Ecological studies of apple snails

Special Publication SJ98-SP6

ECOLOGICAL STUDIES OF APPLE SNAILS

(Pomacea paludosa, SAY)

Philip C. Darby, Patricia L. Valentine-Darby, Robert E. Bennetts,
Jason D. Croop, H. Franklin Percival and Wiley M. Kitchens

Florida Cooperative Fish & Wildlife Research Unit
P.O. Box 110450
University of Florida
Gainesville, FL 32611-0450

December 1997

Prepared under joint contract with:

South Florida Water Management District
(Contract No. C-E6609)

St. Johns River Water Management District
(Contract No. 95D 159)

EXECUTIVE SUMMARY

The Florida apple snail (Pomaceapaludosa, Say) is a critical food web component in

Florida wetlands. Water resource managers and the U.S. Fish & Wildlife Service have

identified apple snail research as a high priority in central and south Florida wetland

restoration efforts. The main impetus for apple snail ecological research is derived from

interests in managing wetland and lake water levels to support Florida's population of

endangered snail kites (Rostramus sociabilis). It is generally accepted that increased

frequency and duration of dry downs beyond natural levels negatively influences snail kite

populations. It is assumed that the negative impact manifests itself through depressed apple

snail populations, although no data from controlled studies exist.

Dry downs are a natural process in the evolution and maintenance of the mosaic of

plant communities within central and south Florida wetlands. The natural hydrologic regime

of these wetlands has, however, been altered substantially due to installation of a network of

canals and flood control structures. The overall goal of this report is to provide information

critical to understanding the ramifications of water management practices on apple snail

populations, and subsequently snail kites, as water resource managers endeavor to achieve

the delicate balance of inundation and dry down in which Florida wetland communities

evolved. The specific objectives of our research on apple snail ecology were to: 1) develop

a reliable sampling technique for estimating snail density and/or relative abundance; 2)

compare density and/or relative abundance of apple snails between a variety of common

habitat types (sawgrass, cattail, wet prairie, and slough); 3) determine the behavioral

responses (migration or aestivation) of apple snails to drying conditions; and 4) estimate

survival of apple snails during a drying event and evaluate the influence of hydrologic

parameters (rate, extent, and duration of dry downs) on survival.

Many basic questions about the ecology or population dynamics of any organism

require some measure of abundance, or at least relative abundance. Over the past two

decades, several direct and indirect measures of apple snail abundance have been proposed,

but none of these methods have been sufficiently evaluated to draw conclusions about their

utility and reliability. In addressing objective 1, we compared three different methods (bar

seine, dip net and suction dredge) for extracting snails from 1-m2 throw traps. We also

investigated the utility of two different trap systems, crayfish traps and trap arrays, to

determine relative abundance. We used crayfish traps to explore the use of a mark-recapture

technique to determine snail density. Finally, we examined the use of apple snail egg cluster

counts as an index to population density.

We found that regardless of the method employed, obtaining reliable estimates of

apple snail density will be time and labor intensive. The egg cluster index, highly desirable

due to its ease and simplicity, was found to be of little value given its high degree of

temporal and spatial variation. The bar seine was eliminated early in the throw trap

investigation due to its low efficiency relative to other extraction methods. The suction

dredge appeared somewhat less sensitive to habitat differences and tended to have slightly

higher overall recovery probabilities than the dip net. However, the dip net required less

effort and may require less initial investment. We encourage use of either the dip net or

suction dredge if a throw trap method is to be employed. We found that it is imperative to

assess efficiency of extraction when using throw traps, especially when sampling across

habitat types. Without information on extraction efficiencies, investigators risk

misinterpreting site to site recovery variability as a real difference in snail density.

The use of crayfish traps or trap arrays as an index of relative abundance may be

appropriate in some situations, but great care must be taken to control for time effects and

site to site variation that might affect capture probabilities. If capture probability is not

directly measured (e.g. a mark-recapture regime), then potential differences in capture

probabilities may be inaccurately interpreted as a difference in snail abundance (as noted

with throw trap recoveries). We believe that the use of crayfish traps within a mark-

recapture sampling regime has the greatest potential to provide reliable estimates of apple

snail densities. Mark-recapture data, obtained and analyzed as described in this report,

provides not only population density information, but also information on survival,

movements, and behavior of snails in the population. All of these parameters effect

sampling protocols.

While testing throw trap and mark-recapture trapping techniques, we were able to

draw some conclusions about the distribution of apple snails in graminoid marshes

(Objective 2). Apple snails were found in all habitat types (sawgrass, prairie, slough, and

cattail) encountered during our research. No consistent pattern of distribution among habitat

types was observed, although it appears that higher densities of snails are more likely to be

found in prairie or cattail habitats in some areas. Our egg cluster data, snail density data, and

telemetry monitoring data revealed that snails utilize densely vegetated areas (e.g., interior

sawgrass and cattail habitats). This contradicts earlier reports that concluded that snails have

difficulty penetrating dense stands of vegetation, and that habitat use is clearly skewed to less

densely vegetated habitats. Our results do not dispute the importance of the prairie/sawgrass

or slough/sawgrass ecotones as being critical for oviposition. We would simply add that the

interior of sawgrass and cattail plant communities be recognized as important apple snail

habitat. We agree with earlier reports that most favorable snail habitat would likely include

a mosaic of densely vegetated and sparsely vegetated habitats within a wetland system.

Based on our experience with site to site variability in the habitats and distributions of snails,

we anticipate that a considerably greater effort will be required to make generalizations

about snail distribution in different habitat types. Understanding snail use of different

habitat types remains an important issue related to natural resource management practices

which affect the plant community (i.e., hydroperiod, fire, aquatic weed control, nutrient

loading), and most certainly in turn, apple snail populations.

Telemetry studies revealed that apple snails do not seek out deep water refuge during

a dry down (Objective 3). Telemetry and crayfish trapping data indicate that apple snail

reproductive ecology drives the movement patterns of snails more so than does hydrology.

Water depth does, however, influence snail movements. An approximate depth of 10 cm

appears to be a threshold level at which snail movements become impeded. At this point

snails settle in one spot and, as residual water recedes, they become subjected to dry down

conditions. They do not burrow, but they do conserve moisture through tight closure of their

operculum.

Finally, we investigated snail survival during the dry season with one field and one

laboratory study (Objective 4). Desiccation during the dry season is not necessarily a

predominate cause of mortality. The 1995 telemetry data, coupled with information gained

through egg cluster counts and crayfish trap surveys in 1996, indicates that snails die within a

few weeks after reproducing. In order to more closely examine the relationship between

hydrology and adult snail survival, we conducted a laboratory study with controls and

experimental tanks from which water was withdrawn to simulate a dry down. Laboratory

studies confirmed that regardless of hydrologic conditions, post-reproductive adult size snails

reach the end of their life span (estimated at 12 to 18 months) and die. Snails which did

survive dry down conditions for 8 weeks (laboratory snails) and 12 weeks (observed in the

field), tended to be juvenile size snails. We hypothesize that snail tolerance to desiccation is

a function of snail size and/or reproductive status.

Further understanding of the relationship between dry down tolerance and snail

physiological status will provide critically needed information about the impact of the timing

and duration of dry downs under the control of water management districts. However, we

already have information on the potential impacts of hydrologic regime on another critical

factor regulating snail populations: recruitment. Our egg cluster surveys in the Blue

Cypress Water Management Area in 1996, as well as earlier surveys done by other

researchers on Lake Okeechobee and in Silver springs, reveal that peak egg cluster

production by apple snails in central and south Florida consistently occurs between March

and July; the majority of eggs are laid over a 4 to 12 week period. Dry downs which

encompass the time period of peak reproduction may reduce or eliminate recruitment in the

effected area. We do not suggest that dry downs be avoided, only that water management

regimes consider their timing. Dry downs occurring later in the reproductive season (i.e.,

after peak reproduction) likely pose significantly less harm to snail populations. Our data

indicate that young of the year snails can survive at least 2 to 3 months in dry down

conditions. We believe rapid early growth enables snails hatched in March and April

(typically a substantial portion of the total hatch) to reach sufficient size to survive a mid- to

late spring dry down. The critical issue for snail populations may not be whether or not dry

downs occur, but rather when they occur.

Our research has revealed the importance of understanding how snail population

dynamics can affect interpretation of study results, and how the timing of hydrologic

changes affects survival and recruitment in snail populations. Balancing the distribution of

water for human use and for maintaining sustainable snail populations may narrow down to a

few critical weeks or months late in the dry season.

ACKNOWLEDGMENTS

The completion of this project and associated report would not have been
possible without contributions from several cooperating agencies and many people. We
greatly appreciate the financial support of the South Florida Water Management District
(SFWMD) and St. Johns River Water Management District (SJRWMD). The Game and
Fresh Water Fish Commission (GFC), Bureau of Nongame Wildlife, funded a project to
assess the impacts of the Lake Kissimmee draw down on apple snails, and this funding
supported graduate students and technicians for a portion of the work which is reported
in this document.
Steve Miller (SJRWMD), Mary Ann Lee (SJRWMD), Ed Lowe (SJRWMD), Pete
David (SFWMD), Paul Warner (SFWMD), and Dale Gawlik (SFWMD) were
instrumental in initiating this project and supporting us throughout its duration. We also
received assistance from David Cook (GFC) and Angela Williams (GFC) in obtaining a
special use permit for the collection of snails used in this study. Patrick Dean (GFC) and
Guy Carpenter (GFC) provided help with disabled trucks and airboats, and on numerous
occasions provided critical information about local resources, services, and field
conditions.
We greatly appreciate the contributions of our field biologists Eric E. Tenoso,
Mauricio Rojas, Betty Jordan, Steve McGehee, Bill Millinor, Mike Betts, Theresa Sitar,
Ron Breeding, Greg Kauffman and Jeff Carter. We also appreciate the assistance from
Traci Dean during the laboratory studies in Kenansville.
Jack Dickson (UF), Jim Nichols (USGS-BRD) and Jim Hines (USGS-BRD)
shared their expertise in statistics and population modeling, and for this we are very
grateful. Frank Jordan (Jacksonville Univ.) provided considerable insight into issues
related to sampling with throw traps. Mike Moulton (UF, WEC), Katie Sieving (UF,
WEC), Clay Montague (UF, EES), Fred Thompson (UF, ZOO) and Don DeAngelis
(USGS-BRD) also provided valuable input regarding study design, data analyses and the
direction of the research.

The laboratory studies on temperature tolerance were performed at the USGS-
BRD facility in North Gainesville. We appreciate the space and resources provided, and
especially thank Nick Funnicelli, Bill Stranghoener, and Anne Keller for their assistance
and support during those studies.
Frank Mazzotti (UF, IFAS Belle Glade Facility) graciously let us borrow his RV-
camper during the mark-recapture studies. We also appreciate Mary Hudson-Kelly's
(UF, IFAS Belle Glade Facility) role in arranging use of the trailer, and Robert
Stubblefield's (UF, IFAS Belle Facility) help in preparing the trailer for use.
We would like to thank Dale Gawlik (SFWMD), Steve Miller (SJRWMD), Ric
Hupalo (SJRWMD), and Mary Ann Lee (SJRWMD) for their helpful comments on drafts
of this report.
We very much appreciate the administrative support provided by Debra Hughes
and Barbara Fesler (Florida Cooperative Fish and Wildlife Research Unit). This study
was conducted under the auspices of the Florida Cooperative Fish and Wildlife Research
Unit (GFC, Univ. of Florida, U. S. Geological Survey- Biological Research Division, and
Wildlife Management Institute, cooperating).

TABLE OF CONTENTS

EXECUTIVE SUMMARY...................................................... .............

ACKNOWLEDGMENTS......................... ................................................... .........v.

1.0 IN TR O D U C TIO N ............................................................................. ...........1

2.0 STUD Y SITES................................... ....................... ............................
2.1 Everglades Water Conservation Areas.............................. ........... 7
2.2 Blue Cypress Water Management Area........................................... 11
2.3 Lake Kissimmee......................................................................12

3.0 METHODS FOR DETERMINING SNAIL ABUNDANCE..........................14
3.1 Throw Trap Extraction Techniques....................................................14
3.2 Movement Based Wire Traps..................................................28
3.3 M ark-Recapture ...................................... ....... ... ...... ... 44
3.4 Egg Cluster Counts.................................................... .....................62
3.5 Conclusions..... ....................................... ............................. ..... 69

4.0 FIELD STUDIES OF MOVEMENTS AND SURVIVAL..............................73
4.1 M ethods................................................ ..................................... ....74
4.2 R esults................................................... ..................................... ...87
4 .3 D iscussion............................................................................ ............ 113

5.0 LABORATORY EXPERIMENTS OF SNAIL SURVIVAL.........................118
5.1 M ethods .................................. ... .................... ................... ...... ......... 119
5.2 R esults.....................................................................................125
5.3 D discussion ...... .. ............. ......... ............. ......... ....... ....130

6.0 SYNTHESIS................................................................................................... 134

LITERATURE CITED .................................................................................... 144

1.0 INTRODUCTION

The Florida apple snail (Pomaceapaludosa, Say) is a critical food web component in

Florida wetlands, one for which further research has been identified as a high priority in

wetland restoration efforts in all south Florida wetlands including the Everglades (Science

SubGroup, South Florida Restoration Task Force, 1996) and the Upper St. Johns River Basin

(USFWS 1986, Turner 1994). The apple snail is the nearly exclusive food of the endangered

snail kite (Rostramus sociabilis) (Snyder and Snyder 1969) and comprises over 75% of the

diet of limpkins (Aramus guarauna) in central and south Florida (Cottam 1936, Synder and

Snyder 1969). Other avian predators include white ibis (Eudocimus albus) (Kushlan 1974)

and boat-tailed grackles (Cassidix mexicanus) (Snyder and Snyder 1969). In addition,

alligators (Alligator missippienisis) (Fogarty and Albury 1968, Delaney and Abercombie

1986), redear sunfish (Lepomis microlophus) (Chable 1947), and soft-shelled turtles (Trionyx

ferox) (Dalrymple 1977) prey on apple snails. Despite their long recognized importance in

Florida wetlands, especially with regard to snail kites (Snyder and Snyder 1969, Perry 1974,

Beissenger 1986, Bennetts and Kitchens 1997), surprisingly little is known regarding the life

history, ecology, and population dynamics ofPomaceapaludosa. Through our efforts

described in this report, we have unraveled some of the mystery associated with snail

ecology, particularly the impact of hydrology on apple snail population dynamics.

Seasonal fluctuations in rainfall occur in Florida, historically providing a hydrologic

regime that supports wetland plant communities that benefit from cyclical drought and

inundation (Gunderson 1994). Populations of P. paludosa, like other members of the

gastropod family Pilidae, must contend with the annual and inter-annual hydrologic

fluctuations which shape the wetland habitats they occupy. The natural hydrologic regime of

wetlands in Florida, however, has been altered substantially due to the installation of water

control structures since the 1930's, particularly in South Florida (Light and Dineen 1994).

Anthropogenic alterations to the natural hydrologic cycle to accommodate agricultural and

urban needs now compromise the success of ecological strategies characteristic of some

wetland species. For example, the untimely and unnaturally large influx of water from

agricultural areas into the Everglades system has increased incidences of alligator nest

flooding (Kushlan 1990). Conversely, diverting water to supply agricultural and urban needs

has shortened hydroperiods in the southern Everglades (i.e., Everglades National Park),

resulting in more frequent and longer dry downs which alter macrophyte community

structure (Davis et al. 1994) and depress fish populations (Loftus and Eklund 1994).

Through their persistence and wide distribution in Florida, it is apparent that apple

snail populations have adapted to the natural hydrologic dynamics of Florida wetland

habitats. Undoubtedly changes in hydrology which deviate from the natural fluctuations will

impact snails, but to what extent is unknown. Anecdotal evidence from observations of

foraging limpkins suggests that snails become stranded during dry downs (Snyder and Snyder

1969). Some studies indicate that stranded snails are intolerant of dry downs as short as two

weeks (Little 1968, Turner 1994). It is also apparent that snail populations respond to

seasonal changes in hydrology, and that snail populations decline during extensive dry

periods (Kushlan 1975). However, sampling efforts to determine snail abundance, and

therefore population trends, have proceeded without knowledge of their efficacy.

