Ecological Applications, 19(4), 2009, pp. 942–960� 2009 by the Ecological Society of America

Effects of agricultural subsidies of nutrients and detritus on fishand plankton of shallow-reservoir ecosystems

ALBERTO PILATI,1 MICHAEL J. VANNI,2 MARIA J. GONZALEZ, AND ALICIA K. GAULKE3

Miami University, Department of Zoology, Oxford, Ohio 45056-1400 USA

Abstract. Agricultural activities increase exports of nutrients and sediments to lakes, withmultiple potential impacts on recipient ecosystems. Nutrient inputs enhance phytoplanktonand upper trophic levels, and sediment inputs can shade phytoplankton, interfere with feedingof consumers, and degrade benthic habitats. Allochthonous sediments are also a potentialfood source for detritivores, as is sedimenting autochthonous phytodetritus, the production ofwhich is stimulated by nutrient inputs. We examined effects of allochthonous nutrient andsediment subsidies on fish and plankton, with special emphasis on gizzard shad (Dorosomacepedianum). This widespread and abundant omnivorous fish has many impacts on reservoirecosystems, including negative effects on water quality via nutrient cycling and on fisheries viacompetition with sportfish. Gizzard shad are most abundant in agriculturally impacted,eutrophic systems; thus, agricultural subsidies may affect reservoir food webs directly and byenhancing gizzard shad biomass. We simulated agricultural subsidies of nutrients andsediment detritus by manipulating dissolved nutrients and allochthonous detritus in a 2 3 2factorial design in experimental ponds.

Addition of nutrients alone increased primary production and biomass of zooplanktivo-rous fish (bluegill and young-of-year gizzard shad). Addition of allochthonous sediments aloneincreased algal sedimentation and decreased seston and sediment C:P ratios. Ponds receivingboth nutrients and sediments showed highest levels of phytoplankton and total phosphorus.Adult and juvenile gizzard shad biomass was enhanced equally by nutrient or sedimentaddition, probably because this apparently P-limited detritivore ingested similar amounts of Pin all subsidy treatments. Nutrient excretion rates of gizzard shad were higher in ponds withnutrient additions, where sediments were composed mainly of phytodetritus. Therefore,gizzard shad can magnify the direct effects of nutrient subsidies on phytoplankton production,and these multiple effects must be considered in strategies to manage eutrophication andfisheries in warmwater reservoir lakes where gizzard shad can dominate fish biomass.

Key words: agriculture; detritivore; detritus; ecosystem subsidy; eutrophication; fish; non-pointpollution; nutrients; reservoir; stoichiometry; water quality; watershed.

INTRODUCTION

Many ecosystems receive large, anthropogenic subsi-

dies of materials from outside their boundaries. Mid-

western reservoirs often receive enhanced nutrient and

sediment subsidies from agricultural watersheds that

lead to eutrophication and other water quality prob-

lems. Eutrophication in these systems is exacerbated by

nutrient cycling by gizzard shad (Dorosoma cepedia-

num), an abundant detritivorous fish that translocates

nutrients from sediments to the water column, thereby

promoting phytoplankton growth. Gizzard shad dom-

inate highly subsidized productive reservoirs in agricul-

tural watersheds, but are relatively scarce in less

subsidized unproductive reservoirs. However, little is

known about the separate and combined effects of

nutrient and sediment subsidies on fish (including

gizzard shad) and water quality in reservoir ecosystems.

We describe an experimental pond study in which we

simulated agricultural nutrient and sediment subsidies to

reservoir ecosystems and quantified the response of

detritivorous and zooplanktivorous fish and water

quality variables to these subsidies.

Ecologists strive to understand how ecosystems are

regulated by subsidies of resources such as detritus and

nutrients (Yang et al. 2008). Several studies have

experimentally examined effects of either detritus (e.g.,

Polis and Hurd 1996, Wallace et al. 1997, 1999, Chen

and Wise 1999, Hall et al. 2000) or nutrient subsidies

(Smith et al. 1999, Cross et al. 2006), but few have

examined how detritus and nutrient subsidies together

regulate ecosystems (Stelzer et al. 2003, Cross et al.

2007). For many years it was believed that lake

consumers are supported primarily by autochthonous

Manuscript received 29 April 2008; revised 13 August 2008;accepted 5 September 2008. Corresponding Editor: K. B. Gido.

1 Present address: Departamento de Ciencias Naturales,Universidad Nacional de La Pampa, Uruguay 151, 6300Santa Rosa, Argentina.

2 Corresponding author. E-mail: vannimj@muohio.edu3 Present address: University of North Carolina at Chapel

Hill, Institute of Marine Sciences, Morehead City, NorthCarolina 28557 USA.

942

primary production, which is often stimulated by

nutrient inputs. However, recent studies demonstrate

reliance on terrestrially derived (allochthonous) carbon

subsidies (Carpenter et al. 2005, Cole et al. 2006).

Allochthonous organic matter can be important in

supporting bacteria (Cole et al. 2002, Kritzberg et al.

2004), zooplankton (Grey et al. 2001, Cole et al. 2002,

Pace et al. 2004), and benthic invertebrates (Solomon et

al. 2008), but less is known about fish (Pace et al. 2007).

Carpenter et al. (2005) found that terrestrial sources can

account for up to 80% of the carbon flow to fish in small

lakes in forested watersheds, and that this percentage is

lower under nutrient-enriched conditions conducive to

high autochthonous (algal) production.

Agricultural activities increase the export of nutrients

and sediments to lakes, often leading to eutrophication

(Fig. 1). Agricultural runoff occurs over extensive,

diffuse areas and is therefore difficult to regulate

(Carpenter et al. 1998). The addition of nutrients to

lakes results in increased phytoplankton production,

some of which is consumed by zooplankton; therefore,

zooplanktivorous fish production might also be en-

hanced through this bottom-up process. Primary pro-

duction not consumed by zooplankton eventually settles

to the lake bottom as phytodetritus, and accumulation

of phytodetritus may enhance the production of

detritivorous fish (Fig. 1). Nutrients also accumulate in

lake sediments, and high rates of nutrient flux from

sediments to the water column can make it difficult to

reverse eutrophication (Carpenter 2005).

Agricultural activities also enhance the export of

particulate matter to lakes (Fig. 1). Indeed, high inputs

of suspended allochthonous sediments into aquatic

systems are characteristic of many agricultural land-

scapes (Lenat 1984, Waters 1995, Johnson et al. 1997).

These sediments are composed of both organic matter

(detritus containing C fixed in terrestrial ecosystems)

and inorganic materials. Inorganic particles such as clay

and silt can impair water quality by covering food

sources, degrading habitat and nesting sites, and

accelerating the rate at which a lake fills with sediment

(Waters 1995, Uri 2001, Allan 2004). Allochthonous

organic detritus may enhance nutrient supply (via

mineralization of this detritus) and/or provide food for

some bottom-feeding detritivores.

In midwestern U.S. reservoirs, where agricultural

activities have great impacts, fish biomass is often

dominated by gizzard shad, Dorosoma cepedianum, an

omnivore that has many impacts on reservoir food webs

and ecosystems (Vanni et al. 2005). This clupeid is

zooplanktivorous during larval stages but in reservoirs

relies almost entirely on sediment detritus in non-larval

stages (Higgins et al. 2006; Fig. 1). Thus, agricultural

subsidy effects may vary with subsidy type and life-

history stage. Larval gizzard shad may benefit from high

inputs of dissolved nutrients because of increased

zooplankton production (Bremigan and Stein 2001). In

contrast, sediment subsidies may have negative effects

on planktivores because suspended sediments may

impair visual feeding (Vinyard and O’Brien 1976,

McCabe and O’Brien 1983).

Detritivorous fish may benefit from both nutrient and

sediment subsidies from agriculture. Gizzard shad

biomass, and its contribution to total fish biomass,

increases greatly with lake productivity (Bachmann et al.

1996, DiCenzo et al. 1996, Michaletz 1997), and lakes

FIG. 1. Simplified conceptual model of how agricultural activities may influence a productive reservoir. Solid lines indicatefluxes of resources, and the dotted line represents excretion by gizzard shad. Nutrients from agricultural activities in the watershedincrease algal production in the lake, and thus zooplankton and planktivorous fish (larval gizzard shad and bluegill) production.The proportion of algae that is not consumed by zooplankton is sedimented to the bottom of the lake (autochthonous detritus).Also, erosion from the watershed increases inputs of allochthonous detritus. Both detritus types form lake sediments, the main foodsource of adult (and juvenile) gizzard shad (GS). Adults translocate nutrients from sediments to the water column via consumptionand excretion of sediment nutrients and have the potential to regulate algal nutrient limitation. Abbreviations are: YOY, young-of-year; POM, particulate organic matter.

June 2009 943SUBSIDY EFFECTS ON FISH AND PLANKTON

with watersheds that are dominated by agriculture often

have the highest gizzard shad biomass (Vanni et al. 2005,

2006). Juvenile and adult gizzard shad stimulate

phytoplankton by translocating nutrients from sedi-

ments to the water column through consumption and

excretion (Fig. 1); both the flux of nutrients through

gizzard shad populations and its relative importance in

sustaining phytoplankton production increase with lake

productivity, i.e., with increasing watershed agriculture

(Vanni et al. 2006). In addition, effects of gizzard shad

on zooplankton and fish are often stronger in productive

reservoirs (reviewed in Stein et al. 1995 and Vanni et al.

2005).

Because the abundance and effects of gizzard shad

vary greatly with productivity, it is important to

understand how sediment and/or nutrient subsidies

affect gizzard shad abundance. Specifically, it is not

clear if gizzard shad biomass is high in agriculturally

impacted lakes because of increased allocthonous

subsidies of dissolved nutrients, sediment detritus, or

both. These subsidies can also have direct and indirect

effects on many other aspects of shallow-reservoir

systems, which may involve feedbacks through gizzard

shad. For example, agricultural subsidies may stimulate

primary production and gizzard shad biomass, and

enhanced gizzard shad biomass may further stimulate

primary production (Vanni et al. 2005). In this research

we aimed to experimentally elucidate how subsidies of

sediments and nutrients affect shallow-reservoir systems,

with emphasis on gizzard shad populations and the

feedbacks between gizzard shad and phytoplankton.

This experiment, conducted in experimental ponds, was

designed to mimic conditions in midwestern U.S.

reservoirs; the rates of nutrient and sediments inputs

reflect those to Acton Lake (Vanni et al. 2001, Renwick

et al. 2005), a hypereutrophic reservoir with an

agricultural watershed (Knoll et al. 2003). We analyzed

fish growth and survival as measures of their response to

different subsidies, and gizzard shad excretion rates as

measure of their potential ecosystem impacts. We tested

three hypotheses:

1) Detritivorous fish (i.e., juvenile and adult gizzard

shad) biomass will be maximal with elevated inputs of

both allochthonous sediments (which provide a direct

food resource) and dissolved nutrients (which provide

food via increased phytodetritus deposition) because

under these conditions there is more detritus to

consume.

