Richard O. Flamm, Brad L. Weigle, I. Elizabeth Wright, Monica Ross, and Sherry Aglietti. 2005. Estimation of manatee (Trichechus manatus latirostris) places and movement corridors using telemetry data. Ecological Applications 15:1415–1426.

Appendix A. Details of the method used to generate the manatee movement paths.

Derivation of shoreline map.--The Florida coastline map was a GIS coverage in vector format that was digitized from a mosaic of 1:40,000 NOAA navigation charts and 1:24,000 USGS quadrangle maps (Flamm et al. 2000).  Areas were labeled as either land or water.  We used ARC/INFO’s GRID module to convert this vector coverage to a raster map having 25 × 25 m cells.  The conversion from the vector map to the grid resulted in some peninsulas being separated from land and streams being closed off at their narrows.  To restore the lost features of the shoreline, we developed a browsing and raster editing tool using ARC/INFO’s Macro Language (AML).  The browsing tool allowed us to scan a display of the rasterized shoreline that was overlaid by its vector representation.  Using the application’s raster editing component, cells that actually represented streams but had been mapped by the polygon-to-grid transformation as land were changed to water.  Peninsulas were reconnected to the mainland by converting the necessary pixels to land.  This editing was necessary because telemetry points located upstream, beyond narrows in the waterway, might be isolated from the Gulf of Mexico, and the separation of peninsulas from the mainland would allow the movement-path algorithm to delineate paths across areas where land would have prevented manatee travel.

Transformation of the bathymetry map to a cost surface.--Vector coverages of bathymetry extending from the Florida Panhandle to the Florida Keys and updated through 1997 were used. The coverages were joined into a single map, projected from ALBERS (NAD83), in double precision to UTM projection (NAD27), converted to a 25 × 25 m raster map, rectified to the rasterized Florida shoreline map, and then transformed to a cost surface for the movement-path model. A cost surface is a raster map where cell values represent the "cost" associated with the passing of an object, in this case a manatee, through a cell.  In a cost-path analysis, the least costly path between two points is delineated.

Costs were derived in a two-step process: first, we conducted quantitative analysis to generate starting values for the costs and second, we instituted an iterative process of evaluating the costs and then refining them based on outputs from the movement-path model and expert opinion.  The quantitative analysis involved comparing a distribution of water depths associated with satellite location-class-3 telemetry points, the most accurate, which served as the observed distribution, to an expected distribution.  The expected water-depth distribution was calculated by multiplying the proportion of cells in each bathymetry class found in Tampa Bay by the number of class 3 points used to create the observed distribution.  Expected values were based on the assumption that manatees demonstrate no bathymetric preferences.    In effect, this analysis estimated the strength of the manatee’s preference or avoidance of a bathymetry class by measuring the deviation between the observed and the expected.  This deviation was calculated for each bathymetry class by dividing the larger cell count value, whether that of the observed or expected count, by the smaller.  An observed count exceeding the expected suggested preference for that bathymetry contour and the resulting quotient was given a positive sign.  An expected count larger than the observed suggested avoidance or no interest, and the quotient was negative.  The results were sorted from largest to smallest and rescaled to reflect their difference from the largest positive number using the following formula:

P_i = (((Q_T – Q_i) / Q_T) + 1) * 10,

where P_i is cost associated with bathymetry class i, Q_i is the quotient from dividing the larger of the expected or observed cell counts by the smaller for bathymetry class i, and Q_T is largest calculated Q_i.  Quotients for the 30' and greater depth classes were excluded from qualifying for Q_T because of a lack of class 3 telemetry points located in those bathymetric classes.  The resulting values were used to transform bathymetry to a cost surface.   The cost surface was evaluated by using it in the movement path model to delineate paths.  A subset of cells traversed by the paths was selected randomly, and the resulting distribution of water depths was compared to the observed distribution.  Expert opinion was then used to adjust the costs so that the distribution of water depths overlapping the paths approached that of the expected.  This expertise reflected the combined opinion of two experienced naturalists, one in particular was very familiar with the movement behavior of manatees in Tampa Bay.  We recognized that there were biases inherent in the derivation of the costs because the accuracy of the class 3 telemetry points used to initiate the iterative process for deriving the costs reflected manatees primarily resting or feeding in relatively shallow water rather than moving.  However, we believe that the biases did not prevent us from deriving costs that were useful for delineating travel paths that was representative of manatee movement.  For example, many of the corridors delineated by the model made sense and were suspected to exist prior to the study.  One classic corridor is the route between the TECO warm-water discharge site, passing near McDill AFB, and ending at the warm-water discharge of the Florida Progress Bartow Plant.  Another is along the grass flats between the TECO discharge canal and the Little Manatee River.

Quality assurance-quality control of telemetry points.--Telemetry data were a combination of satellite-based Argos locations and visual sightings assisted by VHF radio telemetry.  All points and their attributes were exported from an ARC/INFO point coverage to an ASCII file and then sorted sequentially.  An ARC/INFO AML program was written to verify that the points were located in water.  Telemetry points situated on land occurred because of positional errors inherent in Argos’s estimation of the manatee’s locations and reductions in map resolution resulting from the conversion of the Florida shoreline vector coverage to a raster map.  The fact that manatees live close to land means that even relatively small inaccuracies in position estimates can result in manatees’ locations appearing to be on land.  Points farther than 1 km from water were not used in this study.  Telemetry locations on land and within 1 km of water were moved to the nearest cell of water identified by the Euclidian-distance ARC/INFO GRID utility Eucpoint.  If several water pixels were equidistant from the satellite-estimated position, one was selected randomly.  All points moved by the AML program were verified by field biologists familiar with the movement histories of the tagged animals and the habitat use patterns of manatees in the area.  Points moved to locations considered unreasonable by the AML program, e.g., areas too shallow for manatees, were either deleted or moved by the biologist to the nearest more likely places.  In addition, points situated in power-plant intake canals were moved to their corresponding outflow canals. 

Generation of manatee movement paths.--The movement-path model processed three telemetry points at a time to generate two movement paths.  First, an analysis window was defined that was large enough so that the three telemetry points were not separated from each other by land.  Second, a raster map representing the cost of moving through every cell in the analysis window to the cell containing the second telemetry point in the sequence was generated using GRID’s costdistance command.  This map was used to identify the least-cost manatee movement paths from points 1 to 2 and from points 2 to 3 by applying GRID’s flowpath command.  The movement paths were stored as a line coverage in ARC/INFO and attribute data were attached, including the number of minutes spent travelling between the two points, the number of cells crossed, and the residence time (minutes/cell).  Manatees that were tagged for fewer than 30 consecutive days were not included in this study.

LITERATURE CITED

Flamm, R. O., L. I. Ward, and M. White.  2000.  Atlas of Marine Resources on CDROM.  Florida Fish and Wildlife Conservation Commission, Florida Marine Research Institute.  Version 1.2.