The main impetus for apple snail ecological research is derived from interests in

managing wetland and lake water levels to support Florida's population of endangered snail

kites (USFWS 1986). It is generally accepted that increased frequency and duration of dry

downs beyond natural levels negatively influences snail kite populations (Sykes 1983,

Takekawa and Beissenger 1989, Bennetts and Kitchens 1997). It is assumed the negative

impact manifests itself through depressed apple snail populations, although no data from

controlled studies exist. Providing sufficiently long, and appropriately timed, wetland

hydroperiods to support apple snails falls largely under the auspices of the South Florida and

St. Johns River Water Management Districts. Bennetts and Kitchens (1997) point out that

effective water management for kites does not mean excluding dry downs, which are a

natural process in the evolution and maintenance of Florida's mosaic of wetland plant

communities (Davis et al. 1994). The overall goal of this report is to provide information

critical to understanding the ramifications of water management practices on apple snails,

and subsequently snail kites, as the districts endeavor to achieve the delicate balance of

inundation and dry down in which Florida wetland communities evolved.

The main goals of the research described herein were to evaluate apple snail

sampling techniques and to investigate the extent of the impacts of drying events on snail

populations. The specific objectives of our research on apple snail ecology are as follows:

1) Develop a reliable sampling technique for estimating snail density and/or

relative abundance.

2) Compare density and/or relative abundance of apple snails between a variety

of common habitat types (e.g., sawgrass, cattail, wet prairie, and slough).

3) Determine the behavioral responses of apple snails (e.g., migrate or aestivate)

to drying conditions.

4) Estimate survival of apple snails during a drying event and evaluate the

influence of hydrologic parameters (e.g., rate, extent, and duration of dry

down) on survival.

The research was conducted in two phases in the period from March 1995- June 1997, in

four different wetland systems (Water Conservation Areas 3A and 2B, Blue Cypress Water

Management Area, and the Lake Kissimmee littoral zone) and in a laboratory setting. Data

from both years and from more than one system were compiled to address each objective

stated above. However, we report the results to reflect the objectives, not necessarily the

chronological order in which the data were obtained. The following synopsis clarifies when

and where the data were collected. The study sites are described in Chapter 2.

Phase I

Water Conservation Area 2B. A pilot study comparison of throw trap extraction

techniques was conducted. A portion of the egg cluster index data was also

collected during this effort. Sampling was conducted from May 1995 -

February 1996.

Blue Cypress Water Management Area. The first field study of snail movements

and survival over the course of a marsh dry down was performed. Monitoring

occurred from March August 1995.

Lake Kissimmee. A study of snail movements and survival during an extreme

draw down of the lake was conducted. Monitoring occurred from October

1995 February 1996.

Phase II

Water Conservation Area 3A. A comparison of dip net and suction dredge

extraction of throw traps was conducted in three sites in southwestern

WCA3A. This data was collected from May 1996 through August 1996. The

mark-recapture method was also tested in WCA3A, but in the eastern section.

The mark-recapture data was collected from March 1997 through May 1997.

During both the throw trap and mark-recapture data sampling periods, we also

collected data on egg clusters as an abundance index, and compared snail

densities in different vegetation types (sawgrass, cattail, and wet prairie).

Blue Cypress Water Management Area. A follow up study on snail movements

was conducted to assess the relationship between snail breeding season and

movements. This work was conducted from February 1996 through August

1996.

Laboratory Studies. A laboratory study was performed to compare the survival

of snails exposed to dry down conditions to those under control (inundated)

conditions. The study was started in May 1996 and completed in August

1996. We also performed temperature tolerance studies in October -

November 1996.

In the process of achieving the project objectives, we also collected data on mating behavior,

temporal and spatial variation in snail egg production, growth and senescence, and

temperature tolerances. In addition we have developed novel techniques for conducting

radio-telemetry studies and for collecting large numbers of apple snails for scientific study.

2.0 STUDY SITES

Our two year study on apple snail ecology included the three largest freshwater marsh

systems in Florida: The Everglades, The Upper St. Johns Marsh, and the Kissimmee Marsh

(Kushlan 1990). The hydrology of each of these areas has been altered by construction of

levees and canals and installation of water control structures (Lowe 1983, Light and Dineen

1994, GFC 1995). The natural intra-annual and inter-annual variation in rainfall results in

seasonal dry downs (typically in late Spring) which contribute to the maintenance of

community structure of these and all central and south Florida wetlands (Kushlan 1990,

Davis et al. 1994). Understanding water management impacts on apple snails in these

particular wetland systems has been identified as important to recovery of the Florida snail

kite (Bennetts and Kitchens 1997).

Apple snails inhabit many types of freshwater habitats in Florida, including forested

swamps, rivers and streams, lakes, and agricultural canals and ponds (pers. obs.). Our study

results and conclusions apply only to the graminoid wetlands and littoral zone habitats in

which the research was performed. Important distinctions between graminoid wetlands and

other aquatic systems, especially in terms of hydrology, may preclude extending our

conclusions about snail ecology to non-graminoid aquatic habitat types.

2.1 Everglades Water Conservation Areas

The Everglades Ecosystem is part of a watershed that includes the Lake Kissimmee

Chain of Lakes, Lake Okeechobee, the Everglades Water Conservation Areas, and

2.0 STUDY SITES

Our two year study on apple snail ecology included the three largest freshwater marsh

systems in Florida: The Everglades, The Upper St. Johns Marsh, and the Kissimmee Marsh

(Kushlan 1990). The hydrology of each of these areas has been altered by construction of

levees and canals and installation of water control structures (Lowe 1983, Light and Dineen

1994, GFC 1995). The natural intra-annual and inter-annual variation in rainfall results in

seasonal dry downs (typically in late Spring) which contribute to the maintenance of

community structure of these and all central and south Florida wetlands (Kushlan 1990,

Davis et al. 1994). Understanding water management impacts on apple snails in these

particular wetland systems has been identified as important to recovery of the Florida snail

kite (Bennetts and Kitchens 1997).

Apple snails inhabit many types of freshwater habitats in Florida, including forested

swamps, rivers and streams, lakes, and agricultural canals and ponds (pers. obs.). Our study

results and conclusions apply only to the graminoid wetlands and littoral zone habitats in

which the research was performed. Important distinctions between graminoid wetlands and

other aquatic systems, especially in terms of hydrology, may preclude extending our

conclusions about snail ecology to non-graminoid aquatic habitat types.

2.1 Everglades Water Conservation Areas

The Everglades Ecosystem is part of a watershed that includes the Lake Kissimmee

Chain of Lakes, Lake Okeechobee, the Everglades Water Conservation Areas, and

Everglades National Park. A history of the watershed and its management can be found in

Light and Dineen (1994). Most of our work on apple snail survey techniques occurred in the

Water Conservation Areas.

Water Conservation Area 3A (WCA3A)

WCA3A is a 237,000 ha wetland impoundment in Dade and Broward Counties

(Figure 1). Historically, the area now confined within WCA3A was continuous with the

Everglades system topographic gradient, and it conveyed the southerly flow of water into

what is now the Everglades National Park. Water control structures have since diverted that

flow toward the east, and now the eastern section of 3A, especially adjacent to Levee 67A,

has deeper water and longer hydroperiods relative to western 3A. Because we sampled two

distinct areas in WCA3A, we describe them separately.

South Western WCA3A

The plant community in south western WCA3A is a mosaic of sawgrass (Cladium

jamiacense) marsh interspersed with wet prairie and slough, and dotted by tree islands of bay

(Perseapalustris), pond apple (Annona glabra) and/or willow (Salix caroliniana). Wet

prairie habitats are characterized by an abundance of periphyton and the presence of

emergent macrophytes such as spike rush (Eleocharis cellulosa), maidencane (Panicum

hemitomon), beak rushes (Rhynchospora spp.), and arrowhead (Sagittaria spp.) Typical

slough habitats support floating-leaved plants such as water lily (Nymphaea odorata) and

submerged plants such as bladderwort (Utricularia spp.). We found gradations between wet

Figure 1. Map of south Florida showing the location of field study areas as described in Chapter 2 and referred
to throughout the report (L. Kiss = Lake Kissimmee, BCWMA = Blue Cypress Water Management Area,
WCA3A = Water Conservation Area 3A). Specific sampling sites noted by O. Cities (*) and Lake
Okeechobee (L. Okee) included for reference.

Rt. 512
Orlando*

L. KISS aB P
.BCWMA BCWMA West

": LGLra

SMiami

prairie and slough difficult to distinguish within sites sampled, so we refer to these habitats

collectively as prairie/slough. The marsh substrate consists of fibrous peat, but we did

encounter some limestone rock in localized areas within sampling sites. Water depths at

sampling sites ranged from 60 to 45 cm during our May through July 1996 sampling period.

Eastern WCA3A

We determined snail density in three distinct habitat types in eastern WCA3A; two

sites each of cattail (Typha sp.), prairie/slough, and sawgrass. All sampling locations were

within 2.0 km of the L-67A canal. The cattail sites were dominated by 2.5 to 3-m high

cattail, with scattered stems of sawgrass and small pockets of water lily. Emergent

vegetation in the prairie/slough habitat was dominated by spike rush. Sawgrass,

approximately 2.0 to 2.5-m high, dominated the third habitat type. In general, sawgrass sites

contained more open pockets of water lily and pickerelweed (Pontederia cordata) than the

cattail sites. Water depths ranged from 89 to 72 cm in slough, 93 to 63 cm in cattail and 91

to 68 cm in sawgrass over the March May 1997 sampling period. The underlying substrate

in all sites was fibrous peat. Cattail sites maintained a flocculent organic debris layer over

the peat; this flocculent debris layer was not observed in the prairie/slough or sawgrass sites.

The slough sites supported a thick layer of periphyton. Periphyton was limited to small open

patches in sawgrass sites, but was not discernible in cattail sites.

Water Conservation Area 2B (WCA2B)

WCA2B is an 11,300 ha impoundment in Broward County (Figure 1). The plant

community is similar to that described for southwestern WCA3A, except the introduced

Melaleuca tree (Melaleuca quinquenervia) is much more common. The substrate here is

also a fibrous peat, but unlike WCA3A no limestone rock was encountered. Water levels in

1995 ranged from approximately 20 to 100 cm, but note that these measurements were

collected over 9 months. We moved to WCA3A in 1996 to complete the sampling study due

to low water levels in WCA2B.

2.2 Blue Cypress Water Management Area

The Blue Cypress Water Management Area, a part of the Upper St. Johns River

Basin, located in Indian River County, FL, is a large graminoid marsh system which, like the

Everglades, is undergoing a substantial restoration effort. Interest in the apple snail

population of the BCWMA grew substantially when the endangered snail kite returned to this

area to nest and roost after years of infrequent kite sightings (Turner 1996). Our research on

snails included two distinct areas within the Upper St. Johns Basin, BCWMA East and

BCWMA West (Figure 1).

BCWMA East

The majority of our field investigations of snail movements and survival were

conducted in the eastern-most quarter of BCWMA East. Work was conducted there in

March through July 1995 and February through August 1996. The pilot study for the mark-

recapture method was done in BCWMA East in 1996. Trap arrays and crayfish traps were

also deployed in this area in 1996 to collect snails for laboratory studies.

The BCWMA East plant community consists of patches of sawgrass surrounded by

mixed emergents (Panicum spp., Eleocharis spp., Sagittaria spp., Pontederia cordata, and

in shallower areas, Xyris sp.). Periphyton forms thick layers over the sand substrate which

characterizes BCWMA East. In 1995, portions of our study area dried to below ground level;

no dry down occurred in this area in 1996. During our 1995 study of dry season survival,

water depths throughout the study site (excluding canals) ranged from 15 to 70 cm in March

- April, and from 0 to 50 cm in May July. In 1996, water depths throughout the study site

ranged from 35 to 90 cm in February April and from 20 to 70 cm in May-August

BCWMA West

During the 1995 study on survival and movements, we also monitored snails in

BCWMA West. The hydrology of this section of the Upper St. Johns marsh is distinct from

that of the BCWMA East, reflected in part by the dominant vegetation and substrate.

Sawgrass, water lily and cypress (Taxodium spp.) are the dominate vegetation. In contrast to

the sand substrate of BCWMA East, the substrate of BCWMA West is a fibrous peat. During

April June 1995, water depths ranged from 40 to 80 cm.

2.3 Lake Kissimmee

Lake Kissimmee is an approximately 14,000 ha lake in Osceola County (Figure 1).

The extensive littoral zone supports numerous species of macrophytes including maidencane,

pickerelweed and cattail. Long term stabilization of water levels has resulted in an

accumulation of muck in many localized areas of the lake. Muck sites (n=3) supported

dense growths of spatterdock (Nuphar luteum). Three sites on the lake had a predominately

sand substrate with a thin (< 15 cm), patchily distributed, flocculent organic layer, while the

remaining two sites had a clean sand substrate. The 1995 trap array validation was

conducted in a range of water depths (45 to 100 cm) depending upon how close we could get

to shore. Lake levels were stable during the trap validation study. During the 1995-1996

draw down, the snail movement study site initially had water depths of 50 to 200 cm

(depending on distance from shore); by February 1996 most of the site had no standing

water.

3.0 METHODS FOR DETERMINING SNAIL ABUNDANCE

Many basic questions about the ecology or population dynamics of any organism

require some measure of abundance, or at least relative abundance. Indeed, one of the

reasons for the paucity of information on snail ecology is a lack of reliable techniques for

sampling apple snail populations. Over the past two decades, several direct and indirect

measures of snail abundance have been proposed (Owre and Rich 1987, Bennetts et al.

1988), but none of these methods have been sufficiently evaluated to draw conclusions about

their utility and reliability.

In this chapter, we compare several direct and indirect measures of snail abundance.

We compare three different methods for extracting snails from 1-m2 throw trap sampling

(3.1). We also investigate the utility of two different trap systems to determine relative

abundance (3.2), and we further explore one of these trap systems for estimating snail density

using a mark-recapture approach (3.3). Finally, we examine the use of apple snail egg

cluster counts as an index to population density (3.4). In each section we end with a

discussion of the results for that particular method. The chapter concludes with an overall

discussion of the relative utility of all the methods investigated.

3.1 Throw Trap Extraction Techniques to Determine Snail Density

Throw traps which encompass a small area (approximately 1 m2 to 2 m2) have been

used for sampling fish and macro-invertebrates in the Everglades system (Kushlan 1981,

Owre and Rich 1987, Chick et al. 1992, Jordan et al. 1996). A metal sided 1-m2 throw trap

3.0 METHODS FOR DETERMINING SNAIL ABUNDANCE

Many basic questions about the ecology or population dynamics of any organism

require some measure of abundance, or at least relative abundance. Indeed, one of the

reasons for the paucity of information on snail ecology is a lack of reliable techniques for

sampling apple snail populations. Over the past two decades, several direct and indirect

measures of snail abundance have been proposed (Owre and Rich 1987, Bennetts et al.

1988), but none of these methods have been sufficiently evaluated to draw conclusions about

their utility and reliability.

In this chapter, we compare several direct and indirect measures of snail abundance.

We compare three different methods for extracting snails from 1-m2 throw trap sampling

(3.1). We also investigate the utility of two different trap systems to determine relative

abundance (3.2), and we further explore one of these trap systems for estimating snail density

using a mark-recapture approach (3.3). Finally, we examine the use of apple snail egg

cluster counts as an index to population density (3.4). In each section we end with a

discussion of the results for that particular method. The chapter concludes with an overall

discussion of the relative utility of all the methods investigated.