2) The success of zooplanktivorous fish (i.e., age-0

gizzard shad and bluegill) will be maximal under

conditions of elevated nutrients because of high

zooplankton production via increased phytoplankton

production.

3) Gizzard shad feeding on high-quality phytodetritus

will have the greatest potential ecosystem impact (per

fish) through nutrient recycling. The rationale is that

with nutrient enrichment, detritus will have high

amounts of N and P per unit sediment because it will

be largely phytodetritus. Thus, shad will excrete more N

and P than when feeding on allochthonous detritus,

which should have less N and P per unit sediment.

METHODS

Experimental design

We used a 2 3 2 factorial design with blocks in which

we increased the availability of allochthonous and/or

autochthonous detritus (see Plate 1). The treatments,

each with three replicates, were (1) no nutrient or

sediment addition (control), (2) nutrient addition (þN),

(3) sediment addition (þS), and (4) addition of nutrients

and sediment combined (þNþS).All treatments designed to stimulate autochthonous

detritus production (þN and þNþS) received weekly

additions of ammonium nitrate (NH4NO3) and sodium

phosphate (NaPO4 H2O) at loading rates of 150 lg N/L

pond water and 15 lg P/L pond water per week. These

loading rates are typical inputs to Acton Lake, a

hypereutrophic reservoir (Vanni et al. 2001). All

treatments with allochthonous detritus (þS and þNþS)first received a layer of Acton Lake sediments (see

Methods: Experimental ponds and setup), and then

received weekly additions of Acton Lake sediments

(0.06 m3; ;70 kg) as a slurry that was sprayed with a

pump. This weekly input rate mimicked the long-term

sedimentation rate in Acton Lake (1.7 cm/year; Renwick

et al. 2005) and represented twice the estimated

consumption rate of the shad population in each pond

at 258C (Vanni and Headworth 2004). In terms of

stoichiometric ratios, this slurry (C:P¼ 102.2, C:N¼ 9.9,

N:P ¼ 10.2) was similar in composition to particulate

allochthonous detritus entering this lake from its

watershed (C:P ¼ 121.3, C:N ¼ 13.3, N:P ¼ 9.1 over an

8-year period; for data from 1994 to 1998 see Vanni et

al. 2001; for data from 1999 to 2001 see M. J. Vanni,

A. M. Bowling, and A. D. Christian, unpublished data).

It is important to note here that the slurry added to the

ponds was composed of both allochthonous organic

detritus and inorganic sediments (e.g., clays), and it is

likely that a significant fraction of the P associated with

‘‘sediments’’ was adsorbed to inorganic particles (Now-

lin et al. 2005).

Experimental ponds and setup

The experiment was conducted in the experimental

pond facility at the Ecology Research Center, Miami

University in Oxford, Ohio, USA. The facility consists

of six rectangular ponds ;45 3 15 m, and 2.5 m

maximum depth, grouped in three pairs along an

elevation gradient (each pair represents a block; see

Plate 1). The ponds are lined with heavy-duty plastic

(XR-5 polyester geomembrane; Seaman Corporation,

Wooster, Ohio, USA) and thus have no sediments in

them. Ponds were divided in half with vinyl curtains,

which present effective barriers to water movements

between halves as revealed by the use of a rhodamine

ALBERTO PILATI ET AL.944 Ecological ApplicationsVol. 19, No. 4

tracer (M. J. Vanni, unpublished data). For simplicity, we

refer to these ‘‘half ponds’’ as ‘‘ponds’’ in this paper. All

four treatments were represented and randomized within

each block, except that the sediment treatments (þS and

þNþS) were always on the southern half of the ponds.

This is because the inlet pipe from which water enters

the pond is on the northern half, and we were concerned

that when filling up the ponds the inflow of water would

resuspend and carry sediments to ponds that were

supposed to be sediment free.

Before starting the experiment, we added ;5500 kg

(wet mass) of sediments from Acton Lake to ponds in

the þS and þNþS treatments so that some sediments

were present when fish were added. The sediments were

spread to form a ;2–3 cm layer in the bottom of the

ponds; because ponds were always well oxygenated,

sediments were always available to gizzard shad. This

amount of sediment represents eighty times the annual

consumption of gizzard shad at 258C (Vanni and

Headworth 2004). Sediments were collected from

shallow areas of Acton Lake, where the lake’s largest

inlet (Four Mile Creek) empties into the lake. This

shallow area is subject to high sediment deposition from

the watershed (Renwick et al. 2005), which is mainly

dominated by agricultural activities (Vanni et al. 2001,

Knoll et al. 2003). Thus, most of the sediments in this

area are likely allochthonous and derived from agricul-

ture.

Ponds were filled by gravity on 29 May 2004 with

water from an oligotrophic supply pond (6.8 lgchlorophyll/L) and then inoculated with 400 L of Acton

Lake water to provide an inoculum of species typical of

productive environments. On 10 June, all ponds were

stocked with gizzard shad at a biomass of ;175 kg wet

mass/ha. Gizzard shad biomass in Acton Lake is usually

higher than this on an areal basis (mean late summer

biomass ¼ 290 kg/ha, 1999–2005; M. J. Vanni, unpub-

lished data). However, the experimental ponds are

shallower than Acton Lake (mean depth 2 vs. 3.5 m),

so the biomass in the ponds was similar to that in Acton

Lake on a volumetric basis. We added gizzard shad of

two size classes: 40 individuals between 120 and 140 mm

(135 6 4.7 mm total length [TL]; mean 6 SE; referred to

as ‘‘juveniles’’) and 60 individuals between 180 and 200

mm (193 6 4.8 mm TL; referred to as ‘‘adults’’). Of

those, 10 fish of each size class were marked by clipping

the last dorsal fin ray to obtain an estimate of growth.

Based on Mundahl and Wissing (1988) and Schaus et al.

(2002), the small size class corresponded to age 1, and

the large size class to age 2þ. Our objective was to

replicate conditions of midwest U.S. reservoirs. Thus, we

also added 100 bluegill sunfish (Lepomis macrochirus; 92

6 1.8 mm; mean 6 SD) to each pond on 22 June (1.42

kg/pond which represents 50.5 kg/ha). Bluegill are often

the most abundant species after gizzard shad in

midwestern U.S. reservoirs (Garvey et al. 1998) and

are at least partially zooplanktivorous at this size.

Therefore we expected the bluegill to reduce zooplank-

ton abundance and body size to an extent similar to that

observed in reservoir ecosystems, where small zooplank-

ton taxa dominate assemblages.

We began sampling on 14 June, one day before we

added sediment slurries and nutrients for the first time.

The experiment lasted 11 weeks, and the last sampling

was on 30 August. On 8 September, we drained the

ponds, collected all fish, and recorded total length and

mass of all fish.

Response variables and analytical procedures

To measure gizzard shad success, we focused on

growth rate, survival, total biomass, and condition

factor over the duration of the experiment. Gizzard shad

growth was calculated from individually marked fish as

final minus initial mass. We did not directly measure

initial mass (to avoid stressing fish), but instead

estimated it using a length–mass relationship obtained

for shad from Acton Lake (W¼ 0.00963L2.9686; n¼ 67

fish, r2 ¼ 0.997, P , 0.001). Biomass at the end of the

experiment was obtained by weighing each fish with an

Ohaus (Pine Brook, New Jersey, USA) top-loading

balance. We also analyzed the condition factor (K) of all

gizzard shad and bluegill individuals as K ¼ 100 W/L3,

where W is the mass in grams and L is the total length in

centimeters (Williams 2000).

Gizzard shad spawned during the experiment, and

this presented an opportunity to assess young-of-year

(YOY) performance. We quantified YOY recruitment,

defined as the number and mass of YOY fish at the end

of the experiment, as well as YOY biomass.

Ponds were sampled weekly for physical parameters

(temperature, light, and oxygen); total suspended solids

(TSS) and non-volatile suspended solids (NVSS, i.e.,

inorganic suspended particles); and nutrients and

phytoplankton (as chlorophyll) to see if these factors

could help predict gizzard shad success. Temperature

and oxygen were measured with a YSI Model 58 meter

(YSI, Yellow Springs, Ohio, USA), and photosynthet-

ically active radiation (PAR) with a LI-COR model 1400

spherical quantum radiometer (LI-COR, Lincoln, Ne-

braska, USA). For all other parameters except zoo-

plankton, we collected water samples with a Jabsco Par-

Max 4 diaphragm pump (Jabsco, White Plains, New

York, USA), integrating the top 1.5 m of the water

column. Total suspended solids (TSS) was estimated by

filtering pond water onto an A/E glass fiber filter of

known mass and drying the filter for 24 h at 608C. NVSS

was calculated by ashing the same filter used for TSS

(5508C for 4 h) to burn off organic matter; filters were

then reweighed. Unfiltered water samples were assayed

for total N (TN) and total P (TP); samples were digested

with potassium persulfate and analyzed as soluble

reactive phosphorus (SRP) and NO3�, respectively,

using a Lachat auto-analyzer (Lachat Instruments,

Lakeland, Colorado, USA).

Chlorophyll samples were filtered onto A/E filters,

frozen, and extracted with acetone. The extract was read

June 2009 945SUBSIDY EFFECTS ON FISH AND PLANKTON

in the fluorometer before and after acidification with 0.1

mol/L HCl to account for the presence of phaeopig-

ments. Primary production (PPr) was assayed on five

dates by generating photosynthesis–irradiance curves in

each pond using 14C uptake measurements (Knoll et al.

2003). From the five PPr measurement dates we,

estimated pond-scale PPr over the duration of the

experiment using ambient light data from the Ecology

Research Center meteorological station (U.S. EPA

Clean Air Status and Trends [CASTNET] program;

data available online),4 weekly light profiles in the ponds,

and weekly chlorophyll values, using the program

PSPARMS (Fee 1990).

We also sampled zooplankton every two weeks with a

63-lmmesh net. Zooplankters were identified as nauplii,

copepods (calanoids vs. cyclopoids), cladocerans (to

genus), and rotifers (to species). Dry biomass was

calculated by measuring 10–20 randomly chosen indi-

viduals of each taxonomic group and using length–mass

relationships (Dumont et al. 1975, McCauley 1984).

Rotifer volume was calculated using geometric formulas

(McCauley 1984). Then the volume was converted to

fresh mass assuming a specific gravity of 1 and then

converted to dry mass assuming a dry : wet ratio of 0.1

(Doohan 1973).

Finally, because we hypothesized that the addition of

dissolved nutrients would increase shad growth via

increased phytodetritus production, we quantified fluxes

of algal and non-algal carbon and nutrients reaching the

bottom of the ponds with sediment traps (PVC, 5.2 cm

diameter and 30 cm tall) that were retrieved weekly. The

content of each trap was poured out and traps were

immediately returned to the bottom of the pond.

