# 4. Population Survey Methodology

## 4.1 General Methods

In most instances use of direct count methodology to investigate distribution, abundance, and density of a large animal population is impractical or impossible. A viable and practical alternative is to use sampling to investigate a representative portion of the population. Sampling techniques are many and varied but ultimately can be classified into 2 categories-

1. Area Samples, in which all animals within a known fraction of the total area under investigation are counted. Meristic data is then converted to an estimate of the total animal population using the formula:

(P=population, S=counted animals in sample, A=total area under investigation, a=area sampled).

2. Time samples, in which there is an incomplete count conducted of animals in the entire area under investigation. The size of the animal population is estimated by calculating the fraction of animals “re-observed” (ie. individuals that were counted during the initial count) during at least one subsequent count - this is the basis of Mark, Release and Recapture methodology.

Coconut crabs, being relatively immobile, are well suited to both area and time sampling. However the varied habitat and relatively large size of the Niue survey area (ie. the entire island) renders use of time sampling, in the form of mark-recapture, impractical over the relatively short term of the present study. Area sampling, in the form of strip-count surveys, has been successfully used in estimating the size of the coconut crab population on Christmas Island in the Indian ocean (Schiller 1988a). Strip-count surveys are one of the simplest methods of obtaining meristic data to estimate the size and density of animal populations. The method involves an observer walking along a predetermined line (transect) and counting all individuals observed within a predetermined distance either side of the centre line and no baits are used. This technique produces count data that is readily converted to absolute density and abundance values. However the initial pilot survey of coconut crabs on Niue (Schiller 1988b) found that use of absolute count techniques was impossible due to the extremely rugged terrain of Niue, particularly the fringing coastal region in which limestone pinnacles to 6m are common.

The basic premise of absolute density survey techniques is that all individual animals (to be surveyed) within a survey region can, and are, observed and recorded. The ubiquitous limestone formations covering much of Niue provide such good refuge to coconut crabs that it is impossible to observe and count all individuals within even extremely limited search areas. As a result it was necessary to use the less direct survey technique of an analysis of Catch Per Unit Effort (CPUE) using count data obtained from baited transects. CPUE is a relative estimate of abundance and can be defined as the number of coconut crabs caught/observed along a transect compared to the number of baits along that transect. CPUE values can be considered as an index of absolute abundance and can be used as a means of comparing density between different areas.

## 4.2 Transect Selection and Establishment

On the basis of data from the previous pilot stock survey it was known that the distribution of coconut crabs on Niue was non-random and that the observed distribution heterogeneity was in part a function of the vegetation/substrate type. In addition previous experience indicated that the density of coconut crabs in all habitat types decreases with increasing distance from the coast. In Vanuatu Fletcher (1988) found that few coconut crabs occurred more than 2 km from the ocean. As a consequence transect placement was determined via stratified random selection in which a 1:30000 vegetation map of Niue (DAFF 1990) was used to divide the island into 4 strata based on the main vegetation categories of Coastal Forest, High Forest, Regenerating Forest and Fernland [ie. herbaceous community dominated by Nephrolepsis hirsutula (Sykes 1970)] (Plates 1–4 respectively). Each strata was then examined in terms of its proximity to the coast and where necessary subdivided into 2 further categories:

1. Coastal- occurring within 2km of the coast.
2. Inland- occurring greater than 2km from the coast.

All vegetation strata except Coastal Forest incorporated both coastal and inland components, resulting in a total of 7 different vegetation categories- 1. Coastal Forest, 2. Coastal High Forest, 3. Inland High Forest, 4. Coastal Regenerating Forest, 5. Inland Regenerating Forest, 6. Coastal Fernland, and 7. Inland Fernland (refer Figure 2- due to the extreme fragmentation of the fernland and regenerating forest regions it is impractical to depict individual land 'parcels' within each zone. As a result fernland and regenerating forest have been combined into a. Coastal Fernland and Regenerating Forest and b. Inland Fernland and Regenerating Forest).

The number of transects allocated to each vegetation category was based on 1. the known habitat types in which coconut crabs normally occur and 2. the relative size of the category in relation to the total sample area. The total area of each vegetation category was obtained by digitising the 1:30000 vegetation map of Niue using a Scriptel digitiser and then using Autodesk Autocad (version 11) software on a personal computer to calculate the areas via the polygon method (refer section 4.6 for more detail).