3.1 Throw Trap Extraction Techniques to Determine Snail Density

Throw traps which encompass a small area (approximately 1 m2 to 2 m2) have been

used for sampling fish and macro-invertebrates in the Everglades system (Kushlan 1981,

Owre and Rich 1987, Chick et al. 1992, Jordan et al. 1996). A metal sided 1-m2 throw trap

has been found most effective for sampling fish and invertebrates in vegetated habitats

(Chick et al. 1992, Jordan et al. 1996). Extraction techniques for sampling fish and

invertebrates from throw traps include a dip net (Kushlan 1981, Jacobsen and Kushlan 1987,

Chick et al. 1992), bar seine (Rozas and Odum 1988, Chick et al. 1992), and suction dredge

(Brook 1979, Owre and Rich 1987, Bennetts et al. 1988). Although each of these methods

has also been used to sample apple snails (Owre and Rich 1987, Bennetts et al. 1988, F.

Jordan, pers. comm.), no quantitative comparison has been made among these extraction

techniques.

Methods

During 1995 in WCA2B, we conducted a pilot investigation of throw trap sampling to

(1) refine our sampling protocols in preparation for a second season, (2) compare the

proportion of marked animals that were recovered from throw traps using three extraction

techniques, and (3) evaluate the effort required for each extraction method.

Sampling was accomplished using a throw trap, which quickly encloses a 1-m2 area

after being thrown into the marsh (Kushlan 1981, Chick et al. 1992). The throw trap was a 60

cm high by 100 cm x 100 cm box constructed of welded aluminum pipe enclosed with

aluminum sheeting that lacked a top and bottom (Chick et al. 1992). A 40 cm extension was

attached to the top of the trap, as necessary, to permit sampling in water depths up to 100 cm.

The trap was hand thrown from a standing position in a randomly selected direction. The

trap was immediately pushed into the substrate to prevent animals from escaping under the

trap. All vegetation was then uprooted, rinsed vigorously, and examined for snails.

Three sites for throw trap sampling were selected in southwestern WCA3A based on

signs of snail presence (i.e., egg clusters on emergent vegetation and/or catching a few snails

in a preliminary trapping effort) and the presence of sawgrass stands adjacent to

prairie/slough habitat. We selected juxtaposed habitat types to minimize marked substrate

differences between habitats which might occur over large spatial scales. Since we selected

sites based in part on snail presence, these sites may not have represented snail densities

throughout the area. Because our purpose was to compare methods, rather than to estimate

wild snail densities, having a sufficient sample of snails was of greater concern than having

representative densities. All throw traps were placed at least 10 m from the ecotone defined

by the two habitat types to avoid edge effects. Within each of these habitats, at least 50

throw trap samples were collected. Bennetts and Kitchens (1993) calculated coefficients of

variation (CV) based on throwing up to 80 traps per site, and estimated that at least 50, and

maybe up to 100, throw traps per site were required to obtain reasonable precision (CV of 20

to 30%). They suggested that obtaining substantially lower coefficients of variation would

not have been logistically feasible given the labor intensity of the methods and variability in

snail abundance.

During 1996 we compared the number of snails extracted from throw traps using

either a dip net or a suction dredge. One of these two methods was randomly selected and

used to clear the trap. The dip net was constructed of welded aluminum pipe, consisting of a

1.5 m handle centered on a 0.30m (h) x 0.66m (1) frame, which supported 1.3 cm mesh

netting. It required two sweeps of the net to cover the entire trap. Once the vegetation was

removed, the net was passed through the trap until 20 consecutive sweeps (10 in each half of

the trap) devoid of snails were obtained. The suction dredge consisted of a self-priming 2-

cycle 5-hp pump, a Mays fluid transformer (Keene Engineering, Northridge, CA) to induce

suction (Brook 1979), and a reinforced rubber intake hose (7.5 cm diameter) [Neither the

State of Florida nor the Federal Government endorses commercial products mentioned in this

text]. The hose was attached to a 7.5 cm diameter aluminum handle with a 15 cm x 15 cm

box on the end. The dredge was operated until we had extracted the top 8-10 cm of the 1-m2

area enclosed by the trap. All material extracted through the suction hose passed over a

sorting tray of 1.3 cm wire screen and into a 1.3 cm mesh bag at the end of the sorting tray.

Extracted material was then sorted and all snails were removed. During the 1995 pilot study

we also explored the use of a bar seine for extracting snails from throw traps. The bar seine

was a Im x Im aluminum frame with two handles extending 0.5 m from each side of the

frame. The frame was covered with 1.3 cm mesh netting. The bar seine was swept through

the trap until 10 consecutive sweeps devoid of snails were obtained (one sweep covers the

entire trap).

For throw trap sampling to provide reliable estimates of snail density, it either must

be assumed that all animals within each throw trap are counted or the proportion of animals

counted must be estimated (Burnham 1981, Nichols 1992). We estimated the proportion of

animals counted using marked snails, which were placed in some throw traps after

deployment but prior to plant removal. The number of snails placed in the traps ranged from

0 to 5, which reflected the number of snails collected in 1-m2 traps (Bennetts and Kitchens

1993). The proportion of animals recovered from throw traps was then estimated as the

proportion of snails recovered to marked snails placed in the trap. This procedure was

intended to be "blind" (i.e., the person collecting the sample did not know if or how many

snails were placed in each throw trap); however, we later realized that our "blind" protocol

had not been strictly adhered to during our 1995 pilot study.

Effort was measured as the time required to clear a given throw trap at four sites

during 1995 using each of the initial three extraction methods. Extraction time was the time

from when the throw trap was positioned until it had been completely cleared; thus, it

included the removal of vegetation. This procedure was only intended to compare

extraction time among methods and did not reflect the total time required for sampling,

which would have included additional time for transportation, setup, and equipment

maintenance.

Data Analysis

A preliminary analysis of the throw trap data revealed that they were not normally

distributed (Shapiro Wilk test, P< 0.001) (SAS Inc. 1988), but were reasonably well fitted by

an unconstrained negative binomial distribution (G=14.96, 12 df, P=0.244) (White and

Bennetts 1996). Consequently, we used the likelihood-ratio testing framework for a negative

binomial distribution described by White and Bennetts (1996) to test for all main effects

attributable to site, habitat, and extraction method. The negative binomial distribution has 2

parameters: m (the arithmetic mean) and k (a dispersion parameter) (Bliss and Fisher 1953).

White and Bennetts' approach uses a combination of likelihood-ratio tests, Akaike's

Information Criteria (AIC) (Akaike 1973, Shibata 1989), and goodness-of-fit tests to

determine if m and/or k differ among treatment groups. A disadvantage of this approach is

that it is computationally difficult and limited software is available to analyze more complex

designs (White and Bennetts 1996). Consequently, we used ANOVA to further explore the

full suite of potential interaction effects. Although these data do not meet the assumptions of

ANOVA, it has been shown that ANOVA is quite robust to violation of its assumptions when

the data are distributed as negative binomial (Mitchell 1977), even when the variances are

unequal (White and Bennetts 1996).

We began our analysis of effort by modeling the relationship between extraction time

and the total number of marked or wild snails extracted using an ANOVA. Finding a snail in

a trap inherently increased the time required for clearing the trap because our search criteria

for both the dip net and bar seine were based on the number of sweeps without finding a

snail (i.e., finding a snail would result in 10 and 20 additional sweeps for the bar seine and

dip net, respectively). Consequently, we used the residuals from the first ANOVA as a

dependent variable to examine the additional effects of extraction method, habitat, and site,

having already accounted for increased time due to the number of snails extracted. We then

examined the effects of site, habitat, and extraction method on the residual times. Because

our "blind" protocol for estimating recovery of marked snails had not been adhered to during

1995, we included an additional effect in our analysis for whether or not marked snails had

been placed in the throw trap. A significant interaction between this treatment effect and

extraction method would have indicated if any bias was unequally distributed among

treatments.

Results

During the pilot study of 1995, one to five marked snails were placed in 112 throw

traps and extracted to examine sampling recovery efficiency. The proportion of marked

snails recovered did not differ among extraction methods in prairie habitats (X2=0.47, 2 df,

P=0.792), but was relatively poor using the bar seine in sawgrass habitats (X2=5.45, 2 df,

P=0.065) (Figure 2). Based on this result, the bar seine was dropped as an extraction

method for our 1996 effort in order to concentrate on the dip net and suction dredge.

In 1996 we extracted 610 throw traps to compare the performance of the dip net and

suction dredge in assessing snail density. The most snails extracted from a single throw trap

in 1996 was four, but most traps contained either one snail (94 traps) or no snails (492 traps).

A negative binomial model in which dispersion (k) was constant (i.e. no variation due to

habitat, site or method), but the mean number of snails per throw trap (m) differed among

Figure 2. The proportion of marked snails recovered from 112 throw traps in prairie/slough and sawgrass
habitat using a bar seine, dip net, and suction dredge. Proportions were obtained by dividing the number of
marked snails extracted by the number of marked snails put in the throw trap.
1.0
0.9 Prairie
0.8 Sawgrass
0.7
0.6
S0.5
c 0.4
g 0.3
e
0 0.2
0.1
0.0 1
Bar Seine Dip Net Suction Dredge
Extraction Method

extraction methods, sites, and habitats was supported by our data (Table 1). This model had

the lowest AIC score and was further supported by all likelihood-ratio tests (at P=0.05). We

also had no evidence for lack-of-fit of this model (G=17.64, 23 df, P=0.777). An ANOVA

also supported the conclusion of our negative binomial model indicating that mean number

of snails per throw trap differed among all main effects (habitat, site, method), and further

supported the inclusion of interaction effects (Table 2).

The number of snails/m2 was substantially higher in prairie/slough habitat using the

suction dredge at site 2 compared to other sites, habitats, or extraction methods (Figure 3).

Table 1. Description of negative binomial models and their corresponding Akaike Information
Criteria (AIC) scores. Lower AIC scores indicate more parsimonious models. "None" refers to
models not accounting for contributing effects of either Site, Habitat, or Method or any
combination (i.e., potential effects are unknown or random). Also shown is the parameter structure
(i.e., whether m [the arithmetic mean] and/or k [dispersion] differed among extraction methods,
habitats, or sites).
Model Source(s) of Source(s) of No. AIC
Variation (m) Variation (k) Parameters
1 None None 2 749.25
2a Site, Habitat, Method None 13 714.89
3 None Site, Habitat, 13 767.89
4 Site, Habitat, Method Site, Habitat, 24 734.21
5 Site None 4 736.34
6 Habitat None 3 734.93
7 Method None 3 740.64
8 Habitat, Site None 7 719.26
9 Method, Site None 7 730.74
10 Habitat, Method None 5 724.50

a Model selected based on AIC, likelihood-ratio tests and goodness-of-fit.

Figure 3. The mean (SE) number of apple snails per m2 in prairie and sawgrass habitats at each of 3 sites in
WCA3A during 1996 using a dip net and suction dredge.

Site 1
Prairie
Sawgrass

Dip Net Suction
Dredge

Site 2

!

-~Di Net Suction-

Dip Net Suction
Dredge
Site/ Method

Site 3

Dip Net Suction
Dredge

Table 2. Analysis of variance (ANOVA) table for fully saturated model of wild apple snail numbers
in relation to habitat (HAB), site (SITE), and extraction method (METH) from throw traps. Sums of
squares (SS) are type III partial SS, which are adjusted for all other terms in the model (SAS Inc. 1988).
Source df SS MS F Source
HAB 1 4.629 4.629 17.57 <0.001
SITE 2 5.282 2.641 10.02 <0.001
METH 1 3.029 3.029 11.50 0.001
HAB*SITE 2 4.952 2.476 9.40 <0.001
HAB*METH 1 2.659 2.659 10.09 0.002
SITE*METH 2 1.677 0.838 3.18 0.042
SITE*HAB*MET 2 1.914 0.957 3.63 0.027
Error 598 157.553 0.263

1.2

1.0 -

0.8

0.6

0.4

0.2

0.0

Higher numbers of snails tended to be extracted using the suction dredge in all prairie/slough

habitats compared to sawgrass habitats (Figure 3). The suction dredge also extracted more

snails than dip net in the prairie habitats of each site. Snail densities appeared similar among

sites in sawgrass habitats using either extraction method (Figure 3).

We evaluated extraction time (effort) from 955 throw traps during 1995. Our

analysis indicated that after accounting for the number of snails (marked or unmarked), the

time of extraction was influenced by site, habitat, extraction method, and whether or not

marked snails had been placed in the throw trap (Table 3). Sawgrass habitats tended to take

Table 3. Analysis of variance (ANOVA) table for our final model of residual extraction time after having
taken into account the time attributable to the number of snails extracted. Effects were habitat (HAB), site
(SITE), extraction method (METH), and whether or not marked snails had been placed in the quadrat
(MARK). Sums of squares (SS) are type HI partial SS, which are adjusted for all other terms in the model
(SAS Inc. 1988).
Source df SS MS F Prob >F
HAB 1 1391.699 1391.699 54.43 <0.001
SITE 3 3246.463 1082.154 42.32 <0.001
METH 2 371.247 185.623 7.26 <0.001
MARK 1 116.109 116.109 4.54 0.033
SITE*HAB 3 2263.432 754.477 29.51 <0.001
MARK*SITE 2 304.762 152.381 5.96 <0.001
Error 942 34468.146

more time than prairie/slough habitats, particularly at site 1 (Figure 4). Our final model also

indicated a site*habitat interaction effect and an interaction between sites and whether or not

marked snails had been placed in the throw trap. Overall, more time was expended on throw

traps in which marked snails had been placed (residual = 2.70 minutes) compared to throw

traps in which marked snails had not been placed (residual = -0.36 minutes). However, our

data did not support the inclusion of an additional interaction between extraction method and

whether or not marked snails had been placed in the throw trap (F2,940=1.99, P=0.138).

Figure 4. The mean (SE) residual extraction time in prairie and sawgrass habitats at each of 4 sites in WCA2B
during 1995 using a bar seine (BS), dip net (DN), and suction dredge (SD) after having accounted for number of
snails in the throw trap.

25

20 U Prairie
I Sawgrass
15

E
F: 10

(5

0 -

-5
Site 1 Site 2 Site 3 Site 4

-10
BS DN SD BS DN SD BS DN SD BS DN SD
Extraction Method

Discussion

Snail extraction using the suction dredge yielded the highest estimates of snail

density. Because removal of snails precludes multiple counting of individuals, we also

suggest that higher estimates probably were more accurate. This conclusion is further

supported by our independent estimates of recovery probabilities. The suction dredge

consistently had the highest recovery of marked individuals, regardless of habitat type.

The relative effectiveness of these extraction methods probably reflects how effective

they are at removing snails from uneven substrates. Plant removal creates numerous small

depressions and holes into which snails may fall, thereby avoiding collection by the dip net

and bar seine. We have confirmed this by retrieving unrecovered marked snails by hand

following attempted extraction. The use of marked snails to assess sampling efficiency,

regardless of the extraction method employed, is strongly recommended when using throw

traps.

Differences among habitats in extraction time from throw traps were not surprising.

Sampling snails in sawgrass habitat, with its greater vegetation density and more rigid

structure, takes longer regardless of the extraction method used. Differences in sites also

were not surprising, given variability in substrates and vegetation density. In addition, site 1

was the first site of our pilot study and may have taken longer due to lack of experience

working with the extraction techniques. Longer time for throw traps having marked snails

was likely due to "observer expectancy bias" (Balph and Balph 1983). During the 1995

pilot, when extraction time was measured, observers were aware of whether or not marked

snails were in the throw trap. When an observer knows that a snail is present in the throw

trap, whether intentional or not, effort may be increased to insure its recovery. We did not

measure extraction time during 1996, but we have no reason expect this bias was again

present, since observers were "blind" to whether or not marked snails were in the throw trap.

The lack of an interaction between whether or not a marked snail was in the throw trap and

the extraction method indicates that this bias was not differentially distributed among

extraction methods. Consequently, we believe that our comparisons among capture

probabilities of different methods during 1995 were reasonable. However, we strongly

suggest that estimation of recovery probabilities always be "blind" as to whether or not

marked snails are placed in a throw trap.

Logistical constraints precluded evaluation of some approaches that have been used

to sample apple snails. For example, Donnay and Beissinger (1993) recently sampled

Pomacea doliodes using a seine in conjunction with transects, rather than throw traps.

Although using a seine without the constricting sides of a throw trap may improve its

performance, it is likely to suffer from the same problems we encountered with missing

snails in the substrate. Thus, estimation of recovery probabilities would likely be necessary

for reliable results.