To calculate the algal-sedimentation flux, first we

measured the total amount of particulate C, N, P,

chlorophyll, and phaeopigments in sediment traps. To

estimate the amount of C, we filtered two aliquots of the

sediment trap slurry onto GF/F glass fiber filters. We

used one (non-ashed) filter to measure the amount of

total carbon present in the sample using a PerkinElmer

Series 2400 elemental analyzer (PerkinElmer, Waltham,

Massachusetts, USA). The second filter was ashed at

5008C for 4 hours, and then we measured the amount of

inorganic carbon left in the sample. Particulate-organic-

carbon (POC) flux (grams per meter squared per day)

was calculated as

POC flux ¼ total C� inorganic C

sediment trap area 3 time:

To estimate the fraction of organic-carbon flux com-

prised of algal carbon, we calculated the total amount of

pigments in the trap as the sum of chlorophyll (chl) and

phaeopigments (processed in the same manner as water

samples, as just described). To account for the phaeopig-

ments resulting from chl degradation so that we could

estimate the original amount of chl hitting the trap, we

multiplied the total amount of pigments by 0.875 (the

chl:total pigments ratio from seston samples). As

sestonic particles are composed not only of algae but

also of zooplankton and detritus, we then used the

POC:chlorophyll ratio (POC:chl) to estimate algal C.

We obtained the POC:chl ratio using the model

proposed by Cloern et al. (1995), which estimates the

ratio as a function of temperature, irradiance, and

nutrient-limited growth rate. Then we obtained algal C

as chl 3 POC:Chl and non-algal C as the difference

between total organic C and algal C. We then used the

Redfield ratio (mol/L C:N:P ¼ 106:16:1) to estimate

algal N and P being deposited in traps. Although we had

data on seston C:N:P, we did not use those ratios to

calculate algal N and P because non-algal material

contributed significantly to seston nutrients, especially in

treatments receiving sediment additions. Non-algal N

and non-algal P were calculated as the difference

between total particulate N and total particulate P in

the traps minus the estimated algal N and algal P,

respectively. Total particulate nitrogen from traps was

measured on non-ashed filters using a PerkinElmer

Elemental analyzer. Total particulate phosphorus in the

traps was measured using an HCl digestion to convert

the particulate phosphorus to SRP (Stainton et al. 1977).

To estimate total daily fluxes in the traps, we measured

particulate organic carbon (POC), total N (TN), and

total P (TP) in the traps, then estimated the flux (grams

per meter squared per day) by dividing the sediment trap

area (square meters) by the number of days the traps

were in the water (7 days). Mol/L ratios were calculated

using POC, TN, and TP present in the traps.

At the end of the experiment, we also took sediment

samples from the bottom of all ponds, dried them for 48

h at 608C, and processed them for C, N, and P in a

manner identical to sediment trap samples. Carbon and

nutrients in sediment traps reflect the deposition each

week, presumably with minimal resuspension (based on

trap dimensions) or processing by gizzard shad. In

contrast, sediments collected from the bottom of the

ponds at the end of the experiment reflect food

availability at that time, under conditions that include

resuspension and processing by shad. We measured C

and nutrients in both sediment traps and bulk sediments

because it is not clear which is the best indicator of food

availability for adult gizzard shad.

Excretion experiments

Differential gizzard shad excretion rates among

treatments could affect pond productivity and algal

biomass. To assess this possibility, we conducted

excretion experiments in each pond (treatment) in one

randomly selected block, at the end of the experiment (7

September), i.e., after fish had responded to the

conditions created during the experiment. Fish were

collected via electroshocking, as this seems to have little

effect on excretion rates (Mather et al. 1995). Collected4 hhttp://www.epa.gov/castneti

ALBERTO PILATI ET AL.946 Ecological ApplicationsVol. 19, No. 4

fish were kept in 2-L containers filled with filtered lake

water (GF/F glass fiber filter), and they were allowed to

excrete for 30 minutes to 1 hr. At the end of the

experiment, fish were euthanized and their total length

and mass were recorded. Water samples were filtered in

situ with GF/F filters, and soluble reactive phosphorus

(SRP) and ammonia (NH4þ) were measured using the

ascorbic acid (Greenberg et al. 1992) and the phenate

methods (Solorzano 1969), respectively.

Statistical analysis

All statistical analyses were done on log-transformed

data to stabilize variances, but figures are shown with

raw data to provide a better visual portrayal of trends.

We analyzed final fish biomass, growth, fish survival,

and condition factor using a two-way ANOVA with a

complete block design (PROC GLM; SAS release 9.1;

SAS Institute, Cary, North Carolina, USA). The

addition of nutrients (þN) or sediments (þS) were the

two factors, and position (elevation) was the block. In

general, the block effect and the interaction terms were

not significant, so we do not mention them unless they

were significant. We report the terms that were

significant at the 0.05 level. A post hoc Tukey test was

used to compare the treatment means.

Statistical analyses for some limnological variables

collected weekly were done using a two-way repeated-

measures ANOVA with a complete block design (PROC

GLM; SAS release 9.1). The two factors in the ANOVA

were sediment and nutrient additions. To assess

differences among the specific treatments, we used the

mean of all dates per pond, and these means were used

as observations using a post hoc Tukey test.

Multiple step-wise regressions were also used to better

understand factors that can predict YOY and non-YOY

final fish biomass. Predictor variables were algal C flux,

chlorophyll, total carbon (TC) flux, cumulative PPr, and

zooplankton biomass. Response and predictor variables

were log transformed before running the multiple

regressions.

To compare gizzard shad excretion rates among

treatments, we used analysis of covariance (ANCOVA),

with log N and P excretion rates (per individual fish) and

ratios (N:P) as dependent variables, treatment as the

categorical variable, and fish size (log wet mass) as a

covariate. To further investigate the diet effect, an equal

slope ANCOVA model was then fit and intercept

parameters were compared.

RESULTS

Phytoplankton and nutrients

Phytoplankton biomass (chlorophyll concentration)

was highest in treatments with nutrients added (þN and

þNþS; Fig. 2A). Control chlorophyll values remained

quite low during the experiment and never exceeded 14

lg/L. In contrast, the other treatments showed a more or

less consistent increase after a lag of three weeks from the

beginning of the experiment. Chlorophyll concentrations

(lg/L) averaged over the whole experiment were 9.7 6

3.0, 21.3 6 11.6, 31.6 617.8, and 40.9 6 15.6 for the

control,þS,þN andþNþS, respectively (all data shown

are means 6 SD). Two-way repeated-measures ANOVA

indicated that chlorophyll changed with time and

increased with the addition of nutrients and sediments

(see Appendix).

Primary production (PPr; Fig. 2B) peaked in late July

to early August in all treatments. Cumulative PPr during

the study period was 33.9 6 6.6, 58.4 611.5, 78.0 6 9.1,

and 92.5 6 2.6 g C/m2 for the control, þS, þNþS, andþN, respectively. The effect of nutrients and sediments

on PPr was significant, as was the interaction between

nutrients and sediments (Appendix).

Total phosphorus (TP) increased with the addition of

nutrients or sediments (Fig. 2C); the interaction was not

significant (Appendix). Mean TP values averaged over

the experiment were 26.9 6 3.8 for the control, 54.7 6

17.4 for theþN, 62.8 6 23.1 for theþS, and 106.3 6 44.2

lg P/L for the þNþS treatment.

Total suspended solids (TSS, which includes organic

and inorganic matter) increased over time (Fig. 2D) and

was affected by the addition of nutrients or sediments;

the N 3 S interaction was not significant (Appendix).

TSS in the control and the þN treatments were mainly

composed of organic particles (i.e., phytoplankton,

zooplankton, and organic detritus), because concentra-

tions of non-volatile suspended solids (NVSS) in these

treatments were almost negligible; in contrast, NVSS

was high in treatments with sediment addition (Fig.

2E).

Seston C:N, C:P, and N:P ratios decreased with the

addition of sediments (Fig. 2 F–H, Appendix). N 3 S

interactions were not significant for any of the three

seston ratios. Thus, all three stoichiometric ratios

(C:N, C:P, and N:P) were highest in the control and

the þN treatments, and lowest in the þS and þNþStreatments, and for all three seston ratios, sediment

additions had much stronger effects than nutrient

additions (Fig. 2F–H). The mean mol/L C:P ratio was

253 6 72.5 and 226 6 51.9 for the control and the þNtreatments, while the þS and þNþS treatments had

lower ratios (123 6 20.0 and 121 6 36.0, respectively).

The mean C:N seston mol/L ratios were 12 6 2.6, 12 6

2.6, 10 6 2.1, and 10 6 2.3 for the control, þN, þS,and þNþS, respectively.

Zooplankton

After addition of fish, maximum zooplankton bio-

mass was observed in the þS (117 lg dry mass/L) and

control (97 lg/L) treatments while the lowest biomass

was observed in the þN (58 lg/L) and þNþS (68 lg/L)treatments. A two-way repeated-measures ANOVA

indicated that total zooplankton biomass after the

addition of planktivorous fish was reduced by the

addition of nutrients (Appendix).

June 2009 947SUBSIDY EFFECTS ON FISH AND PLANKTON

The addition of planktivorous bluegill on the day

after the first zooplankton sampling apparently reduced

the biomass of large bodied calanoids and small bodied

nauplii. These groups declined by a factor of seven in the

control, a factor of four in theþN, a factor of 10 in the

þS, and a factor of 17 in theþNþS treatment (Fig. 3A–

D). In the þN, þS, and þNþS treatments, where the

highest abundance of planktivorous fish (bluegill plus

YOY shad) was observed (see Results: Zooplanktivorous

fish), calanoid biomass was reduced and never recovered

(Fig. 3B–D). In contrast, after the initial reduction,

nauplii started to recover immediately in the control and

the þS treatments (Fig. 3A, C), while in the þN and

þNþS (Fig. 3B, D) their biomass increased after three

weeks. The biomass of cyclopoids, which were never

abundant, also declined in the control andþN after fish

FIG. 2. Time series of (A) chlorophyll, (B) primary productivity (PPr), (C) total phosphorus (TP), (D) total suspended solids(TSS), (E) non-volatile suspended solids (NVSS), (F) seston C:N, (G) seston C:P, and (H) seston N:P for the different treatments:control (solid circle), þN (open square),þS (closed triangle), and þNþS (open diamond).

ALBERTO PILATI ET AL.948 Ecological ApplicationsVol. 19, No. 4

introduction, and never recovered (Fig. 3E, F). On the

other hand, cladocerans (dominated mainly by Diaph-

anosoma sp.) and rotifers increased during the experi-

ment (Fig. 3E–H). Cladocerans and rotifers were never

abundant (note scale differences in Fig. 3A–D vs. E–H),

except in the þN treatment, where they contributed up

to ;70% of the total zooplankton biomass in some

ponds late in the experiment (Fig. 3F).

