CRAMP Home

Long Term Monitoring

Status of the Reefs

Site Selection Process

Methods

Diving Protocols

Permits

Site Survey Protocol

Benthic Monitoring

Reef Fish Monitoring

Database

Archiving of CRAMP Survey Data

Study sites

Results

Cooperative Programs

Rapid Assessment

Mapping

Watersheds

Publications



Search the CRAMP Website

Navigate the CRAMP Website

Bibliographic Search

Training Divers to Estimate Fish Lengths Underwater Using Fish Models

Alan M. Friedlander¹ and Eric Brown ^2,3

¹ Oceanic Institute, Makapuu Point, Waimanalo, Hawai‘i

² Department of Zoology, University of Hawai‘i at Mānoa, Honolulu, Hawai‘i (Affiliation during study)

³ National Parks Service, Kalaupapa, Moloka‘i. (Current Affiliation)

Introduction

Underwater visual censuses of fishes are a valuable method to obtain information on the patterns and change of fish populations. Although underwater visual census may underestimate nocturnal and cryptic species, comparisons with the results of complete (destructive) sampling of whole fish assemblages indicate that it can give good estimates for most diurnally active fishes (Brock, 1982). No single method exists for accurately censusing a diverse group such as reef fishes, which often have varying degrees of delectability, mobility, and wariness (Thresher and Gunn, 1986). Despite these limitations, underwater visual census is the best single non-destructive method for obtaining estimates of abundance for an entire fish assemblage in hard-bottom habitats.

Biomass analysis is an important consideration both ecologically and from a fisheries management perspective. In Hawai‘i, where a large number of reef fishes are targeted by fishers, most of the biomass of fishes observed on the reef is exploitable. Length estimates of fishes from visual censuses can be converted to weight using the following length-weight conversion:

W = aSL^b - the parameters a and b are constants for the allometric growth equation where SL is standard length in mm and W in grams. Length-weight parameters are available for 150 species commonly observed on visual fish transects in Hawai‘i (Friedlander 1997). In the case where length-weight information does not exist for a given species, the parameters from a congener species can be used. Additional information is available from the Hawai‘i Cooperative Fishery Research Unit. It all this fails, the length³ gives a rough estimate of weight for many fish species.

Methods

Study area

This study was conducted on the forereef at Ho‘ai Bay on the island of Kaua‘i. The areas consisted of basalt boulder habitat in ca. 25’ of water. Visibility on average was greater than 50’.

Fish models

Fish models were created from scanned photographs of a variety of reef fishes that exhibiting a wide range of body shaped and color patterns. These included members of the following families: parrotfishes (Scaridae), wrasses (Labridae), surgeonfishes (Acanthuridae), damselfishes (Pomacentridae), and goatfishes (Mullidae). The scanned images were imported into an image-processing program to create left and right-sided mirror images of each fish. Several different sizes of each fish were created and printed on a color ink-jet printer. Left and right-sided mirror images of each fish were cut out and thermal laminated with a thin layer of foam placed between the two sides to create positive buoyancy. Snap swivels and monofilament line were attached to the ventral surface of each fish to allow for easy deployment and manipulation of fish position along the transect line.

Estimating fish length and dive training

Typical fish census transects were 25 x 5 m with the diver swimming down the center of the transect and estimating width 2.5 m to either side of the transect centerline. To estimate fish lengths, each diver swam along a transect line that was ca. 2.5 meters from a parallel transect line with attached fish models. On the first run, 17 fishes were haphazardly positioned along the fish model transect and the diver swam at a constant speed while estimating the standard length of each fish model. The diver then returned along the fish model transect and measured the actual standard length of each model. On the second run, six additional models were added and the location of all the existing models was haphazardly changed. The diver again estimated fish lengths and measured the actual lengths on the return swim. One additional run was conducted with all 23 fish models again being haphazardly changed along the transect.

Statistical analysis

A two-way ANOVA was conducted with observers (2) and runs (3) as fixed factors in the ANOVA model. The dependent variable in the model was the absolute difference in the observed vs. the actual standard length of each fish model.

Results

The difference in the observed vs. the actual standard fish lengths was significantly different between observers and among trials (Table 1). Overall, mean length estimate differences for observer 1 (Mean = 0.702 cm, SEM = 0.185) were significantly lower (t = 4.420, P < 0.001) than for observer 2 (Mean = 1.853, SEM = 0.184; Table 2).

Table 1. Two-way ANOVA with observers and trials as fixed factors. The dependent variable in the model was the absolute difference in the observed vs. the actual standard length of each fish model.

For both observers pooled, the mean difference in estimated vs. actual length was greatest for the first trial (2.235 cm) and became progressively lower with each subsequent run (trial 2 = 0.891 cm; trial 3 = 0.707 cm). The mean difference in estimated vs. actual length was not significantly different between trial 2 and 3 (Table 3).

Table 2. Least square means for observers and samples.

Observer 1 showed no significant difference in mean estimated length differences among the 3 trial (Table 3). The mean estimated vs. actual length difference for observer 2 after the first trail was 3.647 cm (SEM = 0.350). After the first trial, observer 2’s estimated vs. actual length difference declined to 1.00 cm (SEM = 0.301) and showed no significant difference in mean estimated length difference between trial 2 and 3 (Table 3). After 3 trials, observer 1’s mean estimate vs. actual fish lengths was 0.500 cm (SEM = 0.308) and observer 2’s mean estimates vs. actual fish length was 0.913 cm (SEM = 0.301). Observer 1 and 2 differed significantly in their estimates vs. actual difference for trial 1 but these differences were not significant for the subsequent trials (Table 3)

Table 3. All Pairwise Multiple Comparisons Procedures (Bonferroni t-test).

Discussion

Observer 1 had extensive previous experience in estimating fish length and his estimates did not change significantly during the training period nor did his estimates differ much from the actual lengths. Observer 2 was relatively inexperienced at estimating fish lengths underwater, but after one trial his estimates were comparable to observer 1 and did not differ significantly from the actual lengths. This small experiment shows that with minimal training, divers can learn to estimate fish lengths underwater. Both observers obtained less than one cm accuracy in estimating actual fish length after one 2 trials. Re-calibration should be done on a periodic basis to insure that length estimates do not differ significantly from actual lengths.  It is apparent that divers can be trained to obtain accurate fish length data as part of the underwater transect technique.  The use of this method enables CRAMP to obtain biomass estimates for reef fish. This parameter is of utmost importance.

Figure 1.  Leaning curve for trained diver (blue) and untrained diver (purple) during the three trials. (Click for a larger view)

References:

Brock (1982)

Friedlander (1997)

Thresher and Gunn (1986)

Last Update: 04/21/2008

By: Lea Hollingsworth

Hawai‘i Coral Reef Assessment & Monitoring Program

Hawai‘i Institute of Marine Biology

P.O. Box 1346

Kāne‘ohe, HI 96744

808-236-7440 phone

808-236-7443 fax

email: jokiel@hawaii.edu