Despite its superior performance relative to the dip net and bar seine, the suction

dredge does have limitations as a throw trap extraction technique. The pontoons supporting

the dredge pump are cumbersome to maneuver through vegetation during sampling,

especially in sawgrass. When sampling, the handle and hose are filled with water, which

adds considerable weight (estimated at 10 kg) which the operator must move up and down to

cover the 1-m2 trap area. Also, the suction dredge cannot function adequately in less than 15

cm of water, a depth common to much of the apple snail's range, especially late in the dry

season. The intake hose requires submersion in at least 15 cm of water. Another drawback

with the suction dredge is the wire screen in the sorting tray; it had a tendency to damage the

snails forced through the dredge. This may not be acceptable for a series of sampling efforts

over time in the same site, for example. All the throw trap methods required vegetation

removal, which alters the habitat and may influence subsequent sampling results.

An important consideration of any potential sampling method is the amount of effort

involved in sampling. The throw trap we used to penetrate vegetated habitats to study snails

weighs in excess of 18 kg. This, in combination with the effort required to uproot vegetation

from the trap, makes this method very labor intensive regardless of the extraction technique

employed. A large number of 1-m2 throw trap samples also are needed to make

comparisons of apple snail density (e.g. 492 of 610 traps yielded no snails) (Bennetts and

Kitchens 1993). We deployed at least 50 traps per method per site to gain enough precision

to compare the extraction methods in each habitat type.

Finally, the throw trap data provides some indication of the distribution of apple

snails among two habitat types, sawgrass and prairie/slough. Owre and Rich (1987) and

Turner (1994) hypothesized that apple snails do not use interior sawgrass marsh to any great

extent. Their suggestion was based on egg cluster indices, which we have found are not

reliable due to the high temporal and spatial variation (see Section 3.5 Egg Cluster Counts).

Our data from the throw trap effort indicate that apple snails regularly occur within stands of

sawgrass, although densities may be lower than in adjacent prairies or sloughs. Sawgrass

represents a substantial proportion of South Florida's wetlands. For example, in WCA3A,

sawgrass covers more area than wet prairie (Davis et al. 1994). Thus, even at lower

densities, the numbers of snails occurring in sawgrass may constitute a substantial portion of

the overall apple snail population. Some additional data on habitat distribution are described

in the context of our mark-recapture study (Section 3.3).

3.2 Movement Based Wire Traps

In addition to analyzing which extraction technique was most effective for

determining snail densities from throw traps, we planned to investigate less labor and time

intensive methods for determining relative abundance. For this purpose, we used throw traps

as a direct measure of snail density to validate other potential methods of sampling snails. In

this section, we present results of using wire traps as a tool for assessing relative snail

abundance.

Owre and Rich (1987) had suggested using traps made from plastic containers in their

treatise on snail sampling, but after limited success abandoned the idea. Mike Miltner (GFC,

pers. comm.) at times captured nearly equal numbers of snails as crayfish while using wire

crayfish traps baited with gizzard shad. Steve Miller (SJRWMD, pers. comm.) also found

snails in baited crayfish traps in the BCWMA East. Tim Towles (GFC, pers. comm.) found

dozens of snails in funnel traps used in wetland herpetological surveys. This anecdotal

evidence indicated that the number of snails moving into wire traps warranted investigation

as an indicator of snail abundance. In this section we describe two trapping systems which

can be used to collect snails, and we compare the snail catch in these traps to snail densities

determined by throw traps extracted with a suction dredge.

Methods

Crayfish Traps

In a 1995 pilot study in BCWMA East we used 7 crayfish traps with and without dead

fish as bait. After this pilot study and a concurrent telemetry study (Chapter 4), we realized

bait was unnecessary and proceeded without bait with equal success. Upon ordering

additional traps in 1996, we switched to a slightly modified version of a commercially

available crayfish trap (Sam Lemmond Enterprises, Salt Springs, FL). These traps consist of

three components: the funnel base, the stack, and a lid. The funnel base is constructed of

vinyl coated wire with 19 x 24 mm hexagonal mesh. The base supports three funnel

entrances which have an outer diameter of 15 cm and an inner diameter (i.e., inside the trap)

of 5 cm. The mesh size and funnel openings limit our snail catch to snails with shell lengths

of approximately 22 to 50 mm. [Mean snail ( S.D.) size for all BCWMA snails collected in

1995 was 36.9 3.6 mm]. The funnel base is 75 cm and tapers to 18 cm where it attaches to

the stack. The stack is made from wire with a 25 x 13 mm rectangular mesh. The lid (not

sold with the trap) is made from extruded polyethylene plastic with 13 mm mesh and is

attached to the stack with nylon cable ties. The trap is secured in place by attaching a nylon

cable tie through the stack wire and around the pvc pole driven into the substrate. The total

trap height is 90 cm; use in water depths greater than 90 cm does not permit snails (or

incidentally trapped animals) to breathe air.

Crayfish traps were deployed in BCWMA East and in southwestern WCA3A. These

traps have been used in all wetland habitat types encountered, including wet prairie, slough,

sawgrass and cattail. Deployment requires pushing aside only a 1 m diameter area of

vegetation. Effort was made to place the trap as near to the substrate as possible, but

successful use did not require that the trap rest firmly on the bottom (which would require

labor intensive plant removal). We have observed snails in the field moving along vegetation

and on periphyton suspended loosely above the substrate, and they likely can enter the funnel

traps via plant stems, roots or a flocculent periphyton or organic layer.

BCWMA study- In 1996, we used crayfish traps in a study of snail movements in

relation to reproductive activity in BCWMA East (see chapter 4). Data from this study are

included here to illustrate the utility and limitations of using these traps to study snail

populations. During the reproductive ecology study, we deployed 54 traps split into 3 groups

of 18 traps (data from some traps were censored if they fell over). Traps were placed

approximately 5 meters apart. We monitored traps each month from February through May,

in late June/early July, and in August, for a total of six trap sessions and 622 trap checks. A

trap session consisted of checking each trap on two occasions over a 7 to 12 day period.

Traps were not moved during a trap session. We checked traps at 3 day intervals for 5 of the

12 occasions, 4 day intervals for 4 of 12 occasions, and for one occasion each a 5, 7, and 9

day interval. For each trapping session (n=6) we compared the total snail catch from the

shortest time interval (i.e., 3 or 4 days) to the longer of the two time intervals (i.e., 4, 5, 7 or

9 day interval) to assess time effects within trap sessions using a one-way ANOVA (snail

catch= trap interval + trap session + interactions). This was done using the General Linear

Model (GLM) procedure (PROC) in SAS (SAS Inc. 1988). Due to interaction effect between

trap interval (short vs. long) and session, the short vs. long time interval within a session was

compared using the "slice" option in PROC GLM in SAS (SAS Inc. 1992), wherein the

interaction sums of squares were partitioned by the session sums of squares, and the

denominator for the F-test (to compare trap interval within sessions) was the error term from

the ANOVA. Because of the interaction effects, we could not perform a meaningful

regression of snail catch to trap interval across sampling sessions. Based on our results of

the effect of trap interval, we adjusted total snails per session to reflect a 7 day trap interval

to give a snail catch per trap-week. The adjustment was made by dividing the number of

snails by trap interval (days) and multiplying by 7 days/week.

We studied escape rates by placing marked snails (plastic i.d. numbers affixed to the

shell with marine epoxy) in crayfish traps and checking traps 1 to 8 days later. We placed 21

marked snails in 16 traps for an 8 day check, 30 marked snails in 20 traps for a 2 day check,

and 14 snails in 13 traps for a 1 day check.

WCA3A study- In southwestern WCA3A in 1996, we evaluated the use of crayfish

traps as a method for determining relative snail abundance. Throw traps with suction dredge

extraction was the standard for validation. The crayfish traps were deployed in two habitat

types, sawgrass and wet prairie, at three different sites (i.e., six areas sampled) (sites

described in section 3.2). Crayfish trapping was performed within one week of sampling

using the throw trap and suction dredge. Two rows of 10 traps were placed at each area

sampled; this (i.e., 20 traps) was considered one sampling unit. Rows were 10 m apart and

traps within rows were separated by 6 m. Traps were checked 3 days following initial

placement, and then moved 10 meters to a new trap line. Traps were checked again after a 4

day interval, which concluded our 7 day trap effort within a habitat type. Since traps were

moved after 3 days, we treated the second location as a replicate sample within the same

habitat. We compared the mean snails captured per crayfish trap to the mean snail

density/m2 using linear regression (GLM procedure in SAS)(SAS Inc. 1988).

Trap Arrays

Trap arrays consist of a series of barriers which direct moving animals into funnel

traps (Enge 1997). They are most frequently used for trapping amphibians and reptiles in

uplands and wetland habitats (Enge 1997, Dodd 1991). Initially, we constructed a trap array

with walls of silt fencing (woven polypropylene) and funnels of window screening as

described by Enge (1997). After initial tests with these arrays, we decided to enhance their

utility for trapping apple snails in wetland and lake habitats.

Our snail trap consists of three major components: 1) center funnel unit; 2) three

single funnel units; and 3) three walls to divert animals into funnel units. The walls were

configured in a Y-shaped array. At the end of each of the three walls was a funnel trap. We

constructed the funnel traps to extend above water level (up to about 1 meter) in order to

allow snails, or incidentally trapped animals, to breathe air. Plastic coated wire was used to

construct the frame of the funnel units to increase stability. The funnel entrances were made

from 13 mm netting, fabricated from extruded high density polyethylene (Nalle Plastics,

Austin, TX). We also used the 13 mm plastic netting for construction of the 5-meter long

walls. Unlike plastic sheeting, the mesh stands up to wave action, which can be substantial

due to boat traffic and winds. The extruded plastic material has excellent durability yet is

sufficiently pliable to roll the walls and partially compress the funnel traps for transport.

Four pvc poles, each two meters long, were attached to the walls using nylon cable ties, and

these poles were driven into the substrate to hold up the walls and funnels during sampling.

For practical use, these traps are generally limited to wet prairie and slough habitats

(e.g., areas with rushes, maidencane, and floating-leaved macrophytes in moderate to low

densities). Proper deployment requires that the three 5-meter long barrier walls and funnel

traps rest on the bottom, which requires uprooting or cutting down relatively large

macrophytes such as sawgrass, cattail and pickerelweed. Engaging in this labor intensive

process was counterproductive to our goal of developing a sampling method less labor

intensive than throw traps, so we restricted trap array use to wet prairies and sloughs.

Practical use of trap arrays is also limited to water depths of less than one meter, because

deployment requires workers to handle trap components at the substrate level.

We compared snail densities, as determined by suction dredge extraction of 1 m2

throw traps, to the number of snails captured in trap arrays at eight different sites on Lake

Kissimmee. Sampling was conducted for 33 days from 9 October through 10 November

1995. For each of the eight site comparisons, 50 randomly placed throw trap samples were

taken. Only snails larger than 13 mm in length were sampled due to the mesh size of our

traps. We placed 13 trap arrays in the 8 sites. Vegetation in the immediate vicinity of the

trap array walls and funnels was removed with a hoe, as necessary, in order to ensure that the

trap system rested on the substrate. Five traps were moved within a site (to obtain a more

representative sample), and 5 traps were checked twice in one location for a total of 23 trap

checks. A single trap array check was considered a sample unit, so n varied from 1 to 4 per

site. The one site with n =1, we realized after a very difficult trap deployment, was

extremely difficult to work in so we elected to limit this site to one check after a 23 day trap

interval. Trap interval ranged from 7 to 23 d among the 23 checks. Six trap units were also

deployed in BCWMA East in 1996 for collecting snails; data from this effort are included to

illustrate general trap utility. The relationship between trap array snail catch and throw trap

snail density was analyzed by linear regression (GLM procedure in SAS) (SAS Inc. 1988).

Results

Crayfish Traps

Crayfish traps proved to be an effective way to capture snails with much less labor

than throw traps. A set of 20 traps can be set up in sawgrass, cattail, slough or wet prairie

habitats by one person in less than 30 minutes. Twenty traps can be checked in 10-15

minutes. Monthly trapping efforts in BCWMA in 1996 yielded 93 to 226 snails per 100 trap

checks. We examined the effect of sample size on the precision of the abundance estimate

by plotting CV as a function of the number of traps deployed. It appears that using

approximately 30 traps per area results in a stable CV of approximately 20 to 15% (Figure 5).

Total catch from crayfish traps was weakly related to snail density as determined by

throw traps extracted with a suction dredge (R2= 0.31, n= 6, P= 0.051) (Figure 6). However,

ten of the crayfish trap catches were between 0.18 and 0.3 snails/m2. If the two replicate

crayfish trap catches from WCA3A site 2-prairie are removed from the analyses, leaving only

Figure 5. Change in coefficient of variation of snails captured in crayfish traps in the BCWMA with increasing

sample sizes. Each sample represents the number of crayfish traps, all having a 7 day trap interval.

0.3

0.25
i

0.2.

0
0.15

0.1
0.1 -- -- --i--- i -- ] -- -- -- i -- -- -- l l I i
6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51
Number of Crayfish Traps

Figure 6. Snails caught in crayfish traps and estimated snail densities. Crayfish trap data are the total catch of

snails per 20 traps placed in each site (n=6 sites). Snail density estimates were derived from throw trap sampling

using a suction dredge in prairie/slough habitat.

45
40
| 35
S 30 -
S 25
o 20 -
N
15
I 10
5
0
0 0.2 0.4 0.6 0.8 1
snailslm2

the snail catch from sites with densities between 0.18 and 0.3 snails/m2, the relationship falls

apart (R2-0.006, n=4, P=0.82).

In the BCWMA study, trap interval (number of days) did effect the number of snails

captured in crayfish traps ( F1,15 = 3.86, P=0.0011) (Table 4), however there was an

interaction between trap interval (short vs. long) and trap session (Table 4). The analysis

between short and long trap interval within each session resulted in only 2 sessions (April

and May) showing significant differences (Table 5). This did not affect the crayfish trap vs.

throw trap regression, since trap interval (7 d) was constant for all sites. We decided to

adjust all snail catches for the BCWMA study to reflect a one week trap interval, even

though the difference in trap intervals was observed only in the April (3 versus 4-day trap

interval) and May (3 versus 9-day interval). The adjustment was made by dividing the snail

Table 4. Analysis of variance (ANOVA) table for the model describing snail catch as related
to trap interval (TI)(short vs. long) and trap session (SESS). Sums of squares are type III
partial SS, which are adjusted for all other terms in the model (SAS Inc. 1988).
Source of Variation df SS MS F Pr>F
TI 1 28.47 28.47 10.73 0.0011
SESS 5 155.33 31.07 11.71 <0.001
TI*SESS 5 65.25 13.05 4.92 <0.001
Error 600 1592.00 2.65
Total 611 1841.06

catch by trap interval (days) and multiplying by 7 days/week. Total catch/week did vary by

trap session (Figure 7), and likely relates to changes in snail breeding activity, snail

movements, and population turnover over the seven month period of sampling (see Chapter

5).

Escape rates from crayfish traps were 0 for both one and two-day trap intervals (n=14

and n=30, respectively). Mean 8-day escape rate was 28.6%.

Crayfish traps and trap arrays collect a variety of animals other than snails, some of

which are air breathers. Out of 460 crayfish trap checks in BCWMA, we counted 127

incidental captures; these consisted of 15 gar, 88 centrarchid fish, 5 snakes (predominately

green water snakes), 5 turtles (mud turtles), and 11 amphiumas. Snakes and gar, whose jaws

Table 5. Tests for differences between short and long interval within trap sessions.
Comparisons were made using the results of ANOVA (Table 4) and partitioning the
INT*SESS sums of squares by SESS sums of squares using the "slice" option in PROC
GLM in SAS (1992). The denominator for the F-test is the error term from the ANOVA.
TRAP SESSION F-Statistic PR >F
FEB 0.85 0.36
MAR 0.44 0.51
APR 8.11 0.0045
MAY 16.99 0.0001
JUN/JUL 0.46 0.50
AUG 0.0012 0.97

tended to get stuck in the mesh, suffered the greatest mortality (20% of snakes and 13% of

gar captured).