Zooplanktivorous fish

Gizzard shad spawned in all treatments, resulting in

variable biomass of their YOY (Fig. 4A). Neither the

addition of nutrients nor sediments (Table 1) had a

significant effect on YOY recruitment at the a ¼ 0.05

level, but both effects were marginally significant (P ,

0.10). We also observed large differences in mean YOY

FIG. 3. Stacked areas showing the trend of the zooplankton major groups, calanoids and nauplii (A–D), and minorzooplankton groups, cyclopoids, cladocera, and rotifers (E–H), for the control (A, E), þN (B, F), þS (C, G), and þNþS (D, H).Note that major group biomass is plotted on a different scale than biomass of minor groups. The arrow on the x-axis indicateswhen zooplanktivorous bluegill were added. All dates are in 2004.

June 2009 949SUBSIDY EFFECTS ON FISH AND PLANKTON

size among treatments. The control ponds had the

lowest total YOY shad abundance (1–75 individu-

als/pond) but the largest individual size (11.2 6 2.8

g/individual), while treatments with nutrients (þN,

þNþS) had the highest abundance (320–2700 individu-

als/pond) but the smallest size (3.5 6 1.9 g/individual).

TheþS treatment had intermediate abundance (128–509

individuals/pond) and size (8.7 6 5.9 g/individual).

FIG. 4. Final biomass of (A) planktivorous gizzard shad young-of-year (YOY) and (B) bluegill. Also shown is the final biomassof (C) adult and (D) juvenile detritivorous gizzard shad found in the different treatments. Growth (final mass – initial mass) ofindividually marked (E) adult and (F) juvenile gizzard shad are shown for the different treatments. Adult mass decreased even whenthere was a slight increase in length, while juveniles increased their mass throughout the study period. Error bars are standarddeviations. Different letters indicate differences at P , 0.05 (Tukey test).

TABLE 1. F values and P values (in parentheses) of different factors from a two-way ANOVA.

Treatment Adult biomass Juvenile biomass YOY biomass Adult CF Adult growth Juvenile growth

N 7.46 (0.034)* 6.71 (0.041)* 4.81 (0.071) 7.29 (0.036)* 4.31 (0.093) 4.47 (0.088)S 11.08 (0.016)* 7.02 (0.038)* 5.97 (0.050) 10.92 (0.016)* 22.06 (0.005)* 13.46 (0.015)*N þ S 6.62 (0.042)* 1.50 (0.267) 1.99 (0.208) 2.10 (0.198) 1.58 (0.264) 0.42 (0.546)Block 0.83 (0.488) 1.99 (0.218) 1.76 (0.251) 1.51 (0.294) 1.42 (0.326) 0.51 (0.630)

Notes: The degrees of freedom for nutrient addition (N), sediment addition (S), and NþS are (1, 6), while for the block they are(2, 6). The variables are adult, juvenile, and young-of-year (YOY) gizzard shad biomass, condition factor (CF), and growth rate ofadult and juvenile gizzard shad.

* P , 0.05.

ALBERTO PILATI ET AL.950 Ecological ApplicationsVol. 19, No. 4

Based on step-wise multiple regressions, total carbon

flux to sediments and primary productivity together

explained 79% of the variation in YOY biomass (F2,11¼16.770, P ¼ 0.001; Table 2).

Bluegill biomass increased with addition of nutrients

(F1,6 ¼ 48.31, PNut , 0.001) but did not respond to

sediments (F1,6 ¼ 1.49, PSed ¼ 0.257; Fig. 4B). Bluegill

condition factor was also enhanced by the addition of

nutrients; thus, bluegill condition was lowest in the þStreatment (1.78 6 0.08) and highest in the þNþStreatment (1.95 6 0.08). Bluegill survival overall was

very high (on average .95% in all treatments) and was

not affected by treatments.

Detritivorous fish

The two non-YOY cohorts of gizzard shad responded

similarly to the addition of autochthonous and alloch-

thonous detritus. However, there were differences

among age groups with respect to specific treatment

effects. Two-way ANOVA indicated that adult shad

responded to the addition to either type of detritus

(Table 1), the interaction term was significant, and

biomass was higher in all three subsidy treatments

relative to the control (Fig. 4C). There was also a

significant effect of nutrients and sediments on juveniles

(Fig. 4D, Table 1). However, theþN andþS treatments

were not different from the control, while juvenile

biomass in theþNþS treatment was significantly higher

than the control. Based on step-wise multiple regres-

sions, seston P, chlorophyll, total carbon (TC) flux, and

zooplankton biomass together explained 89% of the

variation in non-YOY biomass (F4,11¼ 13.82, P¼ 0.002;

Table 2).

Adult gizzard shad seemed to be the only age class

that differed in condition factor among treatments. The

addition of nutrients and sediments increased the

condition factor over the control (Table 1). Thus, the

condition factor ranged from 0.70 6 0.07 in the control

to 0.84 6 0.07 in the þNþS treatment.

Adult and juvenile gizzard shad survival was similar

for all treatments, and it was overall very high during the

whole experiment (;92% averaged across treatments;

Pilati 2007). Survival was not significantly affected by

the addition of nutrients or detritus.

Gizzard shad growth obtained from marked individ-

uals was affected by the addition of allochthonous

detritus. Adult shad lost a mean of ;12% of their

biomass (Fig. 4E) in all treatments (possibly due to

spawning activity and an associated decline in feeding

rate), but mass loss was less with the addition of

allochthonous detritus (Table 1). Nutrient addition had

only a marginally significant (P , 0.10) effect on mass

loss (Fig. 4E). Two-way ANOVA of individually

marked juveniles indicated that the addition of alloch-

thonous sediments had a positive effect on juvenile shad

growth (Table 1). Furthermore, effects of nutrients and

sediments were additive; nutrients enhanced juvenile

growth by approximately two times, sediments by

approximately five times, andþNþS by eight times over

the control (Fig. 4F).

Sedimentation fluxes and sediment composition

Cumulative C, N, and P sedimentation fluxes (from

sediment traps) generally increased with addition of

nutrients, sediments, or both (Fig. 5). The sediment

treatments (þS andþNþS) were significantly higher than

the non-sediment treatments for C, N, and P fluxes,

except that N flux in the þS treatment was not

significantly different from that in the þN treatment

(Fig. 5A–C). Carbon sedimentation flux was always

dominated by organic C (Fig. 5A) and was increased by

the addition of sediments (F1,6¼32.44, PSed¼0.001) and

nutrients (F1,6¼ 6.54, PNut¼ 0.043). In comparison with

the control, the C flux was about two times higher in the

þN treatment, about three times higher in the þStreatment, and about four times higher in the þNþStreatment. N sedimentation flux (Fig. 5B) followed the

same pattern as C, but it was significantly increased by

both nutrients (F1,6¼ 9.69, PNut¼ 0.013) and sediments

(F1,6 ¼ 33.74, PSed , 0.001). Net P sedimentation flux

was also significantly affected by the addition of

nutrients (F1,6 ¼ 22.03, PNut ¼ 0.001) and sediments

(F1,6¼ 132.04, PSed , 0.001). As in the case of C and N,

nutrients increased the P sedimentation flux two times

over the control, but sediment addition enhanced P

deposition more than that of other elements (12 times in

the þS treatment and 19 times in the þN treatment,

relative to the control; Fig. 5C).

We also analyzed element ratios (C:P, C:N, and N:P)

of sediment trap flux (cumulative flux of one element

divided by cumulative flux of the other element). The

addition of sediments significantly decreased the C:P

ratio of sedimenting material (F1,6 ¼ 139.82, PSed ,

0.001). Therefore, the C:P ratio of sedimenting material

was highest in the control and theþN treatment, and the

TABLE 2. Results of stepwise multiple regressions.

Response variable andpredictor variable Estimate r2

Model Pvalue

Non-YOY biomass

Intercept �0.7431 0.002Seston P �0.5661 0.64Chlorophyll 0.6471 0.75Total carbon (TC) flux 0.2819 0.80Zooplankton biomass 0.2237 0.89

YOY biomass

Intercept �10.1519 0.001TC flux 1.7216 0.68Primary production 1.3879 0.79

Notes: Non-YOY biomass includes the biomass of age-1 andage-2þ gizzard shad together. The response variables are inkilograms of fish per pond obtained at the end of theexperiment. The predictor variables are the means during thewhole experiment: seston P (lg P/L), chlorophyll (lg/L), TCflux (mg C�m�2�d�1), zooplankton biomass (lg dry mass/L),and primary productivity (mg C�m�2�d�1).

June 2009 951SUBSIDY EFFECTS ON FISH AND PLANKTON

lowest in the þS and þNþS treatments (Fig. 5D). The

C:N ratio of sedimenting material significantly de-

creased with the addition of nutrients (F1,6 ¼ 10.19,

PNut ¼ 0.019) and sediments (F1,6¼ 9.85, PSed ¼ 0.020).

Thus, the highest sediment C:N ratio was found in the

control and the lowest C:N ratio in theþNþS treatment

(Fig. 5E). The N:P ratio of sedimenting material

decreased with the addition of sediments (F1,6 ¼100.50, PSed , 0.001), but nutrient addition had no

effect. N:P ratio (Fig. 5F) was higher in the control and

þN treatment (mean 25:1), and lower in the sediment

treatments (;12:1). In general, trends among treatments

in C:N:P ratios of sedimenting material were similar to

those for seston ratios (Fig. 2F–H).

The analysis of bulk sediments from the bottom of the

ponds at the end of the experiment indicated that C:P

and N:P ratios of these sediments followed trends

similar to those ratios obtained from the sediment traps.

That is, these ratios were significantly higher in the

control and þN treatments than in the þS and þNþStreatments (Fig. 5D, F). The lower C:P and N:P of

sediments in theþS andþNþS treatments might indicate

a strong influence of P adsorbed to inorganic particles in

these treatments. However, C, N, and P concentrations,

per milligram of dry sediment, were lowest in theþS and

þNþS ponds, and the highest in the control and þNponds (Table 3). This suggests that food quantity (C, N,

and P available) was diluted by inorganic materials in

FIG. 5. (A) Net carbon, (B) nitrogen, and (C) phosphorus sedimentation fluxes as obtained from sediment traps. Panel (A) alsoshows the proportions of particulate organic carbon (POC) in white and particulate inorganic carbon (PIC) in black. (D) C:P, (E)C:N, and (F) N:P elemental mol/L ratios of the sedimentation flux (solid bars) and bulk sediments obtained from the bottom of theponds at the end of the experiment (open bars). The carbon : element ratio refers to the total carbon found in the sample (organicþinorganic). The error bars indicate the standard deviation, and different letters show differences among the treatments at the P ,

0.05 level (Tukey test): a–c refer to sedimentation fluxes, while x–z refer to bulk sediments. The error bars in panel (A) represent thestandard deviation calculated from the total particulate carbon (PICþ POC).