The 2km2 of tapu area within the Huvalu Forest is comprised of both Coastal Tall Forest and Inland Tall Forest. Because the area is protected and entry by humans is prohibited it was not possible to conduct population surveys of any type in that part of Huvalu forest. In later calculation of the total coconut crab population, CPUE values for the tapu area of Huvalu Forest were assumed to be the same as those of the appropriate vegetation categories. The area involved is quite small and therefore any error inherent in such an assumption will be similarly small.

Starting points for transects within each category were decided by simple random selection of longitude and latitude co-ordinates using a 1:30000 map of Niue (Niue Department of Agriculture, Forestry and Fisheries, 1990). A compass heading for each transect was randomly generated by computer, where terrain prevented following a precise compass heading a path of least resistance was used.

Establishing a transect involved:

1. ensuring general access was possible.

2. cutting a path on a pre-determined compass heading through the vegetation leaving brightly coloured plastic/material markers for future workers to follow. A cotton ‘Hip Chain’ thread, used for measuring distance travelled, was left in place as an additional guide.

3. commencing at the start of the transect, a numbered (1–10) plastic ribbon was placed every 40 m to indicate the position of the coconut baiting points. When a transect was baited prior to a survey the coconut baits were placed at these baiting points.

Initially thirty-three transects were completed, each extending for 360 m and incorporating ten marked coconut baiting points (i.e. every 40 m). During early August 1990 two additional transects were completed to bring the total number of transects to 35 (Figure 3). Thirty-five transects were considered appropriate given that the number of survey transects had to meet the following criteria, in decreasing order of importance:

1. That each coconut crab habitat type have sufficient transects to provide statistically tenable survey results.

2. That the survey programme was able to be conducted by two people.

3. Transects had to be sampled at least once every 4 weeks over the course of the project. This was to ensure that any temporal variation in coconut crab abundance, size etc. associated with moulting and reproductive cycles would be represented in the survey results.

Experience in Vanuatu indicated that only 1 transect could be surveyed a night because of the need to commence surveys within 1 hour of dusk when coconut crabs are most active. In Vanuatu by the time one transect had been surveyed and observers had travelled to a second transect it was usually too late to survey that transect. Surveys were not to be conducted during, or 1 night either side of, a full moon (coconut crabs are less active during periods of full moon). Therefore over a 4 week period, with a 5 day working week, there would be 17 working nights per observer. Hence if a transect is to be sampled once every 4 weeks, and there are 2 people conducting surveys, a total of 34 transects can be accommodated. This number of transects was deemed sufficient, although not optimal, to provide meaningful and statistically relevant population data.

During the early stages of the survey programme it was found that the small size of Niue, together with the small number of coconut crabs being encountered during surveys, permitted 2 geographically adjacent transects to be surveyed in one night with both transects being commenced within 1 hour of dusk. This was most fortuitous because 3 weeks into the survey programme local project staff was reduced from 2 to 1, leaving only Mr Colin Etuata to conduct all transect surveys for the duration of the project.

## 4.3 Transect Surveys

Transect surveys were conducted over a 71/2 month period from 11-5-90 to 28-12-90. On the day a transect was to be surveyed, coconut baits (1/2 coconut fruits tied to a tree or rock) were put out at each bait marker along the transect. Beginning at dusk, a searcher traversed the transect and coconut crabs encountered both at and between baits were numbered, marked and certain morphological features (thoracic length, abdominal expansion and weight) recorded as per a pro-forma record sheet. Thoracic length as measured on coconut crabs refers to the distance, when measured along the carapace, from immediately behind the rostrum to the rear border of the carapace (refer Figure 4). Crabs were individually numbered on the thorax using both a black felt pen and a scribe to etch/scratch the number into the carapace. Marking of crabs permitted recognition of crabs previously recorded.

Transect surveys yielded data of 2 types:

1. Morphological data to be used in investigation of the mean individual size and the size-structure of the population.

2. Count (meristic) data in the form of CPUE values to be used in estimation of the distribution, abundance, and density of the coconut crab population. CPUE values were calculated for each transect survey as follows:

It was assumed that 1. CPUE is linearly related to absolute density and 2. the effective range of attraction of coconut crabs to baits, together with feeding activity, (ie. overall crab ‘catchability’) were constant amongst different regions.