Trap Arrays

The trap arrays proved effective as a method for determining relative abundance of

apple snails. Linear regression analysis of the data for snails collected using throw trap

sampling versus trap arrays indicate a strong relationship (R2=0.90, n=8, P <0.001) (Figure

8). Length of trap interval did not appear to affect trap catch, but we had a small set of

samples from which to work (only 3 situations where traps in the same location were

checked twice and that also had at least one snail captured). The reason for the longer

Figure 7. Total catch of snails from crayfish traps (adjusted to reflect a 7 day trap interval) from six trapping
sessions conducted February August 1996 in BCWMA East. Data for each session are from either 52 or 53
traps.
200

180

160

140

S120

J* 100

80 -

60

40

20

Feb Mar April May JuniJul Aug

interval was poor weather (approximately October 15 through October 30), which may have

caused the lower catch for those three particular occasions.

Out of 22 trap array checks (4 traps per array) in Lake Kissimmee, we counted 26

incidental captures. These captures consisted of 6 gar, 13 centrarchid fish, 1 green water

snake, 2 turtles (one soft-shell), 1 frog, and one unharmed 0.5 m alligator. The deaths

consisted of 4 gar (67%) and one centrarchid fish (7.7%).

Figure 8. Snails caught in trap arrays and estimated snail densities on Lake Kissimmee. Trap array catch is
the mean catch per trap for each site (n = 1-4). Snail density estimates were derived from throw trap
sampling using a suction dredge (n = 50 throw traps per site). Eight sites were sampled (3 points overlap at
0 snails/m2).
so

50

40

*
D 30

20

10

0 20 40 60 80 100 120 140 160
snailslSO m2

Discussion

One of the main reasons the ecology of apple snails has remained a mystery is the

difficulty associated with collecting snails from the graminoid marshes and littoral zones

they inhabit. As a result of our effort to develop a trap-index of relative snail abundance, we

have developed two effective trap systems for collecting large numbers of snails. Both

systems require considerably less effort than dip nets, seines or a suction dredge. Both trap

systems rely on snail movements and do not require bait to attract snails.

We believe that crayfish traps will be an effective tool for collecting apple snails in

most habitats. They are easy to deploy, even in dense vegetation, and they require little, if

any, permanent disturbance of the vegetation. Based on our use in areas where snails are

known to exist, a capture rate of 1 to 3 snails per trap over a 3-5 day period is typical.

Escape studies indicated that trap intervals should be limited to a few days as nearly 30% of

snails escaped after eight days, but none escape after one or two days. Escape studies were

not conducted with trap arrays.

Trap arrays are also effective in collecting snails. Trap arrays left for one week in

eastern BCWMA collected 5 to 18 snails / trap check. On Lake Kissimmee a single trap set

for one week collected from 12 to 91 snails per trap check. Unlike crayfish traps, trap arrays

are not suitable for use in dense vegetation (e.g., sawgrass), and in general require more

effort per snail captured than crayfish traps. Our trap array design proved to offer several

advantages to other trap array designs. They can be set up quickly (typically less than 30

minutes, but up to one hour in muck sites). Compared to funnels made from fine screen and

walls made from plastic sheeting, they collect much less drifting vegetation and debris and

stand up against boat and wave action. Snails and other air breathing animals collected in

the traps (i.e., snakes, turtles) could breathe air in the trap arrays, whereas submerged screen

funnels result in drowning. The impact of restricting snail access to aerial respiration is

unknown; however, we do know that apple snails continue to breathe air in water with high

dissolved oxygen concentrations (McClary 1964, pers. obs.).

The discrepancy between the two movement based methods (crayfish traps and trap

arrays) and a direct measure of snail density (estimated with throw traps) may be explained

by: 1) trap design and function; 2) the different seasons in which these comparisons were

made; 3) different trap intervals; and/or 4) validation study design. The design of both

crayfish trap and trap array sampling depends upon apple snails crawling into the traps.

Neither trap requires bait, because, as will be discussed in chapter 4, snails are frequently on

the move. However, the ways in which snails move into the funnels of the two trap systems

differ. The array barrier walls (total linear surface of 30 meters if both sides of the three

walls are considered) divert snails from their "intended" course of travel and "direct" them

into the funnels. Snails enter crayfish trap funnels only if one of the three 0.012 m wide

funnels lies in their course of travel (which includes horizontal directions of travel as well as

vertical ascents to breathe air).

We have found that in a given area, a snail's tendency to crawl into the crayfish trap

varies with the reproductive activity of the population (see Chapter 4). Since snails are not

directed towards the traps, crayfish trap effectiveness depends more on snail behavior, which

varies considerably over just a few weeks or months during spring and summer. This brings

us to the issue of when the direct method validations were conducted for the trap arrays and

crayfish traps. Trap array validation was completed during Fall 1995. By fall, snail

populations are 3 to 5 months past their peak reproductive activity (Odum 1957, Hanning

1979, this report Section 3.5) and egg production is minimal as populations approach the

non-breeding winter months. In contrast, crayfish trap validation was spread out over June

and July 1996, which typically includes the transitional period between peak reproductive

activity and declining adult survival (see Chapters 4 and 5). Given these observations of

survival and reproduction during spring/summer, the crayfish trap validation likely

incorporated a high degree of temporal variation in snail movements and survival, and this

may have contributed to a lack of a relationship to snail density.

We found considerable variation in snail catch within trap sessions (2 checks in 7

days), and between trap sessions (6 sessions over 7 month period). This variation may relate

to changing weather, fluctuating water temperature, or biological factors affecting snail

movements. Crayfish traps were checked at 3 or 4 day intervals for a total of 7 days. Arrays

were checked at intervals of 7 to 23 day intervals over 21 to 28 days. The longer trap

interval between array checks and the longer amount of total time sampling may negate or

minimize the impact of short term fluctuations in snail movements.

Finally, validation study design may be an issue. In some cases, the crayfish traps

used for validation were deployed in one site and then moved to another. Without some

overlap between all sampling sites, time effects contribute to variation between sites. As

mentioned earlier, considerable variation in crayfish trap snail catch occurs over as little as a

few weeks, especially around peak breeding activity. In contrast, trap arrays were set out in

eight sites with some overlap in sampling for all sites during the 21-28 days of sampling.

The simultaneous trapping among all eight sites controlled for time effects during the trap

array validation.

Both crayfish traps and trap arrays are effective tools for collecting large numbers of

apple snails, tools which already have contributed substantially to unraveling the mystery of

apple snail ecology and life history. However, based on our assessment, crayfish trapping is

not a suitable index of apple snail abundance. Trap arrays do appear to be reliable as an

index, and we have gained important information on snail populations with their use (Darby

et al. 1996). However, their practical use is limited to wet prairie and slough habitat, which

does not permit studies of snail habitat use.

There are other issues which make trap arrays inconvenient to use. Trap arrays are

not available commercially, so individual researchers must spend considerable time and

effort to construct a trap (estimated 8 worker-hours per trap, estimated cost of supplies

$200). Their bulk limits transport to two to three trap units in an airboat. Setting and

checking trap arrays can be difficult, especially in water over 75 cm, because the workers

placing the trap must be able to reach to the bottom of the funnels and barrier walls.

Connecting the walls to the funnel units and lifting the center funnel unit is difficult without

two workers. In contrast, crayfish traps are considerably more convenient to use. Crayfish

traps are available commercially for $12 to $16 each. We frequently transported 30 to 35

traps in our airboat. Crayfish traps are easier to deploy (vegetation need not be cut or

uprooted) and are easily checked by one person. Crayfish traps could conceivably be placed

in any water depth, but if placed in water over 90 cm, drowning mortalities of incidentally

trapped air breathers (turtles and snakes) will increase. We have not tested the rate of snail

mortality if access to aerial respiration is prevented. Crayfish traps are practical to use and

effective in sawgrass and cattail as well as wet prairie and slough, but more data are needed

to understand variations in their capture rates. Our understanding of capture probabilities

with crayfish traps as part of mark-recapture studies is presented in the next section.

3.3 Mark-Recapture

Given their ease of use and efficacy in collecting apple snails, we decided to further

explore the potential use of crayfish traps in determining snail density. Crayfish traps can be

used to determine an actual density, rather than a relative abundance (as described in 3.2), if

there is an associated estimate of capture probability. The method of mark-recapture (or

capture-recapture) determines animal population or density based on such estimates of

capture probability. Since the 1940's, the statistical methods for estimating population size

using mark-recapture models have developed dramatically (Otis et al. 1978, White et al.

1982, Seber 1986, Nichols 1992). Mark-recapture methods have been used with a wide

range of animals, including small mammals, large mammals and birds, and in more recent

years with animals such as lobsters (Evans and Evans 1995) and snakes (Luiselli et al. 1996).

In general, mark-recapture studies proceed as follows. A number of traps are set up

in the study area (e.g., 100 traps in a 10-trap by 10-trap grid). Using the traps, a sample

(designated nl) is taken from the population; the animals are marked and returned to the grid.

After allowing time for the marked and unmarked animals to mix, a second sample

(designated n2) is taken. Of this second sample, the animals that are already marked (the

recaptures) are designated as v2. The proportion of the population that is marked is called

the "capture probability", and its estimate (p-hat) can be calculated by dividing the number

recaptured by the total number caught in the second sample:

p-hat = v2 / n2

An estimate of the total population, N-hat, can then be obtained:

N-hat = n, / p-hat.

To improve the precision of the population estimate (N-hat), one would continue

marking the unmarked animals caught, releasing them, and resampling. A density estimate

can be determined by dividing the population estimate by the area of the grid. However, this

is a 'naive' density estimate, because the effective area of trapping is greater than the actual

area of the trapping grid due to what is known as edge effect (Otis et al. 1978). The

computer program CAPTURE, which we initially used for our snail analyses, takes edge

effect into account by using concentric trap grids to allow concurrent estimation of density

and edge width (Otis et al. 1978, Rexstad and Burnham 1991). A more recently developed

program, TMSURVIV (Jim Hines, USFWS, Patuxent Wildlife Research Center), handles

some violations of closure (explained below), but does not account for edge effects;

therefore, a naive estimate of density is obtained.

CAPTURE and TMSURVIV each include analyses of several model types that are

evaluated for their applicability to a given data set (see Data Analysis, below). Both

CAPTURE and TMSURVIV compute goodness-of-fit statistics and between-model test

statistics that are used to test model assumptions and help select the most appropriate model.

The results from both CAPTURE and TMSURVIV are discussed because they both illustrate

issues involved in applying mark-recapture studies to apple snails.

The closed mark-recapture models in CAPTURE have four general assumptions.

These are: 1) the population is closed; 2) animals do not lose their marks during the study;

3) all marks are correctly noted and recorded at each trapping occasion; and 4) each animal

has an equal and constant probability of capture on each trapping occasion (Otis et al. 1978).

The assumption that the population is closed (#1) means that the population size and

composition are constant over the course of the study; there is no birth or death (meeting

demographic closure) or immigration or emigration (meeting geographic closure). If either

of the two components of closure (demographic and geographic) are not met, the population

is said to be open. In order to increase the likelihood that the closure assumption is met, the

experiment should be conducted over as short a time period as possible and at a time that

will avoid recruitment (e.g., juveniles becoming trappable) and losses (Otis et al. 1978,

White et al. 1982). These closure issues were considered in the grid design and sampling

regime of our snail mark-recapture studies (see Methods, below). However, we found

closure to be an issue for all seven grids, and therefore TMSURVIV, which utilizes open

population models, was an important tool for interpreting the results of our apple snail mark-

recapture studies.

Methods

Mark-Recapture Technique for Apple Snails

Each sampling grid consisted of 10 rows of 10 traps, for a total of 100 traps. In the

pilot study, traps were an average 2.75 meters apart, encompassing a grid of 610 m2. For all

other grids, we used measured pvc poles as a guide to place traps 2.5 meters apart; the

resulting grid size was 510 m2. Inter-trap distances were never systematically tested to

optimize capture probabilities; they were based on our success in collecting snails with

crayfish traps and information on snail movements gained using telemetry (Chapter 4). Each

trap assembly consisted of a modified crayfish trap (described in 3.2) attached to a 1.5-m pvc

pole. Trap poles were numbered to keep track of snail capture location and to ensure release

of marked snails near their trap of origin.

Each sampling grid was checked seven times, allowing for six recapture occasions.

Trap check intervals were three or four days (based on capture rates from preliminary

crayfish trap work), so each grid ran for approximately 21 days. The same procedure was

used on each of the seven trap check occasions. The entire grid was checked, row by row,

and all newly caught snails were placed in plastic bags labeled with the appropriate trap

number. Information for recaptured snails (i.e., date, snail number and trap number) was

recorded on the spot and the snails were immediately returned to the grid approximately 1

meter behind the trap in which they were found.

Unmarked snails were taken to the boat for processing. For each animal, we recorded

the trap number in which it was caught, its size, gender, and designated tag number. Shell

size was determined to the nearest millimeter using a vernier caliper (S-T Industries, MN,

USA). Laminated, 3 mm x 5 mm plastic tags (Floy Tag & Manufacturing, Inc., Seattle, WA)

were used to mark snails. Tags were attached to the shell adjacent to the apex,

approximately 1 to 2 cm from the aperture, using a marine epoxy. Each shell was dried with

a towel, allowed to air dry, and lightly sanded in the area where the tag was to be placed.

Each snail was returned to the water approximately one meter from the trap in which it was

caught.

Pilot Study

The first mark-recapture grid for apple snails was set up in the northeast quarter of

BCWMA East. The sampling was conducted over a 23-day period in June 1996.

Mark-Recapture Studies in Three Habitat Types

This work was conducted in eastern WCA3A. We operated a total of six sampling

grids, three in Site 1 (Lat.: 25 59 13 N, Long.: 080 32 05 W) and three in Site 2 (Lat.: 25 57

67 N, Long.: 080 33 07 W). Three habitat types, cattail, prairie/slough, and sawgrass, were

sampled in each site. Sites were chosen in WCA3A by first scouting for areas which

possessed expanses of the three habitat types large enough for the grid and a 10 m wide

border on all sides. Once the size and vegetation requirements were met, we looked for

indications of snail presence (i.e., snail egg clusters or snails captured in crayfish traps).

Trapping grids were placed only in sites with an indication of snail presence. We selected

sawgrass, prairie/slough, and cattail habitats that were as close together as possible within a

site. The average distance ( SD) between habitats within Site 1 was 1305 m 661 m. The

average distance ( SD) between habitats in Site 2 was 1042 m 314 m. The shortest

distance between the two sites was approximately 3177 m (cattail-1 to cattail-2).

The sampling was conducted over a three-month period, March-May in 1997. Except

for the first and last week of the study, we ran two grids simultaneously. There was always

some temporal overlap between two grids.

Analyses

Each of the seven sampling grids was first analyzed using the computer program

CAPTURE (Otis et al. 1978, White et al. 1982, Rexstad and Burnham 1991). In addition to

density estimation, CAPTURE performs a test for closure (defined on p. 46); the utility of

this test, however, is questionable in situations where a true failure of closure can not be

distinguished from behavioral and some time variations in capture probabilities (White et al.

1982). CAPTURE also performs a uniform density test. This test examines the number of

snails captured across the grid according to the columns, rows, and rings of the grid. The

uniform density test, particularly the ring test, assists in assessing whether or not geographic

closure was met.

Our analyses using CAPTURE indicated that the important assumption of population

closure may have been violated in all seven grids. To further explore this possibility, we

turned to the open population models in TMSURVIV (Pradel et al. 1997). This computer

program is a modification of SURVIV (White 1986). We used six different models which

allow for different combinations of survival being constant or variable, capture probability

being constant or variable, and the absence or presence of transient animals passing through

the sampling grid; transience can be constant or variable. We focused the analysis on this set

of open models because a preliminary analysis of one grid indicated that transient animals

were present. Transient animals are defined as those traveling through the grid, whereas

residents are those that inhabit the grid area throughout the sampling period (Pradel et al.

1997).