ALBERTO PILATI ET AL.952 Ecological ApplicationsVol. 19, No. 4

theþS andþNþS treatments and that detritivorous fish

should have to select against inorganic particles, while

fish on the control and þN should have had access to

nutritional detritus more easily.

From sediment traps, we also estimated fluxes of algal

and non-algal C and nutrients. ANOVA indicated that

algal C sedimentation flux (Fig. 6A) increased with the

addition of sediments and nutrients (Appendix). We did

not run the ANOVA on algal N and algal P because

those values were estimated from algal C using a

constant (the Redfield ratio [see Methods]), so ANOVA

results would be the same as for algal C. Algal N

sedimentation represented ;80% of the total N flux in

all treatments (Fig. 6B). As for C, the highest algal N

sedimentation flux was observed in the sediment

treatments, although algal N comprised a greater

proportion of sedimentation flux in the ponds without

sediments added (Fig. 6B). Proportion of total P flux

composed of algal P was higher in the control and the

þN treatment (Fig. 6C), but the highest absolute flux

was observed in the sediment treatments. Mean algal P

flux during the whole experiment was 20.7 613.4 for the

control, 33.4 6 21.7 mg P�m�2�d�1 for theþN treatment,

56.9 6 30.4 mg P�m�2�d�1 for theþS treatment, and 76.7

6 34.6 mg P�m�2�d�1 for the þNþS treatment.

Gizzard shad excretion

N and P excretion rates and the N:P ratio excreted

varied significantly with treatment and thus presumably

with diet (ANCOVA; Fig. 7). Body size (log wet mass)

was significantly related to N excretion but not P

excretion; the effect of size on N:P excretion was

marginally significant. The interaction between treat-

ment and body size was not significant for either N or P

excretion or N:P excretion, indicating that allometric

scaling of excretion rates was similar among treatments.

For both N and P, there was a significant difference in

excretion rate among the four treatments (FN(4,35) ¼24.17, P , 0.001; FP(4,35) ¼ 8.08, P , 0.001). Also, for

both N and P, excretion rates were significantly greater

in the þN treatment than all other treatments (Fig. 7).

Phosphorus excretion rates did not differ among the

other three treatments, whereas N excretion rates were

significantly greater in the control than in the þS and

TABLE 3. Elemental composition of bulk sediments at the end of the experiment (mean 6 SD), and the estimated elementalcomposition and mol/L ratios of material ingested by detritivorous gizzard shad (Dorosoma cepedianum).

Treatment

Bulk sediment composition(lg/mg dry sediment)

Estimated ingested materialcomposition (lg/mg dry sediment)

Estimated ingestedmaterial mol/L ratios

Organic C N P Organic C N P C:N C:P N:P

Control 114.2 6 5.7 13.3 6 0.2 1.3 6 0.03 114.2 13.3 1.3 10 227 23þN 117.5 6 21.8 16.9 6 2.1 2.5 6 0.26 117.5 16.9 2.5 8 121 15þS 23.0 6 4.8 2.7 6 0.8 0.6 6 0.05 85.1 23.5 2.2 4 99 23þNþS 27.9 6 3.8 3.2 6 0.3 0.7 6 0.02 103.2 27.8 2.6 4 103 24

Notes: To obtain estimated ingestion composition, we used a feeding selectivity of 3.7 for organic C and P and a selectivity of 8.7for N (Higgins et al. 2006) in the sediment treatments (þS and þNþS). We assumed no selective feeding in the control and þNtreatments.

FIG. 6. Algal and non-algal cumulative sedimentation fluxof (A) particulate organic carbon, (B) nitrogen, and (C)phosphorus during the whole experiment (77 days) as obtainedfrom weekly sediment traps. In panel (A), different lettersindicate differences at P , 0.05 (Tukey test) for the algal fluxonly. Tukey tests were not applied to panel (B) or (C) as the Nand P fluxes were calculated using the Redfield ratio, andtherefore the results are the same as panel (A).

June 2009 953SUBSIDY EFFECTS ON FISH AND PLANKTON

þSþN treatments (the latter two treatments did not

differ from each other). Thus, both P and N excretion

rates were higher in ponds in which only autochthonous

detritus production was stimulated (þN) than in ponds

with allochthonous detritus additions (þS, þNþS) (Fig.7A, B). The excreted N:P ratio was lowest in the þNþStreatment (FN:P(4,35) ¼ 3.56, P ¼ 0.015), supporting the

idea that diet can mediate excretion ratios (Fig. 7C).

DISCUSSION

Our results show that agricultural subsidies have

strong and diverse effects on simulated shallow-reservoir

ecosystems. Nutrient subsidies had a direct positive

effect on primary production and enhanced zooplank-

tivorous fish biomass, while sediment subsidies had

stronger effects than nutrients on seston and sediment

C:P and C:N ratios. Systems receiving both kinds of

detritus showed maximal chlorophyll and total phos-

phorus levels. These subsidies also had diverse effects on

fish. Detritivorous gizzard shad responded similarly to

subsidies of either nutrients alone (which stimulated

production of autochthonous detritus) or allochthonous

sediments alone, suggesting that these fish can take

advantage of either subsidy. Therefore, the hypothesis

that detritivorous fish biomass would be maximal with

addition of both nutrients and sediment was not

supported. The hypothesis that zooplanktivorous fish

biomass would be maximal under conditions of elevated

nutrients was supported for bluegill and YOY gizzard

shad (the latter effect was marginally significant). The

hypothesis that detritivorous fish feeding on detritus

with the highest N and P (per unit mass) would excrete

more N and P than those feeding on detritus with a

lower amount of N and P was also supported, as

excretion rates were highest in the þN treatment. We

also note that we successfully simulated reservoir trophic

status in our treatments. Final chlorophyll and total

phosphorous (TP) values in the þNþS treatment were

similar to those in Acton Lake near Oxford, Ohio, USA,

a productive reservoir with an agricultural watershed. In

addition, chlorophyll and TP values in the control

treatment were similar to those in Burr Oak Lake, a

relatively low productivity reservoir with a primarily

forested watershed located in southeastern Ohio, within

Burr Oak State Park (Knoll et al. 2003).

Phytoplankton and nutrients

We expected low chlorophyll in the þS treatment

because of shading by suspended sediments, but this was

not the case as þS chlorophyll was intermediate that of

the control and the þN treatment (Fig. 2A). The

observed increase in phytoplankton (relative to the

control) in theþS treatment was 53% that of the increase

in theþN treatment (11.6 vs. 21.9 lg chlorophyll/L). We

can potentially account for this increase by considering

three different P fluxes. First, in a short-term experiment

in which we incubated sediment slurries, we observed

that each slurry released the equivalent of 1.2 g P/week,

which corresponds to 15% of P added weekly in

dissolved form in the þN and þNþS treatments (8

g P/week). Second, P released from the sediments on

the bottom of the ponds would have released ;1.4

g P/week, assuming that P is released at the same rate as

from shallow aerobic sediments in Acton Lake (Nowlin

et al. 2005). Because our ponds always had .5 mg O2/L,

it is probably valid to assume similar release rates. This

flux from sediments is 18% of the amount we added as

nutrients. Third, nutrient translocation via excretion by

gizzard shad could potentially explain some of the

phytoplankton enhancement, if excretion rates in theþS

FIG. 7. Mass excretion rates of (A) phosphorus and (B)nitrogen and (C) variation in excretion N:P (mol/L) for thedifferent treatments. In each graph the results of ANCOVAexamine the differences in mass excretion rates and N:P mol/Lratio for fish grown under different diets (ponds or treatments)and the effect of size (covariate). P values , 0.05 indicate asignificant effect of diet or fish size on mass-specific excretionrates or N:P ratio. ANCOVA was run on log-transformed data.

ALBERTO PILATI ET AL.954 Ecological ApplicationsVol. 19, No. 4

andþNþS treatments were higher than those in the non-

sediment treatments (control, þN) early in the experi-

ment (recall that we measured excretion rates only at the

end of the experiment). We suspect that this was the

case, as the sediment layer in the non-sediment

treatments probably was not developed early in the

experiment. Thus gizzard shad probably had more food

available in the treatments with sediments present than

in the other treatments. Using the observed P excretion

value for theþS treatment, and assuming that the daily

P excretion rate is 82% that of the maximum excretion

observed at midday (Schaus et al. 1997), we estimate

that P translocation by gizzard shad was 1.8 g P/week,

equal to 23% of the P added weekly as nutrients. Thus, P

released from the slurries, bottom sediments, and by

gizzard shad could supply a maximum of ;56% of the P

added as dissolved nutrients to the þN and þNþStreatments. Primary production responded similarly; the

increase in theþS was 55% that of the increase in theþNtreatment, compared to the control. Thus, these internal

nutrient cycling processes probably offset some of the

shading effects of sediments on phytoplankton growth.

Zooplankton

The lowest zooplankton biomass was observed in

ponds with the highest zooplanktivorous fish densities.

This low zooplankton biomass might actually corre-

spond to high zooplankton production, as species with

short generation times (rotifers and small cladocerans)

increased after addition of fish. The control and the þStreatments generally showed the highest zooplankton

biomass. The control ponds had the lowest shad YOY

biomass and low bluegill biomass. YOY shad in control

ponds grew very quickly, and given their size, most likely

switched to detritivory at an earlier date, releasing

zooplankton from predation pressure. In contrast,

small-bodied rotifers and nauplii dominated zooplank-

ton biomass in the sediment treatments. This dominance

could have been the result of a size-selective predation

(Brooks and Dodson 1965), and/or it could be due to

sediments interfering with larger taxa, especially cladoc-

erans by decreasing carbon ingestion rates (Arruda et al.

1983, McCabe and O’Brien 1983, Kirk 1992). In our

case, this effect was clearly seen in calanoids and

cyclopoids, whose biomass was reduced after the

addition of bluegill and never recovered to pre-addition

levels. Rotifers, on the other hand, tended to increase in

spite of the presence of planktivorous bluegill and larval

shad, probably as the result of a reduced predation by

other invertebrates (i.e., cyclopoids; Brandl 2005).

Zooplanktivorous fish

Nutrient addition increased the biomass of both YOY

gizzard shad and bluegill. For YOY gizzard shad the

nutrient effect was only marginally significant, and their

growth rate was apparently density dependent; size and

condition factor were highest in the control, where they

were least abundant. Larval gizzard shad are zooplank-

tivorous, and they could have had a strong top-down

effect on zooplankton. Contrary to our prediction, the

addition of nutrients (and the resulting increase in

phytoplankton) did not increase zooplankton biomass,

PLATE 1. Aerial view of one block of experimental ponds, toward the end of the experiment. Half-ponds, which are the units ofreplication, were isolated by dividing ponds with vinyl curtains, which are visible in the midsection of each pond. Each half-pondmeasured ;15 3 22.4 m. Photo credit: A. Pilati.

