Aerial Surveys [larger image] Eight survey flights on fixed, strip transects are flown in summer following protocols by Miller et al. (1998). Two observers in the plane independently record the location, activity and number of manatees observed in each group. Distribution surveys are flown monthly following an irregular predefined path across the mainland waterways, the mangrove islands, and offshore bays. Four habitat types are delineated for both survey types: Freshwater canals and creeks, Inshore Bay, Travel corridors among the mangrove islands, and Offshore bays.

Transect Results [larger image] Summer data from strip transect surveys showed manatees generally were dispersed across the transects in small groups. Some transects were high use areas, while others were low use. The majority of summer sightings on the transects were in Offshore bays in or near seagrass beds (66.2% in 2001 and 76.2% in 2002). Manatee activities recorded during a sighting were comparable between years. The greatest number of sightings was of animals feeding (46.4% in 2001 and 51.0% in 2002).

Distribution Survey Results [larger image] The number of manatees counted in the Ten Thousand Islands area varies seasonally with the fewest in the fall and higher numbers in winter and summer. In winter, higher numbers are observed at warmer Freshwater sites inland. In spring and summer higher numbers are observed Offshore, as observed on the strip transect surveys. Radio-telemetry studies of individual manatees (see poster 2-100 and 2-101) show similar patterns and provide greater insight and detail on seasonal use of habitats.

Double Observers Analysis of the double observer data from the strip transect surveys showed manatees at the surface can be missed by an observer (i.e. perception bias). Detection varied with group size, but not with other factors proposed to affect sightability. Group size Probability of initial sighting (95% Confidence Interval) Probability of resighting1 (95% Confidence Interval) 1 0.657 (0.580-0.727) 0.732 (0.655-0.797) 2 0.746 (0.680-0.803) 0.780 (0.718-0.831) 3 0.818 (0.721-0.887) 0.821 (0.735-0.884) 4 0.873 (0.747-0.941) 0.857 (0.737-0.927) 5 0.914 (0.770-0.971) 0.886 (0.735-0.956) 1resighting by second observer given seen by first observer Huggins Closed Capture model (Huggins 1991) in Program MARK (White and Burnham 1999) was used to model and estimate detection rates. Sources of variation modeled included: group size, habitat type, manatee activity, date of survey, year, and survey conditions. The mean probability of initially sighting a manatee group was 0.715 (95% CI 0.655-0.768). Mean probability of resighting by the second observer was 0.769 (0.709-819). The best model was one with variation due to group size with no variation among surveys or years. Detection increased with increasing group size.

Reflights Reflights of the strip transects showed some manatees were hidden below the surface of the water (i.e. availability bias). [larger image] Sediment plumes were routinely seen muddying shallow water on the strip transect surveys, suggesting a large animal had disturbed the bottom. In 2002, 5 transects during each survey were randomly chosen for an immediate reflight to identify the cause of any disturbance plumes and to locate any additional manatees not seen on the first flight. Of 23 observations of a plume, 14 (61%) were manatees, 6 (26%) were caused by tide action, 2 (9%) were schools of fish, and 1 (4%) was a dolphin. Additionally, 15 manatee groups not seen on the first flight were observed during the immediate reflight.

Major Features Developed by J. D. Nichols and colleagues at the USGS Patuxent Wildlife Research Center (MacKenzie et al. 2002; 2003). Based on presence/absence of a species on study plots randomly or systematically distributed across the landscape. It does not require counts of individuals on a plot. Designed to estimate occupancy rate in the presence of imperfect detection probabilities, which may vary both spatially and temporally. Data from multiple surveys are used to directly estimate detection rates. Provides a flexible, robust modeling framework that incorporates covariates. Habitat traits specific to each plot can be used as covariates to model hypotheses about variation in occupancy and detection rate. Uses latest model comparison approach to select from the competing models those models that best describe the variation in occupancy and detection rate. Developed to estimate occupancy rate within a season or year (MacKenzie et al. 2002) and to estimate rate of change in occupancy over long-term studies (MacKenzie et al. 2003). Incorporated into Program MARK (White and Burnham 1999), the primary computer program used today for capture-recapture analysis.

Application to Manatee Aerial Surveys [larger image] The strip transects and distribution survey routes are segmented into habitat patches. We are exploring different patch sizes to determine an appropriate scale for analysis and monitoring. A priori models have been identified to test for effects on occupancy rate and include habitat type, water depth, proximity of patch to foraging beds, proximity to freshwater sources, and season of the year. Models to test for effects on detection rate include survey conditions, season, and habitat type. Correlation of aerial survey data with telemetry core use data will be examined with a model that uses the number of telemetry locations of tagged animals as a patch covariate.