Results

The mark-recapture approach was successful in determining snail densities in the 7

grids set up in BCWMA and WCA3A. We found no evidence that snail tags were lost, and

the ability to read the tags was not compromised by staining or biotic fouling. The number

of individual snails captured in the 7 trapping grids ranged from 102 in sawgrass (WCA3A,

Site 1) to 561 snails in prairie/slough (WCA3A, Site 2) (Table 6).

As stated above, the CAPTURE analyses indicated that the closure assumption may

have been violated in all of the grids. The TMSURVIV analyses confirmed these results.

Therefore, the snail population and density estimates presented here were obtained from

TMSURVIV for open populations. Note that three different capture-recapture models were

used to estimate densities for the seven grids (Table 7). This is because various mark-

recapture grids had unique data characteristics for which particular models were appropriate.

We found that no one model can be expected to fit all mark-recapture data for apple snails.

In each analysis, the most parsimonious model was selected based on a combination of

likelihood-ratio tests (LRTs), Akaike's Information Criterion (AIC) (Akaike 1973, Shibata

1989), and goodness-of-fit tests.

Table 6. Sampling periods and snail captures for six mark-recapture grids in WCA3A (1997) and one grid
in BCWMA (1996). No. of Snails Captured refers to unique individual snails. Total No. of Captures refers
to all snails captured in traps, including captures of the same individual on more than one occasion.
No. of Snails Total No. of
Dates Sampled Site Habitat Type Captured Captures
6-01 to 6-24 BCWMA-1 Prairie/Slough 291 569

3-12 to 4-03 WCA3A-1 Cattail 121 184

3-18 to 4-09 WCA3A-1 Prairie/Slough 117 153

4-04 to 4-27 WCA3A-1 Sawgrass 102 188

5-09 to 5-31 WCA3A-2 Cattail 392 559

4-28 to 5-22 WCA3A-2 Prairie/Slough 561 844

4-14 to 5-08 WCA3A-2 Sawgrass 127 178

Table 7. Survival and capture probability estimates from mark-recapture grids using TMSURVIV. PR refers to
prairie/slough habitat, CAT to cattail, and SAW to sawgrass. Numbers following habitat designation are site
numbers. Model refers to the most parsimonious model determined by likelihood- ratio tests, AIC and goodness-
of-fit tests.
Grid Model Survival p-hat
Blue Cypress Water Mangement Area
PR-1 SP 0.84 (0.020) 0.39 (0.026)

Water Conservation Area 3A
CAT-1 SPG 0.82 (0.063) 0.43 (0.064)
PR-1 SP 0.64 (0.075) 0.27 (0.065)
SAW-I SPT 0.77 (0.039) 0.44 (0.086)*
CAT-2 SPG 0.73 (0.043) 0.39 (0.040)
PR-2 SP 0.54 (0.023) 0.58 (0.037)
SA W-2 SP 0.72 (0.060) 0.32 (0.058)

In model SP, survival and capture probability are constant, and there are no transient animals.
In model SPG, survival and capture probability are constant, and there are transient animals.
In model SPT, survival is constant, capture probability varies, and there are no transient animals.

This model (SPT) generates p-hats for each trap occasion. The number shown is the mean (SE) of these
individual occasion p-hats. For the other models TMSURVIV directly provides one p-hat applicable to all
trap occasions.

Pilot Study. BCWMA East- The percentage of recaptured snails per occasion was

39.6 % on the second occasion and leveled out to a high of about 65 %. The closure test

performed by CAPTURE (Otis et al. 1978) indicated a violation of this assumption

(z = -3.21, P < 0.001). The uniform density test by rings failed to reject the null hypothesis

(X2 = 5.25, 4 df, P = 0.263), indicating that density was uniform; this test did not suggest a

geographic closure violation. The test further indicated that density was uniform among grid

columns (x2 = 12.07, 9 df, P = 0.209), but not among grid rows (x2 = 28.45, 9 df, P<0.001).

Using TMSURVIV for open populations, the model selected for the pilot study was

SP, the simplest model (Table 7). Under this model both survival and capture probability are

constant over the sampling period and there are no transient animals. This model had the

lowest AIC score of any of the models, the goodness-of-fit was high (G = 32.869, 34 df, P =

0.523), and the LRTs accepted SP over the more general models. Snail survival in this grid

was less than 1.0 (Table 7). This analysis indicates that the population was not open

geographically (there were no transients), but since survival was less than 1.0 between

sampling periods, it was open demographically. The population estimate, N (SE), was

204.56 snails ( 14.02). The density ( SE), calculated by dividing N by the grid size, was

an estimated 0.335 ( 0.023) snails per m2.

Eastern WCA3A Sites

Capture probabilities ranged in the 6 grids from 0.27 (0.065) in prairie/slough-1 to

0.58 (0.037) in prairie/slough-2. Capture probabilities did not differ noticeably between

habitat types (cattail AVG=0.41, prairie/slough AVG=0.43, and sawgrass AVG=0.38) or sites (i.e.,

Site 1AvG=0.38 vs. Site 2AVG= 0.43). The models selected and survival (an issue of closure)

in each grid are summarized in Table 7, and discussed separately as follows.

Cattail-1 Over the course of sampling, the percentage of recaptured snails ranged

from 22% to 48.6 % (Figure 9). The closure test indicated that closure was violated

(z = -2.123, P = 0.017). The uniform density test performed by CAPTURE indicated that

density was uniform throughout the grid (rings: x2 = 8.23, 4 df, P = 0.084; rows: X2 = 15.02, 9

df, P = 0.349; and columns: x2 = 15.02, 9 df, P = 0.090). The results of this test do not

suggest a geographic closure violation

The model selected for this grid using TMSURVIV was SPG, in which survival,

capture probability, and the proportion of transients are all constant. SPG had the lowest

AIC score of any of the models, the goodness-of-fit was the highest (G = 55.673, 33 df,

P = 0.008), and a LRT failed to accept the more reduced model. The survival of nontransient

snails was less than 1.0 (Table 7), indicating that snails were being lost. The population was

open both geographically (i.e., transients were present) and demographically (i.e., survival <

1.0). The proportion of residents (non-transients) among new (not previously marked) snails

was 0.653 ( 0.120 SE). The population estimate, N (SE), over the sampling period was

65.43 snails ( 9.55), and the density estimate was 0.129 snails per m2 ( 0.019).

Prairie/slough-1 The percentage of recaptures per occasion fluctuated (Figure 9).

According to the CAPTURE closure test, closure was violated (z = -2.459, P = 0.007). The

uniform density test indicated that density was uniform among rings (X2 = 8.13, 4 df, P

0.087), rows (x2 = 10.33, 9 df, P = 0.324), and columns (X2 = 7.20, 9 df, P = 0.617).

Using TMSURVIV, the appropriate model for prairie/slough-1 was SP, in which

survival and capture probability are constant and there are no transient snails. SP had the

second lowest AIC score of any of the models, but it was different from the lowest-scored

Figure 9. Percent recaptures per trap occasion in mark-recapture grids in WCA3A.

80.0-

60.0

40.0-

20.0-

0.0-

80.0

60.0

40.0

20.0

0.0

80.0

60.0-

40.0

20.0

0.0

- Slough-1

-

r

3 4 5

Trap Occasion

80.0-

60.0-

40.0-

20.0-

0.0

80.0

60.0

40.0

20.0

0.0

80.0-

60.0-

40.0-

20.0

0.0

7 2 3 4 5

Trap Occasion

Cattail-I

Cattail-2

Sawgrass-2

6 7

2

Slough-2

model by less than 2.0; AIC scores within 2.0 or less are not considered statistically different

(Sakamoto et al. 1986). Model SP also had a reasonable goodness-of-fit (G = 41.817, 34 df,

P = 0.168), and the LRTs accepted SP over the more general models. The survival of snails

was less than 1.0, indicating that snails were dying (Table 7). The population was open

demographically but not geographically (there were no transient snails). The population

estimate ( SE) over the sampling period was 81.64 snails ( 19.55). The estimated density

( SE) was 0.160 ( 0.04) snails per m2.

Sawgrass-1 The percentage of recaptures fluctuated over time (Figure 9) and

ranged from 10.5 % to 75.7 %. The closure test indicated a violation (z = -5.080, P < 0.001).

The uniform density test by rings suggested a geographic closure problem (X2 = 10.89, 4 df, P

= 0.028); more animals were caught in the outer two rings than expected based on a uniform

density. Also, density was not uniform among rows (x2 = 31.68, 9 df, P = 0.0002) or

columns (x2 = 17.43, 9 df, P = 0.043).

The TMSURVIV analysis selected model SPT as the most appropriate model for

sawgrass-1; with this model survival is constant, capture probability varies, and there are no

transients. Model SPT had the lowest AIC score, a goodness-of-fit ofG = 37.596, 29 df, P =

0.132, and LRTs failed to accept the more reduced model and more general models. Snail

survival was again less than 1.0 (Table 7), indicating that the population was open

demographically. The population estimate (SE) for the sampling period was 81.62 snails (

23.77). The estimated density was 0.160 ( 0.05) snails per m2.

Cattail-2 The percentage of recaptures ranged from a low of 15.9 % to a high of

46.4 % (Figure 9). The closure test indicated that closure was violated (z = -4.042, P <

0.001). The uniform density test by rings indicated a non-uniform density (X2 = 17.64, 4 df,

P= 0.0014), which suggests that the population may be open geographically. The test further

indicated that density was uniform among grid columns (x2 = 11.64, 9 df, P-0.234), but not

among rows (X2 =18.62, 9 df, P= 0.029).

Under TMSURVIV, the appropriate model was SPG; it had the lowest AIC score, an

adequate goodness-of-fit (G = 39.021, d f= 33, P = 0.217), and LRTs indicated it was most

appropriate. With SPG, survival, capture probability, and the proportion of nontransients

among new snails are constant. The proportion of residents (or nontransients) among newly

caught snails was 0.275 ( 0.095 SE); in other words, the proportion or probability that a

newly caught snail was a transient was a high 0.725. Survival was again less than 1.0 (Table

7). According to the analysis, this population was open both demographically (survival <

1.0) and geographically (transients present). The estimated snail population ( SE) in cattail-

2 was 210.76 snails ( 21.78). The snail density (SE) was 0.413 ( 0.043) animals per m2.

Prairie/slough-2 The percentage of recaptures per occasion ranged from 19.5 % to

55 % (Figure 9). Again the closure test indicated a closure violation (z = -9.413, P < 0.001).

The uniform density test by rings was not significant (X2 =3.95, 4 df, P = 0.412) and did not

suggest a violation of geographic closure. The three-part test also indicated that density was

uniform among rows (X2 = 10.67, 9 df, P= 0.299) and among columns (x2 =15.31, 9 df,

P=0.083).

The appropriate TMSURVIV model for this grid was SP, in which survival and

capture probability are constant over the course of the study and there are no transient snails.

This model had the lowest AIC score, the goodness-of-fit was good (G = 28.928, 34 df, P =

0.714), and the LRTs accepted SP over the more general models. Survival was substantially

less than 1.0 for this grid (i.e., 0.535) (Table 7), indicating that the population was not closed

demographically. The capture probability (0.584) was the highest of any of the grids. The

estimated population (SE) was 218.15 snails ( 13.47), and the density (SE) was 0.428 (

0.026) snails per m2.

Sawgrass-2 The percentage of recaptures peaked on the third occasion at 42 %

(Figure 9). The closure test indicated a violation of the closure assumption (z= -3.339, P <

0.001). The test of uniform density by rings (X2 =6.97,4 df, P =0.137) did not suggest a

violation of geographic closure. Snail density was uniform among rows (X2 = 14.81, 9 df, P=

0.096), but not columns (x2 = 25.37, 9 df, P=0.003).

Model SP (survival and capture probability constant and no transients) was the

appropriate model under TMSURVIV. It had the lowest AIC score of any of the models, the

goodness-of-fit tests indicated it was best (G = 48.24, 34 df, P = 0.054), and the LRTs

accepted SP over the more general models. Survival was less than 1.0 (Table 7), indicating

that the population was open demographically. For the sampling period, the population size

(SE) was 82.30 snails ( 15.01). The estimated density was 0.161 ( 0.029) snails per m2.

Discussion

We have demonstrated that the mark-recapture method can be applied to apple snails

to determine snail density. We found that interpreting the results of crayfish trapping studies

requires understanding the considerable variability in capture probabilities, and this can only

be achieved with a mark-recapture technique. It is apparent from our mark-recapture data

that snail population behavior is sufficiently complicated to preclude simple generalizations

about snail density or demography among habitat types or during certain phases of the snail

life cycle (e.g. early versus late in the breeding season).

We recommend that initial mark-recapture data analysis be based on the closed

population models contained in CAPTURE. They initially were chosen as a first step in the

analyses for several reasons. First, they tend to require less data than open models due to less

rigorous assumptions and fewer parameters (Otis et al. 1978). Secondly, the closed models

contained in CAPTURE provide information about sources of variation in the capture

probabilities (e.g., time, behavioral response, and heterogeneity among individuals).

CAPTURE also includes some tests of closure, which may or may not direct further analyses

to open models. If closure is not a problem, the results from CAPTURE include a density

estimate that takes into account edge effect. Closure may not be a problem for future apple

snail studies if investigators take into account some of the information we have collected

with regards to snail life history (see Chapter 6).

Despite efforts in this study to avoid expected periods of heavy adult mortality (see

Chapters 4 and 5 regarding survival), demographic closure was violated in all 7 grids due to

survival being less than 100%. Geographic closure violations, caused by snail movements in

and out of the grid, were also an issue in two of the grids. Closure violations resulted in our

use of open models (TMSURVIV) to generate estimates of snail densities. TMSURVIV

output indicated that the simplest open model, SP (survival and capture probability constant

and no transients), was appropriate for four of the seven sampling grids. Note that constant

survival does not refer to an absence of mortalities, but that mortalities occurred at a constant

rate. Model SPT (survival constant, capture probability variable, and no transients) fit only

sawgrass-1 in WCA3A. Transient snails (modeled in SPG, where survival and capture

probability are constant) were identified in only two of the grids, cattail-1 and cattail-2 of

WCA3A. We have no hypothesis as to why transient snails were only indicated for cattail

habitat. Cattail-1 was the first site sampled in WCA3A (initiated 12 March) and cattail-2

was sampled last (initiated 9 May), so time effects would not explain why transients were

only indicated for cattail. In addition, transients were not indicated for other sites sampled

where temporal overlap with the cattail sites did occur (prairie-1 in the case of cattail-1, and

prairie-2 for cattail-2). We question whether habitat structure would contribute to this

phenomenon, at least in terms of the ability of transient snails to penetrate denser habitat

structure. Our sawgrass sites appeared sufficiently similar in both stem density and

submerged vegetation abundance (e.g. periphyton or Utricularia sp., or both), at least within

a range likely to relate to snail movements; however, transients were not identified in

sawgrass sites. Some other aspect of the cattail sites, for example proximity to the canal,

may, in some unknown manner, be related to the transient issue.

Our capture-recapture studies provided some additional information (i.e.,

supplementary to the throw trap data) on habitat distribution among sawgrass and

prairie/slough, as well as cattail habitats. In Site 1, the highest snail densities were found in

prairie/slough and sawgrass, followed by cattail (Figure 10), but the differences were slight.

In Site 2, the highest snail densities were found in prairie/slough, but just slightly higher than

cattail, and sawgrass had the lowest density. These densities and differences between habitat

types are consistent with that of the throw trap; prairie sites appear more likely to have a

higher density of snails than sawgrass, but this relationship varies from site to site. Cattail

does not necessarily, as suggested in a previous study of gastropods in general (Davis 1994),

Figure 10. Snail densities (S.E.) in six locations in WCA3A in 1997 and one location in BCWMA in 1996
obtained from mark-recapture grids. Densities were calculated using population estimates from TMSURVIV
and dividing them by the grid area. CAT = cattail habitats, PR= prairie/slough habitats, and SAW = sawgrass
habitats. Numbers refer to site number.