June 2009 955SUBSIDY EFFECTS ON FISH AND PLANKTON

presumably because zooplankton were kept under a

strong top-down control by the YOY gizzard shad and

bluegill, as predicted by food chain models (e.g.,

Hairston et al. 1960; Fig. 1). Instead, nutrients increased

the biomass at the next trophic level: zooplanktivorous

fish. Treatments with the highest YOY shad and bluegill

biomass corresponded to the treatments with the lowest

zooplankton biomasses and the highest chlorophyll

levels. Also, as expected, bluegill biomass was greatly

enhanced by nutrient addition but was unaffected by

sediment additions. Bluegill are visual carnivores and do

not consume detritus. Suspended sediments could

interfere with their feeding and have negative effects

on their growth (Miner and Stein 1993); however, this

may have been offset by the increase in phytoplankton

(and thus possibly zooplankton production) in the

treatments receiving sediments.

Detritivorous fish

The hypothesis that adult and juvenile gizzard shad

biomass would be maximal with addition of both types

of detritus was not supported for either age class. The

similar response of detritivorous shad biomass to all

subsidy treatments (Fig. 2B, C) may relate to the

combined effects of three interrelated factors: food

quantity, food dilution, and food quality. If fish biomass

was affected only by food quantity, biomass should have

been maximal in theþNþS treatment (because sedimen-

tation flux was highest; Fig. 6), and this was not the case.

If fish biomass was affected only by food availability or

‘‘food dilution,’’ we should have observed the highest

biomass in theþN treatment. This is because detritus in

the þN treatment had much more organic C and

nutrients, per unit mass, than detritus in the sediment

treatments (Table 3). Thus, shad in the þN treatment

may have obtained food more efficiently than fish in the

sediment treatments (þS,þNþS). Dilution of energy and

nutrients by inorganic matter (clays and silt) in

treatments with sediment additions would presumably

result in lower intake rates of energy and nutrients,

unless shad can increase consumption rates under such

conditions. Detritivorous gizzard shad feed continuous-

ly but only during daytime (Pierce et al. 1981); thus it is

unlikely that they could have increased consumption by

feeding for longer periods of time.

If fish biomass was affected only by stoichiometric

food quality and shad are nutrient limited (as opposed

to energy limited), we should have observed maximal

biomass in the þS and þNþS treatments because

sediments in these treatments had lower C:N and C:P.

It is worth noting here that final bulk sediment

composition in the þS and þNþS treatments was very

similar to that of sediments in eutrophic Acton Lake for

C and P (21 lg C/mg and 0.7 lg P/mg; Higgins et al.

2006). This indicates that sediment nutrient concentra-

tions did not change significantly during the experiment,

except for N concentration, which was ;1.5 times

higher than in Acton Lake. It also shows that our

sediment additions effectively reproduced conditions in

productive reservoirs. Our bulk sediment C:P ratio

(;110; Fig. 5D) was also very close to the mean C:P of

particulate matter delivered to in Acton Lake from

streams (C:P¼ 124, averaged across 5 years; Vanni et al.

2001). Also, it is important to note that the bulk

sediment C:P ratio (;110; Fig. 5D) is much lower than

detritus of plant origin. The C:P of terrestrial autotrophs

can range from ,250 to .3500, with a median of ;800

(Elser et al. 2000). The low C:P ratio in our sediments

might be influenced by inorganic particles (e.g., clays) to

which P was adsorbed. Acton Lake sediments contain a

variety of P fractions in sediments (Nowlin et al. 2005).

In shallow areas (near where we collected sediments),

organically bound P (which is presumably relatively

digestible) comprised ;50% of sediment P. Calcium-

bound, Fe-bound, and Al-bound P comprised 19%,

16%, and 8% of sediment P, respectively; presumably

these forms are less available to assimilation by fish.

Because P (and presumably other elements in sediments)

consist of several fractions that likely vary in digestibil-

ity, we cannot necessarily relate C:N:P ratios and overall

food quality.

It is possible that the similarity in responses to

nutrient and sediment additions can be explained by

selective feeding. In reservoirs, gizzard shad selectively

feed on high-quality sediment detritus (Mundahl and

Wissing 1988, Higgins et al. 2006), apparently by

selectively ingesting low-density sediment particles that

are relatively enriched in nutrients and organic C

(Smoot 1999). Mundahl and Wissing (1988) and Higgins

et al. (2006) found that gizzard shad selectively

consumed detritus with higher C, N, and P concentra-

tions. Therefore, we evaluated the possibility that

selective feeding mediated the response of shad to our

manipulations (Table 3). We assume that shad feeding

selectivity was similar in treatments with sediments (þSand þNþS) to that observed in Acton Lake (which has

sediment C, N, and P concentrations similar to these

treatments). Thus, we can multiply the sediment C, N,

and P concentrations in these treatments by the

selectivity coefficients reported by Higgins et al. (2006)

for Acton Lake (see Table 3 notes), where the selectivity

coefficient for each element is the concentration in

ingested detritus (per gram dry mass) divided by the

concentration in sediments (per gram dry mass). Doing

so produces C, N, and P concentrations of ingested

material that are very similar to ambient sediments in

the control and þN treatments (Table 3). Thus, gizzard

shad feeding selectively (to the same extent as observed

in reservoirs) on sediments in the þS and þNþStreatments will ingest similar amounts of C, N, and P

as fish feeding non-selectively in the control and þNtreatments, assuming that overall consumption rates are

similar. This could explain the similar biomass response

in all treatments with subsidies (þN,þS, andþNþS). Thefact that adult and juvenile shad biomass was higher in

the þN treatment than in the control (despite similar

ALBERTO PILATI ET AL.956 Ecological ApplicationsVol. 19, No. 4

stoichiometric compositions) also suggests that food

quantity is important.

Interpreting the response of gizzard shad to these

manipulations is compromised by a lack of knowledge

on the limiting resource, i.e., is gizzard shad limited by

C, N, or P? We evaluated this possibility using data from

theþNþS treatment and from Acton Lake. To do so, we

calculated threshold element ratios (TER) as recently

done for several aquatic consumers by Frost et al.

(2005). TER is the ratio that defines the boundary

between energy (C) vs. nutrient limitation; thus, animals

consuming food with a C:nutrient ratio above the TER

are considered limited by that nutrient, while consump-

tion of food with a C:nutrient ratio below the TER

indicates energy limitation (Sterner and Elser 2002,

Frost et al. 2005). TER was calculated as

TERC:nutrient ¼Anutrient

ðIC 3 ACÞ � RC½ �=IC

� �3

QC

Qnutrient

� �

where Anutrient and AC are the assimilation efficienciesfor the nutrient and for C; IC and RC are mass-specificC ingestion and respiration rates, respectively (mgC�mg C�1�d�1); and QC and Qnutrient are fish C andnutrient contents (proportion of dry mass). Weestimated IC by first calculating sediment dry massconsumption using the equations in Vanni and Head-worth (2004), which predict dry mass consumption as afunction of fish mass and temperature. Then weobtained IC for both juvenile and adult shad usingthe sediment C content (per unit dry mass) in theþNþStreatment and the selectivity coefficient for C (Table 3).This yielded IC values of 0.103 and 0.078 mg C�mgC�1�d�1 for juveniles and adults, respectively. We set AC

to 0.50 and AN to 0.77 (Mundahl and Wissing 1988),and AP to 0.68 (Vanni and Headworth 2004). RC wasset equal to 0.030 for juveniles and 0.028 for adults; thiswas derived from data on growth, assimilation andconsumption in Acton Lake (Vanni and Headworth2004). Finally, QC, QN, and QP were set at 0.40, 0.095,and 0.03 (Higgins et al. 2006).

Calculation of TER suggests that gizzard shad weremore likely to be P limited than energy limited in theþNþS treatment, once we account for selective feeding.The TERC:P was 45 and 65, respectively, for juvenilesand adults, whereas the C:P of ingested detritus wasestimated to be 103 (Table 3). In contrast, the TERC:N

was 16 and 23 for juveniles and adults, respectively,while ingested detritus had a C:N of 4.3; this indicatesthat energy was more likely to be limiting than N.Finally, we also calculated TERN:P and found it to be2.8 for both juveniles and adults; because ingesteddetritus had an N:P of 23.8, gizzard shad were morelikely to be limited by P than by N.

If shad were more limited by P than C or N duringthese experiments, we may have expected shad toperform best under conditions in which they ingestedthe most P. According to our ingestion estimates, which

incorporate selective feeding, shad ingested similar

amounts of P in the þN, þS, and þNþS treatments,

and much less P in the control treatment (Table 3).

Patterns of estimated P ingestion thus correspond well

with biomass response (i.e., similar and high biomass in

the subsidized treatments, much lower biomass in the

control). In contrast, patterns of C or N consumption

(Table 3) do not correspond as well to biomass

responses (Fig. 4 vs. Table 3). Thus, our ingestion and

TER estimates suggest that shad responded similarly in

the þN, þS, and þNþS treatments because these

treatments allowed them to consume P at similar rates.

We acknowledge that this conclusion rests on the

assumption that selective feeding behavior is similar in

eutrophic reservoirs and in ourþS andþSþN ponds, and

that there was no selective feeding in ponds without

sediment additions (this was necessary, as we have no

information on feeding selectivity under such condi-

tions). Further, our analysis assumes that all forms of

elements are equally digestible, which may not be the

case, as already mentioned.

The study of individually marked fish indicated that

juveniles gained mass during the experiment, while

adults lost mass. The cause of mass loss in adults may

be due to spawning activity and reproduction, as well as

high temperatures. Apparently, both sexes of gizzard

shad lost mass at similar rates, as mass loss was not

bimodally distributed (rather, it was normally distribut-

ed; data not shown). The mass loss in females could be

partly the result of release of eggs. Bodola (1966) found

that in Lake Erie, sexually immature age-1 gizzard shad

grew rapidly, while sexually mature females lost 10.7%

of their mass during spawning, on average. This mass

loss is similar to that (;12%) in our study. Mass loss in

males could be due to spawning activity as well, because

both male and female gizzard shad feeding rates decline

greatly prior to and during spawning (Pierce et al. 1981).

Studies in Acton Lake show that growth rates of adult

gizzard shad in some years are very low (essentially 0 in

some cases) in early and midsummer, at which time body

caloric content also declines (Pierce et al. 1980, Mundahl

and Wissing 1987). Low growth rates and depletion of

body energy reserves in summer may be due to warm

temperatures that increase respiratory costs (Pierce et al.

1980). Thus, it is possible that gizzard shad in our ponds

lost mass early in the experiment because of spawning

and reproduction activities, and subsequent growth rates

were low (or zero) because of warm temperatures.

