effect of epiphytes on eelgrass photosyn- ida Dept of Natural Resources,St. Petersburg,FL, pp. 175- thesis. -Aquat. Bot. 3: 55-63. 188. Short, F. T. 1987. Effects of sediment nutrients on seagrasses: Fourqurean,J. W., Zieman, J. C. and Powell, G. V. N. 1992a. literature review and mesocosmexperiment. -Aquat. Bot. Relationships betweenporewater nutrients and seagrassesin 27: 51-57. a subtropical carbonate environment. -Mar. BioI. 114: -, Davis, M. W., Gibson, R. A. and Zimmerman, C. F. 1985. 57-65. Evidence for phosphorus limitation in carbonate sediments -, Zieman, J. C. and Powell, G. V. N. 1992b. Phosphorus of the seagrassSyringodium filiforme. -Estuarine Coastal limitation of primary production in Florida Bay: evidence Shelf Sci. 20: 419-430. from the C: N: P ratios of the dominant seagrassThalassia -, Dennison, W. C. and Capone, D. G. 1990. Phosphorus- testudinum.-Limnol. Oceanogr. 37: 162-171. limited growth of the tropical seagrass Syringodium fil- ., Gleeson, S. and Tilman, D. 1990. Allocation and the transient iforme in carbonate sediments.-Mar. Ecol. Prog. Ser. 62: . dynamics of successionon poor soils. -Ecology 71: 1144- 169-174. ". 1155. Silberstein, K., Chiffings, A. W. and McComb, A. J. 1986. The Grime, J. P. 1974. Vegetation classification by reference to loss of seagrassesin Cockburn Sound, Western Australia. strategies. -Nature 250: 26-31. III. The effect of epiphytes on productivity of Posidonia -1977. Evidence for the existence of three primary strategies australis Hook. F. -Aquat. Bot. 24: 355-371. in plants and its relevance to ecological and evolutionary Tilman, D. 1982. Resource competition and community struc- theory. -Am. Nat. III: 1169-1194. ture. -Princeton Univ. Press, Princeton, NJ. -1979. Plant strategies and vegetation processes.-Wiley, -1985. The resource ratio hypothesis of succession.-Am. New York. Nat. 125: 827-852. Harper, J. L. 1977. Population biology of plants. -Academic -1987. On the meaning of competition and the mechanismsof Press,London. competitive superiority. -Funct. Ecol. 1: 304-315. Inouye,R. S. and Tilman, D.1988. Convergenceand divergence -1988. Plant stategiesand the dynamics and structure of plant of old-field plant communities along experimental nitrogen commuities. -Princeton Univ. Press, Princeton, NJ. gradients. -Ecology 69: 995-1004. -and Wedin, D. 1991a. Dynamics of nitrogen competition Iverson, R. L. and Bittaker,H. F. 1986.'Seagrassdistribution in between successionalgrasses. -Ecology 72: 1038-1049. the eastern . -Estuarine Coastal Shelf Sci. -and Wedin, D. 1991b. Plant traits and resource reduction for 22: 577--6(J2. five grassesgrowing on a nitrogen gradient. -Ecology 72: McLendon, T. and Redente,E. F. 1991. Nitrogen and phospho- 685-700. rus effects on secondarysuccession dynamics on a semi-arid Tomasko, D. A. and Lapointe, B. E. 1991. Productivity and sagebrushsite. -Ecology 72: 2016-2024. biomass of Thalassia testudinumas related to water column McMillan, C. 1981. Seedreserves and seed germination for two nutrient availability and epiphyte levels: field observations seagrasses,Halodule wrightii and and experimental studies. -Mar. Ecol. Prog. Ser. 75: 9-17. from the western Atlantic. -Aquat. Bot. II: 279-296. Tomlinson, P. B. 1974. Vegetative morphology and meristem -1983. Seed germination in Halodule wrightii and Syringo- dependence-the foundation of productivity in seagrasses.- dium filiforme from and the U.S. Virgin Islands. - Aquaculture 4: 107-130. Aquat. Bot. 15: 217-220. Wiginton, J. R., and McMillan, C. 1979. Chlorophyll composi- Patriquin, D. G. 1972.The origin of nitrogen andphosphorus for tion under controlled light conditions as related to the distri- the growth of the marine angiosperm Thalassia testudinum. bution of seagrassesin Texas and the U.S. Virgin Islands.- -Mar.. BioI. 15: 35--46. Aquat. Bot. 6: 171-184. Phillips, R. C. 1960. Observation on the ecology and distribu- Willi..., S. L. 1987. Competition betweentheseagrasses Tha- tion of the Florida seagrasses.-Professional papers series lassia testudinum and Syringodiumfiliforme ina Caribbean No.2, Florida StateBoard of Conservation,Tallahassee, FL. lagoon. -Mar. Ecol. Prog. Ser. 35: 91-98. Pickett, S. T. A. 1976. Succession:an evolutionary interpreta- -1990. Experimentalstudies of Caribbeanseagrass bed devel- tion. -Am. Nat. 110: 107-119. opment. -Ecol. Monogr. 60: 449-469. Platt, W. J.1975. The colonization and formation of equilibrium -and Adey, W. H. 1983. 11zalassiatestudinum Banks ex plant species associations on badger mounds in a tall-grass Konig seedling successin a coral reef microcosm, -Aquat. prairie. -Ecol. Monogr. 45: 285-305. Bot. 16: 181-188. Powell, G. V. N. 1987. Habitat use by wading birds in a sub- Winer, B. J. 1971. Statistical principles in experimental design, tropical estuary: implications of hydrography. -Auk 104: 2nd ed. -McGraw-Hill, New York. 740-749. Zieman, J. C. 1982. The ecology of seagrassesof south Florida: -, Kenworthy, W. J. and Fourqurean, J. W. 1989. Experi- A community profile. -U.S. Fish and Wildlife Service BioI. mental evidence for nutrient lirnita~ion of seagrassgrowth in Rep. OBS-82/25. U.S. Fish and Wildlife Service, Washing- a tropical estuary with restricted circulation. -Bull. Mar. ton. -. Sci. 44: 324-340. -, Fourqurean,J. W. andIverson,R. L. 1989.The distribution, .' , -, Fourqurean, J. W., Kenworthy, W. J. and Zieman, J. C. abundanceand productivity of seagrassesin Florida Bay. - 1991. Bird colonies cause seagrass enrichment in a sub- Bull. Mar. Sci. 44: 292-311. ~ ~~

358 OIKOS 72:3 (1995)