Modeling Manatee Response to Restoration in the Everglades and Ten Thousand Islands
Bradley M. Stith, Jim Reid, Dean Easton, and Susan Butler
USGS, Center for Aquatic Resources Studies, Sirenia Project, Gainesville, FL
The coastal waters of southwest Florida harbor nearly a third of Florida’s population of
the Federally listed West Indian manatee (Trichechus manatus). A large proportion of
this population occurs within the Everglades National Park (ENP) and the Ten Thousand
Islands (TTI), yet only limited information is available from this remote region. Aerial
surveys conducted by ENP personnel in the early 1990s revealed heavy use of both
inshore and offshore areas during most of the year. Manatees feed almost exclusively on
submerged aquatic vegetation (SAV), and most individuals show a strong preference for
marine seagrasses. Because manatees lack the ability to drink saltwater (mariposia), they
need to drink freshwater periodically. Thus, manatees depend on resources in marine,
estuarine, and freshwater zones, making them excellent indicators of the health and
restoration success across all of these zones.
A primary goal of this research is to develop an individual-based ATLSS model to
simulate manatee response to changes in hydrology caused by the Comprehensive
Everglades Restoration Plan (CERP). In support of this goal we have analyzed manatee
telemetry data from the study area to parameterize and calibrate the model. We have also
conducted aerial surveys in the TTI region to provide information on the distribution and
abundance of manatees that could be used to validate the model.
Analysis of telemetry data
Data was analyzed from 20 tagged animals between June 2000 and July 2002,
representing 4,563 manatee tracking days. The attached Argos transmitters were
programmed to fix at least 1 location during 4 time-windows each day, when satellite
overpass geometry was optimal. The actual number and accuracy of fixes obtained
varied due to animal behavior or equipment performance. Filtering out all points with
poor spatial accuracy resulted in 12,600 fixes, a decrease of 31% over what could be
obtained under optimal conditions. We also deployed GPS tags for 104 manatee days;
these tags were programmed to obtain fixes at 15- or 20-minute intervals.
The Argos telemetry data provided valuable information on coarse-scale patterns of
manatee behavior that were incorporated into the ATLSS model. We classified the
landscape into 4 simple aquatic zones: offshore, travel corridors, inshore bays, and
inland riverine systems or canals. We used a GIS overlay analysis to determine the
aquatic zone for each telemetry point. Most manatees showed a consistent pattern of
feeding on marine seagrass beds in offshore zones for a period of several days, followed
by large movements of 5 to 20 km or more up rivers and canals, presumably to obtain
freshwater. Several animals also made heavy use of inshore bays where they could feed
on a suite of SAV different from the offshore areas. The home ranges of all animals
incorporated one or more inland sites which supply freshwater, as well as offshore areas
that provide food resources.
Comparison of the number of inland forays made during wet versus dry season showed a
slightly higher number during the dry season, but the difference was not significant.
Manatees made inland forays in the wet season, even when freshwater was available at
the mouth of rivers or canals. During periods of significant cold weather, manatees
greatly decreased their use of offshore areas and increased their use of inshore locations
with favorable thermal buffering (typically deeper sections of canals, rivers, and inshore
holes).
GPS data provided valuable fine-scale information on manatee behavior, which could not
be obtained from the Argos data. Distribution of movement rates between GPS points
fitted an inverse distribution, with a mean and mode well below 1 km/hour, and
maximum speeds approaching 3 km/hour. Directionality of movement was highly biased
towards small turn angles, and the GPS data provided precise movement pathways
through the complex aquatic landscape in TTI and ENP.
The telemetry data provided valuable information needed to parameterize and structure
the ATLSS model. The broad movement patterns provided by the Argos data were
nicely complemented by the detailed data provided by the GPS tags.
Model structure
The manatee ATLSS model is being developed in C++ using object-oriented techniques.
The model is individual-based, spatially-explicit, and simulates the movements of
individuals on a raster image (20 m cell size) of southwest Florida. A network data
structure of arcs and nodes is used to direct the movement of manatees in an efficient
manner. This network structure was developed from telemetry locations and aerial
survey data, and consists of nodes representing primary drinking areas, feeding areas, and
thermal sheltering areas, connected by travel paths represented as arcs. Standard
algorithms from graph theory are used to access and query this network structure.
Home range allocation – Each manatee is allocated a portion of the total network that
includes one or more freshwater sites and offshore seagrass beds. The initial network
portion allocated to each manatee at the start of a simulation is drawn from a distribution
of home range sizes and geographic positions along the coast as determined from
telemetry data. Manatees born after the creation of the initial cohort inherit their
mother’s network, to reflect maternal transmission of home range.
Movement rules - Manatees move on a network of nodes representing destination sites
for feeding, drinking, and thermal sheltering, all connected by arcs representing travel
corridors. Low water depths limit the movement of manatees along some portions of the
network. Manatees can shift their home range to different parts of the total network if
freshwater or seagrass becomes unavailable within their subset of the network.
Incorporation of environmental variables - Salinities, temperature, and water depth are
modeled only along this network, rather than across the entire grid, to increase
computational efficiency. Until linkages exist to hydrologic models such as TIME, we
are relying on simple surrogate models that simulate observed patterns and possible
restoration scenarios.
Manatee behavioral state switching - A Markov Chain approach is used to simulate the
transition of manatees into different behavioral states that drive the movement patterns of
each individual. Only a few, simple behavioral states (feeding, drinking, traveling,
resting) inferred from the telemetry data are modeled. Transition probabilities were
developed from the telemetry data and from previous research on manatee time budgets
in other areas.
Manatee learning – Several simple learning modules simulate freshwater site switching
by manatees. These modules determine how quickly manatees shift their use of different
parts of the network in response to positive or negative reinforcement in the availability
of freshwater at sites within the home range of each manatee. The simplest algorithm, the
real-time linear operator model, assumes that animals maintain an estimate of resource
availability at time t, which is incremented or decremented during each time step,
depending on whether the resource was found or not.
Key assumptions – A key component of the model is the manatee-learning module,
which determines how quickly manatees shift their use of different parts of the network
in response to positive or negative reinforcement in the availability of a critical resource.
Sensitivity analyses will be used to evaluate the importance of different assumptions and
uncertainty associated with poorly understood parameters. As additional telemetry data
are collected, the model will be refined to incorporate new insights from these data.
Radiotracking and aerial surveys will provide an important means of monitoring manatee
response to natural environmental fluctuations and human-induced alterations associated
with restoration activities.
Brad Stith, USGS, Center for Aquatic Resources Studies, Sirenia Project, 412 NE 16th
Ave., Room 250, Gainesville, FL 32601.
Phone: (352)333-3814, fax: (352)374-8080, bradley_stith@usgs.gov.