## 4.4 CPUE and Absolute Density

Although CPUE is only a relative estimate of abundance it is possible to calculate absolute density values from CPUE data if the relationship between CPUE and absolute density can be quantified. This relationship, in the form of a conversion factor, was to be ascertained by conducting CPUE and strip-count surveys on consecutive nights along the same transects (where terrain permitted). As indicated in section 4.1, the rugged terrain of Niue precluded extensive application of strip-counts, however there were several locations (near Togo and Hakupu) in which such surveys could be conducted. In this manner CPUE relative density and absolute density for the same survey region were to be obtained and the relationship between both density types quantified into a conversion factor. Hence CPUE relative abundance estimates for each terrain/vegetation category can be considered an index from which absolute abundance can be calculated using the following formula:

Absolute Density = Conversion Factor × CPUE

### 4.4.1 Calculation of the CPUE conversion factor

A series of CPUE and strip-count surveys were conducted along transects 2, 3, 4, 11, 18, and 24 (refer Figure 3) with the strip-count survey conducted initially and the CPUE survey on the following night.

CPUE surveys recorded coconut crabs along all comparison transects, whereas strip-count surveys failed to locate any crabs during the entire survey series. As a result it was impossible to calculate a conversion factor between CPUE values and absolute density using Niue data.

In a study similar to that conducted on Niue, Fletcher et. al. (1991) calculated that for Vanuatu the conversion factor for converting CPUE to absolute density (crabs km2) was 12,000. Because coastal vegetation and topography in Vanuatu and Niue are similar, and therefore the available coconut crab habitats are similar, it was considered that the relationship between CPUE and density for the two countries would also be similar. Consequently the Vanuatu conversion factor was used in calculation of absolute coconut crab density from CPUE values for Niue.

Absolute Density = 12000 × CPUE

## 4.5 Calculation of population Size

The number of coconut crabs in each of the 7 vegetation categories (refer section 4.2) was calculated using the mean CPUE for each vegetation category.

Regional Population Size = mean CPUE × 1200 (CV) × Regional Area (km2)

The calculated number of crabs in each vegetation category was then summed to produce a total population for the island. Mean regional CPUE values were calculated from transect survey data in one of two ways:

1. Where hunting effort (numbers of baits used) for surveys of all transects within a vegetation category was constant, mean regional CPUE was calculated as the mean of the CPUE values for each individual transect survey.

2. Where the hunting effort for surveys of all transects within a vegetation category was not constant, mean regional CPUE was calculated as the relationship between the total number of crabs recorded during all transect surveys in that category and the total number of baits used during all such surveys.

## 4.6 Calculation of Land Areas - Effect of Fragmentation

As indicated in section 4.2 above, a 1:30000 forestry map of Niue was digitised and Autocad (V 11.0) software used to calculate the total area of each vegetation category. Through the process of fragmentation formally continuous vegetation categories have been broken up through clearing for agriculture into a number of separate isolates. Initially it was intended that the land areas of all similar category vegetation isolates, however detached and/or small, would be summed to produce a total land area for each of the 7 vegetation categories. However preliminary survey results suggested that such an approach was inappropriate due to the high degree of fragmentation of Inland High Forest and Coastal and Inland Light and Scattered Forest.

Small isolates of Coastal Light and Scattered Forest separated from other coastal vegetation types by relatively large expanses of Fernland were found to have coconut crab CPUE levels the same as for the surrounding Fernland. However the corollary, where a low CPUE vegetation isolate surrounded by a high CPUE vegetation type took on the higher CPUE value, did not occur.

Extensive fragmentation of most vegetation categories on Niue has resulted in the formation of large numbers of isolates surrounded by a 'sea' of differing vegetation type. Where the surrounding vegetation type has a lower CPUE these isolates tend to take on the properties, with respect to coconut crab CPUE levels, of the surrounding vegetation.

On this basis a quasi-quantitative assessment of the degree of 'isolation' of small isolates of high CPUE vegetation surrounded by lower CPUE vegetation was conducted. If the degree of isolation was deemed large enough, the isolate was considered to be of the surrounding vegetation type in terms of both crab CPUE and in later calculation of land areas of each vegetation category.

## 4.7 Statistical Analysis

Sample and population statistics/parameters were calculated using the statistical package SAS Ver 6.03, SAS Institute Inc. 1989.