0.5

0.45 WCA3A
0.4 ENBCWMA (plot)

0.35

0.3

S0.25

0.2

0.15

0.1

0.05

0 .I -
CAT-1 PR-1 SAW-1 CAT-2 PR-2 SAW-2 PR-1
Habitat-Site

suppress apple snail populations. We observed densities in cattail similar to those in prairie

habitats within both WCA3A sites. The uniform density tests from CAPTURE indicated that

snails were uniformly distributed (based on concentric ring, row and column tests) within

three of the grids, but not in 4 others. [These tests are not confounded by violations of

closure.] Our capture-recapture results confirm that snails are patchily distributed over large

scales even with similar vegetation types and hydroperiods, as evident from marked site to

site variations in snail density (Figures 4 and 10). Generalizations about vegetation type and

corresponding snail densities cannot be reliably inferred from our limited amount of data.

However, it is apparent from the site to site variability that considerably more effort will be

required to understand how vegetation type affects snail densities.

When comparing density values reported from our mark-recapture studies to other

published results, recall that TMSURVIV does not account for an edge width, which

increases the effective trapping area beyond the boundary defined by the traps. Therefore,

the densities reported here may be slightly higher than if edge effect were factored in (Otis et

al. 1978).

3.4 Egg Cluster Counts

Prior to this research, our knowledge of apple snail ecology from field studies was

based largely on observations of egg clusters (Odum 1957, Perry 1974, Hanning 1979,

Turner 1996). Unlike the cryptically colored submerged snails, egg clusters are conspicuous

in wetlands; female snails attach clusters of 20 to 30 white eggs to stems of emergent

macrophytes 2 to 200 cm above the water line (Hanning 1979, Turner 1996). Egg clusters

are easy to count along a transect (Hanning 1979) or within a throw trap unit which can be

thrown into the marsh (Bennetts et al. 1988). These characteristics make sampling egg

clusters as an index of snail abundance attractive. Counts of egg clusters have been

explicitly suggested (Perry 1974, Owre and Rich 1987, Bennetts et al. 1988) or implied via

loose application (Takekawa and Beissinger 1989) as an indicator of relative snail

abundance. We used snail densities determined by throw trap and mark-recapture grids to

evaluate the reliability of egg cluster counts as an index of apple snail abundance.

Methods

We examined the relationship between egg cluster counts and apple snail densities

using data from four studies. First, we used data from Bennetts et al. (1988 unpubl. data),

who counted egg clusters and estimated density at two sites in WCA3A during 1987. Their

egg cluster counts were conducted such that each sample represented 15 m2 using a 1m X 5

m polyvinylchloride (PVC) frame flipped end over end 3 times along the ecotonal edge

(Bennetts et al. 1988). Sampling was conducted along the ecotone created by juxtaposed

sawgrass and prairie/slough habitats. This is the habitat most frequently chosen by apple

snails for oviposition in graminoid marshes (Bennetts et al. 1988, Turner 1996). Their

estimates of apple snail density were derived using throw trap sampling with a suction

dredge in the adjacent prairie/slough habitat. Second, we counted egg clusters at four sites

each in WCA2B and western WCA3A during 1995 and 1996, respectively. Our counting

technique in these areas was virtually identical to that of Bennetts et. al., except that each 5

m2 throw trap was one sample. Apple snail densities were estimated using throw traps

extracted by suction dredge. Third, during our BCWMA study of reproductive ecology, we

repeated egg cluster counts at the same site six different times between February and August

1996. Again, we used 5 m2 sampling units. The density estimate, however, came from the

mark-recapture pilot study using the 100-trap trapping grid described in section 3.3. The

fourth source of egg cluster and snail density data came from our work in eastern WCA3A in

the spring of 1997. Again, egg clusters were sampled using 5 m2 sampling units, and

densities were obtained from the mark-recapture experiments.

For a comparison of egg cluster counts in relation to the sawgrass/prairie ecotone, we

used data from Bennetts et al. (1988) in which egg clusters were counted along the ecotone,

7.5 m and 15 m from the ecotone into the interior sawgrass. This evaluation was repeated in

this study in WCA3A in 1995, except that counts were conducted along the ecotone, and 5 m

and 10 m from the ecotone into the interior sawgrass.

Results

We found no reliable relationship between egg cluster counts and estimated apple

snail densities from a pooled sample from 5 separate studies (R2 =0.005, n=14 P=0.81)

(Figure 11). However, we found several potential sources of variation in egg cluster counts

that could influence this result. Repeated sampling at the same site revealed a strong

seasonal pattern in the number of egg clusters, with a peak occurring during April or May

(Figure 12). A pattern also was observed in relation to the prairie/sawgrass ecotone. Marsh

habitats throughout South Florida consist of a mosaic of open prairie or slough habitats

interspersed among stands of sawgrass or cattail. Our data indicated higher numbers of egg

clusters along the ecotone compared to even just a few meters toward the sawgrass interior

(Figure 13).

Figure 11. Egg cluster counts and estimated snail densities from a pooled sample from each of 5 studies. Snail

density estimates were derived from throw trap sampling using a suction dredge in prairie/slough habitat

(Bennetts et al. 1988, WCA3A 1996, WCA2B 1995) and from mark-recapture grids (BCWMA 1996, WCA3A

1997).

3.5

3.0

2.5

2.0

1.5

1.0 4-

0 0.1 0.2

0.3 0.4 0.5
snails I m2

0.6 0.7 0.8 0.9

BCWMA (1996)
o WCA3A (1997)
a Bennetts et al. (1988)
*WCA3A (1996)
SWCA2B (1995)

*
0 0
o

i I ~ ~i !

I r

Figure 12. Mean (SE) number of egg clusters/ 5-m2 sampled at one site in BCWMA 6 times during 1996.
6!

T
4

0
2

0~ ~ ~ L --r --- --- --- -- ---- -- --- -- --- -- --
FEB MAR APR MAY JlJ AUG

Figure 13. Mean (SE) number of egg clusters sampled at 3 distances relative to the sawgrass/prairie ecotone.

Samples from Bennetts et al. (1988) (n=206 for each distance class) were conducted during 1987 and samples

from WCA3A (this study) (n=71 for each distance class) were conducted in 1996.

6

5 (Bennetts et al. 1988)

,4

0

02

1

0
Ecotone Mid Deep
(Om) (7.5m) (15m)

2. (WCA3A 1996)

E
S 1.5

Ecotone Mid Deep
(Om) (7.5m) (15m)
Distance from Ecotone

We examined the precision of our egg cluster estimates from the March through July

clusters counts from BCWMA East in 1996. This effort indicted a relatively stable

coefficient of variation of 10-20% after sample sizes (i.e., number of throw traps) reached 9

(Figure 14). Most of the data used in our analysis had sample sizes above this threshold.

The two sites from Bennetts et al. (1988) had a sample size of 26 and 28. In our study from

BCWMA East, the sample size was 12 throw traps. The data used from our mark-recapture

study in eastern WCA3A in 1997 consisted of 10 throw traps. However, the data from

western WCA3A in 1996 had sample sizes of only 3 per site.

Figure 14. Change in coefficient of variation of egg cluster counts with increasing sample sizes. Each sample
represents the number of egg clusters counted within a 1 m x 5 m PVC frame. Data from surveys in March,
April, May and June/July in BCWMA in 1996.
100

80

Co
2o
60

20 -

0

3 6 9 12 15 18 21 24 27 30 33 36
Number of 5-m2 Samples

Discussion

Our counts of ecotone egg clusters did not correlate to snail densities in adjacent

prairies. Bennetts et al. (1988) also found no relationship between counts of egg clusters and

capture rates of foraging snail kites. We believe that these results are due to high variability

and factors affecting oviposition of apple snails. Apple snail oviposition is influenced by

many factors including temperature and vegetation (Hanning 1979, Turner 1996). Our

results also indicated that egg laying is quite seasonal and the majority of eggs are deposited

over a period of 8 to 12 weeks (Figure 12), which is consistent with reports from Odum

(1957) and Hanning (1979). However, even sampling two sites simultaneously may not

eliminate the problem. Hanning (1979) found spatial variation in peak egg laying among 6

transects in the southwestern littoral zone of Lake Okeechobee. We would also expect

differences to occur along a latitudinal gradient due to the effects of temperature (Hanning

1979). We also found variation attributable to where in relation to the ecotone egg clusters

are sampled, which concurs with observations of Turner (1996) and Bennetts et al. (1988).

Thus, although it is possible that egg cluster counts could provide meaningful results for

studies in which these sources of variation are carefully controlled, our results do not support

the use of egg cluster counts as a reliable method of quantifying snail abundance.

3.5 Conclusions

Obtaining reliable estimates of apple snail density, regardless of the method, will be

time and labor intensive. The Florida apple snail, although the largest aquatic gastropod in

North America, is a relatively small, inconspicuous animal that occurs in relatively low

densities (compared to other invertebrates) in densely vegetated wetlands.

We evaluated one method of assessing snail abundance, egg cluster counts, which did

not require extracting snails from their environment. This is the least labor intensive and

least time consuming method we evaluated; for this reason it has great appeal as an index of

snail abundance. However, we found no support for use of egg cluster counts as a reliable

index of apple snail abundance.

Of the throw trap-based methods, the dip net and suction dredge were similar in

performance. The suction dredge appeared a little less sensitive to habitat differences and

tended to have slightly higher overall recovery probabilities. However, the dip net required

less effort and may require less initial investment. In contrast to these two extraction

methods, the bar seine had a lower overall capture probability and was substantially more

affected by habitat type. Consequently, if a throw trap based method is to be used, we

encourage use of either the dip net or suction dredge. The data for recovery of marked snails

indicates that it is imperative to assess efficiency of extraction when using throw traps. We

agree with previous authors (e.g., Burnham 1981, Nichols 1992) that counts of animals,

whether they be from a throw trap or otherwise, are of questionable value without having an

estimate of the proportion of animals being counted. Recovery studies add some effort to an

already labor-intensive approach to sampling, but without information on extraction

efficiencies, comparisons between sampling areas can not be performed with reliability.

The use of crayfish traps or trap arrays as an index of relative abundance may be

appropriate in some situations, but great care must be taken to control for time effects and

site to site variation that might affect capture probabilities. For example, changes in weather

pattern (e.g., dropping temperature, especially in shallow water) or mating behavior may

affect snail movements, and therefore the likelihood of capture. Site variation may include

the amount of submerged vegetation (e.g., Utricularia spp.) or flocculent substrate which

may hinder movements. If capture probability is not directly measured (e.g., a mark-

recapture regime), then these differences in capture probabilities may be inaccurately

interpreted as a difference in snail abundance. If issues such as changing weather or habitat

structure cannot be controlled for (e.g., by sampling all sites simultaneously, and by sampling

in similar vegetation types), then an estimate of relative abundance using crayfish traps may

not be reliable.

Trap arrays performed more reliably than crayfish traps with respect to their

validation with throw trap data, but trap arrays are not available commercially and are not

suitable for use in densely vegetated habitats. Our analysis of the relationship between

crayfish traps and snail densities derived from throw traps involved a rather low sample size

over a narrow range of snail densities. Given the relative ease and utility of crayfish traps in

a variety of habitat types (e.g., sawgrass, wet prairie, cattail), additional effort to validate

their use as an index of snail abundance is warranted.

We believe that the use of crayfish traps within a mark-recapture sampling regime

has the greatest potential to provide reliable estimates of apple snail densities. Even though

we found that closure could not be assumed for any of the mark-recapture populations, more

recently developed software packages (e.g. TMSURVIV) permit survival and population

estimates to be obtained for open populations. Mark-recapture data provides not only

population density information, but also information on survival, movements, and behavior

of snails in the population. Our mark-recapture studies also revealed that capture

probabilities may vary markedly over relatively short time intervals and from grid to grid

within the same wetland system, thereby providing our justification for exercising caution

with regards to crayfish traps as a tool for determining relative abundance.

We recognize that each sampling regime, throw trapping for snail density, movement-

based traps for relative abundance, and capture-recapture for density, has its unique

advantages and limitations. Amount of effort and expense is a limiting factor in any

monitoring program or research effort. In this case, capital costs may counterbalance labor

costs (i.e., a set of 100 crayfish traps are more expensive, approximately $1600, but require

less overall labor than throw traps). Amount of effort also varies depending on the snail

densities encountered. For example, in areas with high snail densities, fewer throw traps are

necessary to obtain good precision for the snail density estimate. In contrast, higher snail

densities may (depending on capture probabilities) result in more snails captured in trap

grids, and marking and releasing hundreds of snails in one occasion is time consuming

(estimated at 4 hours per 100 snails marked by two workers). Regardless of the investigators

choice of sampling method, it is critical that effects of capture probability on survey results

(or recovery as applied to throw traps) be understood and explicitly stated in the

interpretation.

4.0 FIELD STUDIES OF MOVEMENTS AND SURVIVAL

Seasonal variations in rainfall leading to drying events are common among the

tropical and subtropical wetland habitats occupied by Pilidae species, including the Florida

apple snail (Little 1968, Burky et al. 1972, Kushlan 1975, Haniffa 1978a). However, the

natural hydrology in which the Florida apple snail adapted has been altered considerably

following the installation of canals and water control structures; this includes the three

largest graminoid marsh systems, the Kissimmee Marsh, Upper St. Johns Marsh, and the

Everglades (Lowe 1983, Light and Dineen 1994). Successful restoration of these areas,

currently underway, requires implementation of a hydrologic regime in support of the biota

which historically flourished in them. As increased water demands for agricultural,

industrial, and residential areas have diverted water from their natural course, marsh dry

downs have become more frequent and longer in duration. Earlier research indicated that

these dry downs suppress apple snail populations (Kushlan 1975), and that this may

contribute substantially to snail kite population declines (Takewawa and Beissenger 1989,

Beissenger 1995). Balancing the needs of the human population with that of wetland biota

requires understanding how marsh dry downs affect wetland inhabitants, including the apple

snail, a critical component of wetland food webs in Florida (see Introduction).

Understanding the snail's response to the duration and timing of dry downs will contribute

substantially to the data base from which decisions on the effective allocation of water by the

water management districts can be derived.

Any subtropical or tropical aquatic snail challenged with declining water levels must

contend with higher temperatures and declining dissolved oxygen (DO) in residual water

supplies, followed by desiccation and overheating once water levels drop to ground level

(Haniffa 1978a, Aldridge 1983). Survival strategies involve either burrowing into the

substrate or remaining in residual pools with tolerable conditions (Burky et al. 1972, Haniffa

1978a). Pomacea and Pila snails can aestivate from several months to over a year in dry

conditions (Little 1968, Burky et al. 1972, Haniffa 1978a, Haniffa 1978b). We have not

found any study which documents Pilid snails moving to deep water refugia during a drying

event.

The results presented in this section reveal the movement patterns and survival rates

of Pomaceapaludosa as a function of declining water levels in two contexts. First, we

studied snail movements and survival during a drying event in a graminoid marsh late in the

dry season. Second, we studied snail movements and survival during a lake restoration dry

down which occurred in winter. We also were able to reveal a relationship between

movements and survival as a function of the ebb and flow of seasonal snail reproductive

activity, which spans the hydrologic transition from dry season to wet season.

4.1 Methods

Telemetry Technique

Making inferences about apple snail responses to declining water levels required

monitoring snails frequently enough to keep pace with changing hydrology (e.g., weekly or

biweekly). We selected miniature radio transmitters as the tool for monitoring apple snails

in the field. Radio-telemetry was used to examine patterns of movements correlated with

habitat conditions, to locate stranded snails in dry marsh, and to find snails in deep water

(tested up to 2 meters deep). Telemetry permitted repeated snail location without labor-

intensive sampling and with the greatest reliability of any other method available. The

habitat was not disturbed in the process, except for gaining access to the snail by boat or on

foot. Before releasing snails into the field, we made preliminary observations on nine snails,

weighing approximately 14-28 grams each, to assess behavior while wearing a transmitter.

Snails with transmitters were observed in an aquarium in each of the following behaviors:

crawling on the bottom, climbing on aquarium sides or vegetation, feeding, breathing air,

burrowing (near the substrate surface) in loose sand and peat, mating, laying eggs, and

floating freely on the water surface. We saw no evidence that snail behavior was

compromised by the transmitter.

