NOTAS BIBLIOGRÁFICAS
III. El canon de las grandes re ligiones
The software Distance 6.2 (Thomas et al., 2010) was used to generate different survey designs for the study area. A shapefile of the study area generated in ArcGIS 10.2 was imported into Distance following the steps indicated in the Chapter 5 of the software user guide (Thomas et al., 2010). Coral reef areas were considered like land and removed from the study area surface because reefs emerge from the sea bottom up to few centimetres under the surface making unlikely the presence of cetaceans above them (Figure 2.4). A point grid layer was added to the study area with a grid spacing of 2 km generating 3,247 points. The grid spacing should be larger than the maximum perpendicular distance at which animals are expected to be spotted. This distance was estimated to range from 1,000 m to 1,800 m given the observers’ eye-level above the sea surface (approximately 8 m), the observation method used (naked eyes), the relatively small size of the target species (ranging from 2 to 4 m in length) and the expected sighting conditions.
Chapter 2 – Survey Design
Figure 2.4 - Example of coral reef emerging from the sea bottom to few centimetres below the sea surface.
A systematic design was preferred to a random one, and a continuous zigzag design was preferred to a parallel-line design to decrease navigation off effort between transects. An equal-spaced zigzag sampler was selected to create the survey design as it yielded a lower bias in abundance estimates than other available options (such as the equal-angled samplers) (Strindberg & Buckland, 2004). The shape of the study area approximated a convex hull enabling the zigzag sampling design to be used. Effort was determined by line spacing leaving the software to generate the number of sampler segments (transects) and their aggregated length for each stratum. The same line spacing was allocated for all strata. The angle of the design axis was selected in order for transects not to be perpendicular to the dominant wave direction in each stratum
Chapter 2 – Survey Design
to reduce vessel pitch. The selected strip width was 2 km. The coverage probability was generated empirically using 5,000 simulations.
The total survey effort was estimated considering the budget available, the desire to estimate abundance in the two seasons (winter from mid-October to mid-April, and summer in the remaining six months), and logistic restrictions (Marsa Alam and Hamata located in the northern stratum offered the only harbours in the study area where the survey vessel could be refuelled). The available budget was sufficient to allow for 200 days of vessel charter and associated operational costs to be covered. It was assumed that about 50% of those days would be lost due to bad weather and off- effort navigation to/from Hamata harbour for refuelling. To estimate the overall survey effort (considering the three years of project) the remaining days were multiplied by the hours of navigation per day (i.e. seven hours, taking into account seasonal daylight hours (n summer 13 hours of daylight, and 10 hours in winter were available in the study area based on information for latitude 24°25’ N, source: www.timeanddate.com), time spent in closing-mode with dolphins during sightings - 30-45 min, and travel time required to transit between transects and safe anchorages for the night), and by the average vessel speed (16 km/h). To obtain the survey effort, the overall effort (11,200 km) was divided by the desired number of surveys (five surveys:, three in the summers of 2010, 2011 and 2012 and two winter surveys in 2010 and 2011). A total effort of 2,200 km resulted for each of the five surveys.
The precision of the density (and abundance) of groups was predicted using the formula from Burnham et al. (1980, p. 35):
= ! "# Formula 2.1 Where:
Chapter 2 – Survey Design
$ Total survey effort length
%=&'()
+'*+,(
0)-
.+(0)/ଶ 0 Variance inflation factor +(0)
Probability density function of perpendicular distances from the trackline, evaluated at zero distance (x=0)
Number of groups encountered
Target coefficient of variation of the density estimator
$
A priori known encounter rate for the target species in the area ($ = km navigated; =
number of groups encountered)
The variance inflation factor % includes the variability of sample size and detection probability and is usually considered equal to 3 (Burnham et al., 1980) following the observation made by Eberhardt (1978) that the parameter typically ranges from 2 to 4. As encounter rates for the target species in the region were not available a range of the lowest values available in the literature was used instead (Ballance & Pitman, 1998; Barlow, 2006).
The expected density coefficients of variation for density estimates are shown in Table 2.1. CVs ranging from 0.20 and 0.50 are considered a typical achievement but they might require a big effort if encounter rate is low (Ferguson & Barlow, 2001; Hammond
et al., 2002; Barlow, 2006). In order to obtain a better precision (less than 0.50), a group was expected every 143 km during a survey considering about 2,200 km of survey effort.
Chapter 2 – Survey Design
Table 2.1 – Expected coefficients of variation for density estimates using different combinations of line lengths (L) and potential encounter rates (ER) of group per km. Coefficients of variation were estimated following Formula 2.1. Colour legend: dark grey CV≥ 0.50; grey 0.50 < CV ≤0.20; light grey CV <0.20.
L (km)
Encounter rates (group/ km)
0.001 0.003 0.005 0.007 0.01 0.02 0.03 2,000 1.22 0.71 0.55 0.46 0.39 0.27 0.22 3,000 1.00 0.58 0.45 0.38 0.32 0.14 0.18 4,000 0.87 0.50 0.39 0.33 0.27 0.12 0.16 5,000 0.77 0.45 0.35 0.29 0.24 0.11 0.14 6,000 0.71 0.41 0.32 0.27 0.22 0.10 0.13 7,000 0.65 0.38 0.29 0.25 0.21 0.09 0.12 8,000 0.61 0.35 0.27 0.23 0.19 0.09 0.11 9,000 0.58 0.33 0.26 0.22 0.18 0.08 0.11 10,000 0.55 0.32 0.24 0.21 0.17 0.08 0.10 11,000 0.52 0.30 0.23 0.20 0.17 0.07 0.10 12,000 0.50 0.29 0.22 0.19 0.16 0.07 0.09
The final survey design included 65 transects (25 in the northern stratum, 21 in the central and 19 in the southern), and a total effort of 2,196 km (681 km in the northern stratum, 800 km in the central and 715 km in the southern). The average coverage probability across strata was 0.64 (Figure 2.5).
Chapter 2 – Survey Design
Figure 2.5 - Map of the study area showing (a) the coverage probability in each stratum, (b) and the equal-spaced transects for the pilot survey (white lines).