Multiple regression indicated that gizzard shad

biomass may be predicted by a few limnological

variables. Total carbon (TC) flux (algal and non-algal)

accounted for a high percentage of the variability in

both YOY and non-YOY biomass. Nevertheless,

gizzard shad may have also benefited from phytodetritus

deposition per se (Fig. 1). Algal sedimentation increased

with the addition of nutrients and sediments. This is not

surprising, as lakes with high nutrient inputs are known

to have high sedimentation rates when compared to less

productive lakes (Baines and Pace 1994). High concen-

June 2009 957SUBSIDY EFFECTS ON FISH AND PLANKTON

trations of sediments can also increase algal sedimenta-

tion, and estimated algal flux was always higher in the

sediment-addition treatments (compare Fig. 6A–C).

Avnimelech et al. (1982) and Soballe and Threlkeld

(1988) found that clays increase algal settling by forming

clay–algal flocs. Therefore, the fact that fish growth was

highly affected by the addition of sediments does not

necessarily mean that gizzard shad relied exclusively on

allochthonous detritus, but it could mean that they

benefited from increased algal sedimentation. Unfortu-

nately we cannot distinguish these mechanisms with our

data.

Gizzard shad excretion rates

Gizzard shad excretion rates were higher in ponds

with high sediment N and P content (compare Fig. 7

with Table 3). This agrees with stoichiometry theory

(Sterner and Elser 2002), and with Higgins et al. (2006),

who found that gizzard shad excreted N and P at higher

rates in lakes where ingested sediments had higher

nutrient contents. However, stoichiometry cannot ex-

plain all of the variance among treatments in excretion.

In particular, note that P excretion rates were low in the

þS treatment even though the C:P of sediments in that

treatment was low (suggesting high P availability). This

may be because fish were unable to assimilate some of

the P associated with clays. Excretion N:P ratios were

similar to the Redfield ratio (16:1) except in the þNþStreatment, where it was lower (;13:1). These excretion

ratios fall within the range reported for gizzard shad in

reservoirs (generally 14–17; Schaus et al. 1997, Higgins

et al. 2006).

These generally low excretion N:P ratios may have

potentially important consequences. For example, ex-

periments in Acton Lake show that gizzard shad can

stimulate algal growth through nutrient translocation

when they feed on lake sediments (Schaus and Vanni

2000). Particularly, P translocation by gizzard shad from

sediments to the water column can equal or exceed P

supply from inflow streams in Acton Lake during the

summer dry period (Schaus et al. 1997, Nowlin et al.

2005, Vanni et al. 2005). Since Acton algae are often P

limited (Vanni et al. 2006), shad could enhance algal

growth by reducing P limitation and promote a positive

feedback between phytodetritus production and higher

fish biomass (Fig. 1). The low excretion N:P ratio in the

þNþS treatment could have reduced P limitation and

stimulated algal biomass in the þNþS treatment, thus

increasing algal P deposition, which was highest in the

þNþS treatment.

General implications

Nutrient and sediment inputs into Acton Lake are

declining due to changes in agricultural practices

(Renwick et al. 2005, 2008), including a large increase

in the use of conservation tillage (including no-till) and

better management of fertilizers and manure. In the

Acton watershed, the shift toward conservation tillage

was fostered by economic incentives provided to

farmers, with the goal of reducing lake sedimentation

rates (USDA 1992, Renwick et al. 2008). These changes

in agricultural practices are likely to continue, as they

have in other agricultural watersheds in the U.S.

Midwest (Richards and Baker 2002, Peterson 2005). If

reservoirs receive less nutrients and sediments, they may

become less eutrophic via direct and indirect effects.

Based on our experiment, we predict that reduced

sediment and nutrient subsidies would result in (1) less

recruitment of YOY gizzard shad, although they may

grow faster; (2) slower growth of adult gizzard shad,

which ultimately may lead to reduced population density

and biomass; and (3) increased N:P ratio of gizzard shad

excretion (see Fig. 7C), possibly leading to an increased

phytoplankton P limitation. Nutrient excretion by

gizzard shad can be considered an internal source of P,

and reducing internal P loading is an important

consideration for changing trophic status (Carpenter et

al. 1999). Because nutrient translocation by gizzard shad

can supply as much P as that entering from the

watershed (Schaus et al. 1997, Nowlin et al. 2005),

reduced P excretion may speed the oligotrophication

process.

ACKNOWLEDGMENTS

We are grateful for help in the field and lab from all membersof the Vanni and Gonzalez labs, and to the staff of MiamiUniversity’s Ecology Research Center. Experiments weresupported by a National Research Initiative (NRI) grant(OHOR-2003-01756) from the U.S. Department of Agriculture,NSF grant DEB 02-0235755, and a Summer Workshop grant(Department of Zoology, Miami University).

LITERATURE CITED

Allan, J. D. 2004. Landscapes and riverscapes: the influence ofland use on stream ecosystems. Annual Review of Ecology,Evolution, and Systematics 35:257–284.

Arruda, J. A., G. R. Marzolf, and R. T. Fault. 1983. The role ofsuspended sediments in the nutrition of zooplankton inturbid reservoirs. Ecology 64:1225–1235.

Avnimelech, Y., B. W. Troeger, and L. W. Reed. 1982. Mutualflocculation of algae and clay: evidence and implications.Science 216:63–65.

Bachmann, R. W., B. L. Jones, D. D. Fox, M. Hoyer, L. A.Bull, and D. E. Canfield. 1996. Relations between trophicstate indicators and fish in Florida (USA) lakes. CanadianJournal of Fisheries and Aquatic Sciences 53:842–855.

Baines, S. B., and M. L. Pace. 1994. Relationships betweensuspended particulate matter and sinking flux along a trophicgradient and implications for the fate of planktonic primaryproduction. Canadian Journal of Fisheries and AquaticSciences 51:25–36.

Bodola, A. 1966. Life history of the gizzard shad, Dorosomacepedianum (Le Sueur), in western Lake Erie. Fishery Bulletinof the Fish and Wildlife Service 65:391–425.

Brandl, Z. 2005. Freshwater copepods and rotifers: predatorsand their prey. Hydrobiology 546:475–489.

Bremigan, M. T., and R. A. Stein. 2001. Variable gizzard shadrecruitment with reservoir productivity: causes and implica-tions for classifying systems. Ecological Applications 11:1425–1437.

Brooks, J. L., and S. I. Dodson. 1965. Predation, body size, andcomposition of plankton. Science 150:28–35.

ALBERTO PILATI ET AL.958 Ecological ApplicationsVol. 19, No. 4

Carpenter, S. R. 2005. Eutrophication of aquatic ecosystems:bistability and soil phosphorus. Proceedings of the NationalAcademy of Sciences (USA) 102:10002–10005.

Carpenter, S. R., N. F. Caraco, D. L. Correll, R. W. Howarth,A. N. Sharpley, and V. H. Smith. 1998. Nonpoint pollutionof surface waters with phosphorus and nitrogen. EcologicalApplications 8:559–568.

Carpenter, S. R., J. J. Cole, M. L. Pace, M. Van de Bogert,D. L. Bade, D. Bastviken, C. M. Gille, J. R. Hodgson, J. F.Kitchell, and E. S. Kritzberg. 2005. Ecosystem subsidies:terrestrial support of aquatic food webs from 13C addition tocontrasting lakes. Ecology 86:2737–2750.

Carpenter, S. R., D. Ludwig, and W. A. Brock. 1999.Management of eutrophication for lakes subject to poten-tially irreversible change. Ecological Applications 9:751–771.

Chen, B. R., and D. H. Wise. 1999. Bottom-up limitation ofpredaceous arthropods in a detritus-based terrestrial foodweb. Ecology 80:761–772.

Cloern, J. E., C. Grenz, and L. Vidergar-Lucas. 1995. Anempirical model of the phytoplankton chlorophyll : carbonratio—the conversion factor between productivity andgrowth rate. Limnology and Oceanography 40:1313–1321.

Cole, J. J., S. R. Carpenter, J. F. Kitchell, and M. L. Pace. 2002.Pathways of organic carbon utilization in small lakes: resultsfrom a whole-lake 13C addition and coupled model.Limnology and Oceanography 47:1664–1675.

Cole, J. J., S. R. Carpenter, M. L. Pace, M. C. Van de Bogert,J. L. Kitchell, and J. R. Hodgson. 2006. Differential supportof lake food webs by three types of terrestrial organic carbon.Ecology Letters 9:558–568.

Cross, W. F., J. B. Wallace, and A. D. Rosemond. 2007.Nutrient enrichment reduces constraints on material flows ina detritus-based food web. Ecology 88:2563–2575.

Cross, W. F., J. B. Wallace, A. D. Rosemond, and S. L. Eggert.2006. Whole-system nutrient enrichment increases secondaryproduction in a detritus-based ecosystem. Ecology 87:1556–1565.

DiCenzo, V. J., M. J. Maceina, and M. R. Stimpert. 1996.Relations between reservoir trophic state and gizzard shadpopulation characteristics in Alabama reservoirs. NorthAmerican Journal of Fish Management 16:888–895.

Doohan, M. 1973. An energy budget for adult Brachionusplicatilis Muller (Rotatoria). Oecologia 13:351–362.

Dumont, H. J., I. Van de Velde, and S. Dumont. 1975. The dryweight estimate of biomass in a selection of cladocera,copepoda and rotifera from the plankton, periphyton andbenthos of continental waters. Oecologia 19:75–97.

Elser, J. J., W. F. Fagan, R. F. Denno, D. R. Dobberfuhl, A.Folarin, A. Huberty, S. Interlandi, S. S. Kilham, E.McCauley, K. L. Schulz, E. H. Siemann, and R. W. Sterner.2000. Nutritional constraints in terrestrial and freshwaterfood webs. Nature 408:578–580.

Fee, E. J. 1990. Computer programs for calculating in situphytoplankton photosynthesis. Canadian Technical Reportof Fisheries and Aquatic Science number 1740.

Frost, P. C., M. A. Evans-White, Z. V. Finkel, T. C. Jensen,and V. Matzek. 2005. Are you what you eat? Physiologicalconstraints on organismal stoichiometry in an elementallyimbalanced world. Oikos 109:18–28.

Garvey, J. E., N. A. Dingledine, N. S. Donovan, and R. A.Stein. 1998. Exploring spatial and temporal variation withinreservoir food webs: predictions for fish assemblages.Ecological Applications 8:104–120.

Greenberg, A. E., L. S. Clesceri, and A. D. Eaton. 1992.Standard methods for the examination of water and wastewater. 18th edition. American Public Health Association,Washington, D.C., USA.

Grey, J., R. I. Jones, and D. Sleep. 2001. Seasonal changes inthe importance of the source of organic matter to the diet ofzooplankton in Loch Ness, as indicated by stable isotopeanalysis. Limnology and Oceanography 46:505–513.

Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960.Community structure, population control and competition.American Naturalist 94:421–425.

Hall, R. O., J. B. Wallace, and S. L. Eggert. 2000. Organicmatter flow in stream food webs with reduced detritalresource base. Ecology 81:3445–3463.