We attached 1.6 gram transmitters (ATS, Inc., Isanti, MN) to the outside of the snail

shell using the minimum amount of marine epoxy required for a firm hold. The area for

attachment was towel dried and lightly sanded prior to epoxy application. We placed the

transmitter 1-2 cm up from the aperture. Placement at the apex allows the snail to remain in

an upright position when withdrawn in its shell. Approximately one-half of the transmitters

were equipped with a 12-cm whip antenna positioned to trail behind the snail as it crawled.

The other transmitters had an antenna coiled and encapsulated in the same protective resin

which coats the circuitry and battery. We received signals up to 200 meters from

transmitters with whip antennae (100 meters with encapsulated antennae) submerged in

graminoid marsh.

The transmitter can readily be located within an approximate 2 to 3- meter diameter

area, but quantifying snail response to habitat conditions (e.g., D.O. or water depth)

necessitated obtaining a more precise location. We also anticipated needing to locate snails

in thick vegetation or buried in substrate. We found that a magnet can be used to locate

transmitters precisely. When a magnet (10 cm x 3 cm x 2 cm, 25-kg pull) touched or came

within 13 cm of the transmitter body or antennae, the pulsing signal was turned off or

interrupted, or the pitch of the signal changed. The magnet altered the signal under water,

when the transmitter was buried in sand, or when buried in sediment under water. Through

use of the magnet probe, we reduced transmitter retrieval time from one to several hours (our

experience without a probe) to typically 10-15 minutes.

Tracking snails during the marsh dry down and lake management draw down

required monitoring for several months. The maximum battery life of the 1.6 gram

transmitters used in this study was 60 days. Therefore, apple snail movements for the course

of the dry season were documented by releasing transmitters in a staggered fashion.

Functioning transmitters from dead snails were transferred to newly captured snails to

increase sample size.

Movements During Drying Events

Blue Cypress Water Management Area

We selected our study site in the eastern-most portion of the BCWMA East (Figure

1), since it had the highest ground elevation and was therefore most likely to dry out. We

also monitored snails in BCWMA West to compare and contrast potential differences in

movements and survival based on hydrology, substrate and vegetation type.

We used an airboat to access the study areas. Apple snails were collected

opportunistically during daytime searches in clear water or by spotlight at night. We also

collected snails which were found mating with snails bearing transmitters. We had just

begun deploying crayfish traps (see Chapter 3), so only a few snails were obtained via

trapping.

Once collected, snails were sexed (based on shell morphology; Hanning 1979),

weighed (using a spring scale), and their length measured (using a vernier caliper) prior to

transmitter placement. Each snail was returned precisely to the spot from which it was

taken. These and subsequent snail locations were marked with a pvc pole or flag bearing the

snail's unique identification number.

At the time of collection and subsequent relocation, the following parameters and

habitat conditions were measured:

1) Distance and direction from previous location

2) Water depth at current and previous location

3) Water temperature at current and previous location (mercury thermometer)

4) Dissolved oxygen (D.O.) at current and previous location (YSI 57 D.O. meter)

5) Substrate and plant composition

6) Depth of aperture of shell, if buried

7) Temperature of the sediment (if exposed)

Apple snail movements were examined on three temporal scales. Most movements

were documented at approximately 5-8 day intervals (n= 98) until the snail or the transmitter

battery died. For daily movements, an individual snail's location was documented two times

over a 24 hr interval (e.g., 5 a.m. and 5 p.m.) (n= 48). These data exclude initial snail

locations and locations of snails found dead. We also monitored snail position over a 12

hour interval, including one location at night, to see if snail position was affected by the

diurnal cycle (n= 20).

Lake Kissimmee Draw Down

We monitored apple snail movements in response to draw down of Lake Kissimmee

which occurred over 18 weeks beginning in November 1995. The draw down, part of a lake

restoration project to improve fisheries habitat (GFC 1995), resulted in an approximately 1.7

meter water level drop (from a high water benchmark of 16.7 meter MSL). The telemetry

procedure used to monitor snail movements and survival was similar to that previously

described for BCWMA. We decided not to measure DO, a decision based on the lack of

impact of DO levels on movements observed during the BCWMA study.

Apple snail movements were monitored until the transmitter battery failed or until the

snail was found dead Throughout the 18 week study period, snails were checked at 7 to 11

day intervals in most cases (8 of 11 site visits); the remaining three intervals were 14, 17, and

17 days.

We selected our study site based on substrate and vegetation type, water level, and

boat traffic. We wanted to avoid muck (i.e., flocculent, not fibrous) substrates, floating

islands, and areas of high boat traffic (e.g., the vicinity of boat ramps). We chose the north

end of Brahma Island as our telemetry study site. An airboat was used to access the area.

We collected snails by wading and by using trap arrays. Snails were used only from trap

arrays which were in water over 40 cm (the height of the funnel entrances, which tapered

down to 6 cm) at the time of snail collection. At the time of collection and subsequent

relocation, distance and direction from previous location, and water depth at current and

previous locations were measured. Water temperature was monitored at seven stations

marked by pvc poles, which included the entire range of depths for all snail locations. As the

dry down proceeded, the stations closest to shore could not be used for water temperatures,

but they were used to monitor the exposed substrate temperature. Movements and gradient

data were derived as described for BCWMA.

Survival

Telemetry data from BCWMA and Lake Kissimmee were used to calculate survival

for each study population. Each time a snail was located we checked for mortality. If no

signs of activity were observed with an intact snail, we gently pushed on the operculum to

see if there was resistance. For dead snails this often resulted in breaking the operculum seal

and exposing dead flesh. If we were unsure of the snail's status we carefully inserted a knife

(8 mm blade width) between the operculum and the shell and gently pried it open in order to

see the snail flesh. This was not done with excessive force, since we did not want to damage

the operculum. If we could not pry the operculun open, we left the snail for the next week.

Our experience shows that the operculum, accompanied by snail flesh, is very easily

removed from a dead snail. If an empty shell was found, evidence of predation was noted

(Snyder and Snyder 1969). The two primary signs of predation were 1) location in an

obvious snail kite or limpkin shell pile, and 2) finding that a previously stranded snail had

been extracted (by a predator) from the substrate.

Trapping Study and Egg Cluster Survey in BCWMA East

We supplemented our telemetry work with a snail trapping study to enhance our

understanding of apple snail movements and survival in relation not only to drying events,

but also to coincidentally occurring snail reproductive activity. The trap study was

conducted in the BCWMA East in the same area that we monitored snail movements with

transmitters during spring 1995. The crayfish trapping study was conducted winter through

summer, 1996.

Preliminary observations lead us to two important points regarding the use of crayfish

traps. First, bait is not needed to lure snails into the traps; snails apparently enter the trap

funnels during horizontal movements and/or vertical ascent to breathe air or lay eggs (section

3.2). Second, sex ratios of captured snails suggested that males were lured by females that

had crawled into the traps; whenever we found a female, one to six males were also in the

trap. We hypothesized that as reproductive activity increases, the male to female ratio of

snails captured in traps would also increase. Since the snails must move to the traps, and no

bait attractant is provided, increased movements (whether driven by mating, temperature,

hydrology, etc.) should result in increased captures. If movements are based on males

tracking females, we would expect the M:F ratio to be affected by variation in mating

behavior. If the sex ratio of captured snails varied with some indicator of reproductive

activity, then we would conclude that at least during certain times of the year, movement

patterns change as a function of changing reproductive activity.

The basic trap unit was the crayfish trap (section 3.2). For this study we also wanted

to test whether or not baiting traps with a live conspecific would attract more snails. We

constructed an 8-cm diameter cylindrical enclosure which fit inside the crayfish trap and

permitted the bait (snail) to reach the surface to breathe air. We used from 51 to 54 traps

throughout our 7-month study. A cylindrical bait enclosure was installed in all traps.

Approximately one-third of the traps were baited with adult females (F), a second third of the

traps with adult males (M), and the remaining traps were not baited to serve as controls (0).

Traps were numbered, and the same type of bait was used for each trap each time we set

traps; this was done in case a chemical left by the bait attracted other snails. Six pvc poles

were placed in the study site to monitor temperature and depth throughout the course of our

study. Depth was measured using a one meter rule placed within 10 cm of the pole. These

markers were left in place for the entire seven months of the study.

We conducted six trapping sessions from 30 January to 19 August 1996. A trapping

session was initiated by placing 51 to 54 traps on pvc poles, approximately 5 meters apart,

throughout our study site. On the first occasion, the traps were randomly assigned locations

in order to randomize the distribution of type of bait (M, F, or O). The pvc poles which

support the traps were distributed at the same locations for each trapping session (we made a

map using temperature/depth monitoring stations and vegetation as landmarks). However, at

the initiation of each of the six trapping sessions, individual traps (and therefore bait type)

were randomly distributed among those pvc poles to avoid any potential interactions between

bait type and trap location.

For each trapping session, traps were checked on two occasions. Traps were not

moved between checks. During 12 trap occasions (2 occasions for each of six sessions) we

checked traps at 3-day intervals five times and 4-day intervals four times; the remaining

occasions were at intervals of 5, 8 and 9 days. Snails found in traps were released

approximately 2 meters from the trap (half the distance between traps). During each

occasion, the shell lengths of at least 20 males and 20 females were measured using vernier

calipers.

Egg cluster production was used as our index of reproductive activity over time. We

deployed a 1 x 5 meter pvc quadrat 36 times during each trap session. We sampled the same

transects during each session to control for variation due to sampling location and possibly

due to individual female egg production.

Analyses

Movements During Drying Events

Weekly apple snail movements were examined as a function of sex, time, and water

depth and temperature. We grouped data by biweekly intervals to increase our sample size

for the analyses. If an individual snail's movements were measured twice within a class

interval, the mean value for the distances traveled was used for that individual. Distances

traveled (in meters) were transformed using the function logo (meters +1) in order to meet

assumptions of normality for the analyses of variance. We used a mixed model ANOVA

(snails monitored = random effect; week monitored = fixed effect) for all movement

analyses (movements as a function of time and as a function of depth). Since some

individuals were monitored across several biweekly intervals, each interval included

different sets of individuals. We performed a repeated measures ANOVA (Crowder and

Hand 1990, SAS Inc. 1992) to account for repeated measure of some individuals across

intervals (model logmeter= temperature + sex + time + interactions). The analysis produces

F-statistics that are not a ratio of sums of squares, but are instead from a Wald-test. Sums of

squares (SS) and mean squares (MS) are therefore not reported for those analyses. For more

information on this type of F-statistic see Searle et al. (1992). The same data were analyzed

for distance traveled as a function of depth (depth in which the snail was last found) (model

logmeter = sex + depth + interactions), to test if snails make larger movements to avoid

being stranded when the water level reaches some critical depth. For this and subsequent

analyses (see below), depth was divided into categories of 10 cm intervals.

Movements along gradients of depth, dissolved oxygen and temperature were

analyzed differently. Gradients were calculated based on measurements taken

simultaneously at two consecutive snail locations (Figure 15). Note that gradients refer to

differences between two consecutive snail locations, so that the scale along which the

gradient occurred depends on how far the snail traveled and where the snail was located in

the marsh. Considerable topographic variation occurred within our study area (i.e., wet

prairies adjacent to canals, and areas that eventually went dry adjacent to inundated sloughs).

A mosaic of juxtaposed vegetation types (i.e., sawgrass, Eleocharis sloughs, Panicum wet

prairies) also ensured availability of temperature and dissolved oxygen gradients within the

areas in which snails moved. We were primarily interested in whether or not a snail moved

Figure 15. An example of how gradient data were derived from telemetry data. A snail is located on
May 24 and its location marked with a pvc pole. On May 31, we found the snail at its new location.
We measured depth 1 and depth 2 simultaneously on May 31.

May 24

May 31

DEPTH 1 DEPTH 2
on May 31 on May 31

GRADIENT = DEPTH 2 DEPTH 1

along a gradient (i.e., towards deeper water), not the actual value of the gradient. Water

depth gradient values ranged from -90 cm to + 76 cm. If we had calculated the mean

gradient experienced by a group of snails within some interval, one large change in water

depth for one snail, for example -50 cm, would negate the weight of five snails which moved

along a + 10 cm depth gradient. We therefore tested whether or not snails move along a

positive depth gradient by scoring each individual movement as positive (P) or nonpositive

(NP) (which included zero and negative gradients). The number ofP and NP for each class

interval were added, and a frequency table of the proportions of P and NP was generated.

The association between gradients and either time or depth was tested using the Mantel-

Haenszel chi-square statistic (SAS Inc. 1988, Mantel and Haenszel 1959). We analyzed the

depth gradient both as a function of time (biweekly interval) and previous depth. Due to the

low range of values for temperature and dissolved oxygen, we could not divide the data into

classes of temperature or DO. Temperature and DO gradients were analyzed only as a

function of time as described for depth gradients. For temperature, we analyzed the negative

gradients relative to non-negative gradients (zero + positive) as well as P vs. NP, since we

were also interested in seeing if snails moved to cooler water as water temperature increased.

Survival

Survival of BCWMA and Lake Kissimmee snails with transmitters was estimated at

weekly intervals. The Kaplan-Meier procedure to accommodate staggered release of

transmitters was used for the survival analyses (Pollock et al. 1989). Chi-square tests as

described by Pollock et al. (1989) were used to compare rates of survival between males and

females.

Movements and Population Dynamics Related to Snail Reproductive Activity

We examined the effect of bait type (M, F or 0) using two separate analyses. In the

first, we looked only at the number of males captured as a function of bait type (males= bait

+ session + session*bait). In the second analysis we looked only at females captured

(females= bait + session + bait*session). We defined the number of snails captured per trap

(male or female) as the cumulative data from both trap checks within each trapping session.

We also analyzed snail size as a function of trapping session using ANOVA (size= sess + sex

+ sess*sx).

Male to female ratios were calculated for each individual trap check. We were only

interested in whether males were lured into the traps by females that had crawled into the

traps. Therefore, traps with no captured snails were not included in the M:F ratio data set.

The resultant data also excluded trap checks which contained no females (which would be

division by zero in M:F ratio). The sample size (number of M:F ratios) obtained in this way

varied from 33 to 54 among the six trapping sessions. We analyzed M:F as a function of

time using ANOVA (M:F = trap session). The relationship between M:F ratio and

reproductive activity, as measured by egg clusters, was evaluated using a linear regression

(M:F = eggs).

4.2 Results

Movements During Drying Events

Blue Cypress Water Management Area

We monitored 51 snails with transmitters in BCWMA East and 7 snails with

transmitters in BCWMA West. One of the first major findings during our telemetry surveys

were the weekly distances traveled by snails. Figure 16 presents a histogram of the weekly

distances traveled by snails in BCWMA 1995 (we also included Lake Kissimmee data, which

were similar to those obtained in BCWMA). Thirty-eight percent of the 176 weekly

movements documented were more than 10 meters. The greatest distance measured for one

week of travel was 82.5 meters.

Figure 16. Distances traveled by snails in BCWMA (Spring 1995) and on Lake Kissimmee (Winter 1995-
1996). Data only for snails in water depths > 10 cm.
90

80

650

.. 40o

30

20

10

0
5 10 20 30 40 50 60 70 80 90
meters

MILLISECOND	CLASS.METHOD	MESSAGE
0	sobekcm_page_globals.constructor
0	sobekcm_page_globals.constructor	Application State validated or built
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.constructor	Navigation Object created from URI query string
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.display_item	Retrieving item or group information
0	sobekcm_page_globals.get_entire_collection_hierarchy	Retrieving hierarchy information
0	sobekcm_assistant.get_entire_collection_hierarchy
0	cached_data_manager.retrieve_item_aggregation
0	cached_data_manager.retrieve_item_aggregation	Found item aggregation on local cache
0	item_aggregation_builder.get_item_aggregation	Found 'all' item aggregation in cache
0	system.web.ui.page.page_load (ufdc.page_load)
0	sobekcm_page_globals.constructor.on_page_load
0	html_echo_mainwriter.add_style_references	Adding style references to HTML
0	html_echo_mainwriter.add_text_to_page	Reading the text from the file and echoing back to the output stream
97	html_echo_mainwriter.add_text_to_page	Finished reading and writing the file