Higgins, K. A., M. J. Vanni, and M. J. Gonzalez. 2006.Detritivory and the stoichiometry of nutrient cycling by adominant fish species in lakes of varying productivity. Oikos114:419–430.

Johnson, L. B., C. Richards, G. E. Host, and J. W. Arthur.1997. Landscape influences on water chemistry in Midwest-ern stream ecosystems. Freshwater Biology 37:193–208.

Kirk, K. L. 1992. Effects of suspended clay on Daphnia bodygrowth and fitness. Freshwater Biology 28:103–109.

Knoll, L. B., M. J. Vanni, and W. H. Renwick. 2003.Phytoplankton primary production and photosyntheticparameters in reservoirs along a gradient of watershed landuse. Limnology and Oceanography 48:608–617.

Kritzberg, E. S., J. J. Cole, M. L. Pace, W. Graneli, and D. L.Bade. 2004. Autochthonous versus allochthonous carbonsources: results from whole-lake 13C addition experiments.Limnology Oceanography 49:588–596.

Lenat, D. R. 1984. Agriculture and stream water quality: abiological evaluation of erosion control practices. Environ-mental Management 8:333–344.

Mather, M. A., M. J. Vanni, T. E. Wissing, S. A. Davis, andM. H. Schaus. 1995. Regeneration of nitrogen and phospho-rus by bluegill and gizzard shad: effect of feeding history.Canadian Journal of Fisheries and Aquatic Sciences 52:2327–2338.

McCabe, G. D., and W. J. O’Brien. 1983. The effects ofsuspended silt on feeding and reproduction of Daphnia pulex.American Midland Naturalist 110:324–337.

McCauley, E. 1984. The estimation of the abundance andbiomass of zooplankton in samples. Pages 228–265 in J. A.Downing and F. H. Rigler, editors. A manual on methods forthe assessment of secondary productivity in fresh waters.Blackwell, Hoboken, New Jersey, USA.

Michaletz, P. H. 1997. Factors affecting abundance, growth,and survival of age-0 gizzard shad. Transactions of theAmerican Fisheries Society 126:84–100.

Miner, J. G., and R. A. Stein. 1993. Interactive influence ofturbidity and light on larval bluegill (Lepomis macrochirus)foraging. Canadian Journal of Fisheries and AquaticSciences 50:781–788.

Mundahl, N. D., and T. E. Wissing. 1987. Nutritionalimportance of detritivory in the growth and condition ofgizzard shad in an Ohio reservoir. Environmental Biology ofFishes 20:129–142.

Mundahl, N. D., and T. E. Wissing. 1988. Selection anddigestive efficiencies of gizzard shad feeding on naturaldetritus and two laboratory diets. Transactions of theAmerican Fisheries Society 117:480–487.

Nowlin, W. H., J. L. Evarts, and M. J. Vanni. 2005. Releaserates and potential fates of nitrogen and phosphorus fromsediments in a eutrophic reservoir. Freshwater Biology 50:301–322.

Pace, M. L., S. R. Carpenter, J. J. Cole, J.J. Coloso, J. F.Kitchell, J. R. Hodgson, J.J. Middelburg, N. D. Preston,C. T. Solomon, and B. C. Weidel. 2007. Does terrestrialorganic carbon subsidize the planktonic food web in a clear-water lake? Limnology and Oceanography 52:2177–2189.

Pace, M. L., J. J. Cole, S. R. Carpenter, J. F. Kitchell, J. R.Hodgson, M. C. Van der Bogert, D. L. Bade, E. S. Kritzberg,and D. Bastviken. 2004. Whole-lake carbon-13 additionsreveal terrestrial support of aquatic food webs. Nature 427:240–243.

Peterson, D. 2005. U.S. tillage trends. Land and Water(University of Illinois Extension) 6:1–4.

June 2009 959SUBSIDY EFFECTS ON FISH AND PLANKTON

Pierce, R. J., T. E. Wissing, J. G. Jaworski, R. N. Givens, andB. A. Megrey. 1980. Energy storage and utilization patternsof gizzard shad in Acton Lake, Ohio. Transactions of theAmerican Fisheries Society 109:611–616.

Pierce, R. J., T. E. Wissing, and B. A. Megrey. 1981. Aspects ofthe feeding ecology of gizzard shad in Acton Lake, Ohio.Transactions of the American Society of Fisheries 110:391–395.

Pilati, A. 2007. Stoichiometry and the relative importance ofautochthonous and allochthonous food sources for adominant detritivorous fish. Dissertation. Miami University,Oxford, Ohio, USA.

Polis, G. A., and S. D. Hurd. 1996. Lining marine andterrestrial food webs: allochthonous input from the oceansupports high secondary productivity on small islands andcoastal land communities. American Naturalist 147:396–423.

Renwick, W. H., K. J. Carlson, and J. K. Hayes-Bohanan.2005. Trends in recent reservoir sedimentation rates inSouthwestern Ohio. Journal of Soil and Water Conservation60:72–79.

Renwick, W. H., M. J. Vanni, Q. Zhang, and J. Patton. 2008.Water quality trends and changing agricultural practices in aMidwest US watershed, 1994–2006. Journal of Environmen-tal Quality 37:1862–1874.

Richards, R. P., and D. B. Baker. 2002. Trends in agriculture inthe LEASEQ watersheds, 1975–1995. Journal of Environ-mental Quality 31:17–2.

Schaus, M. H., and M. J. Vanni. 2000. Effects of gizzard shadon phytoplankton and nutrient dynamics: role of sedimentfeeding and fish size. Ecology 81:1701–1719.

Schaus, M. H., M. J. Vanni, and T. E. Wissing. 2002. Biomass-dependent diet shifts in omnivorous gizzard shad: implica-tions for growth, food web and ecosystem effects. Transac-tions of the American Fisheries Society 131:40–54.

Schaus, M. H., M. J. Vanni, T. E. Wissing, M. T. Bremigan,J. E. Garvey, and R. A. Stein. 1997. Nitrogen and phosphorusexcretion by detritivorous gizzard shad in a reservoirecosystem. Limnology and Oceanography 42:1386–1397.

Smith, V. H., D. Tilman, and J. C. Nekola. 1999. Eutrophication:impacts of excess nutrient inputs on freshwater, marine, andterrestrial ecosystems. Environmental Pollution 100:179–196.

Smoot, J. C. 1999. A field study of sedimentary microbiota asfood for detritivorous gizzard shad,Dorosoma cepedianum, inActon Lake: a biomarker approach. Dissertation. MiamiUniversity, Oxford, Ohio, USA.

Soballe, D. M., and S. T. Threlkeld. 1988. Algal–clayflocculation in turbid waters: variations due to algal andmineral differences. Verhandlungen der InternationalenVereinigung fur Theoretische und Angewandte Limnologie23:750–754.

Solomon, C. T., S. R. Carpenter, J. J. Cole, and M. L. Pace.2008. Support of benthic invertebrates by detrital resourcesand current autochthonous production: results from a whole-lake C-13 addition. Freshwater Biology 53:42–54.

Solorzano, L. 1969. Determination of ammonia in naturalwaters by the phenolhypoclorite method. Limnology andOceanography 14:799–801.

Stainton, M. P., M. J. Capel, and F. A. J. Armstrong. 1977. Thechemical analysis of freshwaters. Freshwater Institute,Winnipeg, Manitoba, Canada.

Stein, R. A., D. R. DeVries, and J. M. Dettmers. 1995. Food-web regulation by a planktivore: exploring the generality ofthe trophic cascade hypothesis. Canadian Journal ofFisheries and Aquatic Sciences 52:2518–2526.

Stelzer, R. S., J. Heffernan, and G. E. Likens. 2003. Theinfluence of dissolved nutrients and particulate organicmatter quality on microbial respiration and biomass in aforest stream. Freshwater Biology 48:1925–1937.

Sterner, R. W., and J. J. Elser. 2002. Ecological stoichiometry:the biology of elements from molecules to the biosphere.Princeton University Press, Princeton, New Jersey, USA.

Uri, N. D. 2001. A note on soil erosion and its environmentalconsequences in the United States. Water Air and SoilPollution 129:181–197.

USDA. 1992. Watershed plan and environmental assessmentfor Four Mile Creek Watershed, Ohio and Indiana. SoilConservation Service, Washington, D.C., USA.

Vanni, M. J., L. S. Andrews, W. H. Renwick, M. J. Gonzalez,and S. J. Noble. 2006. Nutrient and light limitation ofreservoir phytoplankton in relation to storm-mediated pulsesin stream discharge. Archives of Hydrobiology 167:421–445.

Vanni, M. J., K. K. Arend, M. T. Bremigan, D. B. Bunnell,J. E. Garvey, M. J. Gonzalez, W. H. Renwick, P. A.Soranno, and R. A. Stein. 2005. Linking landscapes and foodwebs: effects of omnivorous fish and watersheds on reservoirecosystems. BioScience 55:155–167.

Vanni, M. J., and J. L. Headworth. 2004. Cross-habitattransport of nutrients by omnivorous fish along a produc-tivity gradient: integrating watersheds and reservoir foodwebs. Pages 43–61 in G. A. Polis, M. E. Power, and G. L.Huxel, editors. Food webs at the landscape level. Universityof Chicago Press, Chicago, Illinois, USA.

Vanni, M. J., W. H. Renwinck, J. L. Headworth, J. D. Auch,and M. H. Schaus. 2001. Dissolved and particulate nutrientflux from three adjacent agricultural watersheds: a five-yearstudy. Biogeochemistry 54:85–114.

Vinyard, G. L., and W. J. O’Brien. 1976. Effects of light andturbidity on the reactive distance of bluegill (Lepomismacrochirus). Journal of the Fisheries Research Board ofCanada 33:2845–2849.

Wallace, J. B., S. L. Eggert, J. L. Meyer, and J. R. Webster.1997. Multiple trophic levels of a forest stream linked toterrestrial litter inputs. Science 277:102–104.

Wallace, J. B., S. L. Eggert, J. L. Meyer, and J. R. Webster.1999. Effect of resource limitation on a detrital-basedecosystem. Ecological Monographs 69:409–442.

Waters, T. F. 1995. Sediment in streams: sources, biologicaleffects, and control. American Fisheries Society Monograph 7.

Williams, J. E. 2000. The coefficient of condition of fish.Chapter 13 in J. C. Schneider, editor. Manual of fisheriessurvey methods II: with periodic updates. Fisheries specialreport 25. Michigan Department of Natural Resources, AnnArbor, Michigan, USA.

Yang, L. H., J. L. Bastow, K. O. Spence, and A. N. Wright.2008. What can we learn from resource pulses? Ecology 89:621–634.

APPENDIX

Results of repeated-measures ANOVA on sediment and nutrient effects on limnological parameters (Ecological Archives A019-039-A1).

ALBERTO PILATI ET AL.960 Ecological ApplicationsVol. 19, No. 4