Experimental Design - "Problem Focused" Research
Download the publication (PDF file):
Brown E, Cox E, Jokiel P, Rodgers K, Smith W, Tissot B, Coles S and Hultquist J. (2004) Development of Benthic Sampling Methods for the Coral Reef Assessment and Monitoring Program (CRAMP) in Hawai‘i . Pacific Science 58 (2): 145-158.
CRAMP long-term monitoring experimental design allows detection of changes that can be attributed to various factors such as:
The emphasis of the program is on the major problems facing Hawaiian coral reefs as listed by managers and reef scientists during workshops and meetings held in Hawai‘i (1997-1998). These are:
CRAMP experimental design gives priority to areas where baseline data relevant to these issues were previously collected. The program is synthesizing existing data into a comprehensive Hawai‘i-wide database and is incorporating current Hawai‘i coral reef research into a common experimental design. Transect dimensions, number of replicates, and methods of evaluation have been selected to detect changes with statistical confidence. Standard techniques include the establishment of permanent transects to quantify fish, coral, algae, and invertebrates at study sites.
CRAMP researchers are quantifying changes that have occurred on coral reefs subjected to varying degrees of fishing pressure, sedimentation, nutrification, and algal growth and are conducting experimental work in order to test hypotheses concerning the role of these environmental factors in the ecology of coral reefs. We are also in the process of resurveying, updating and integrating existing ecological information on an array of coral reefs that have been designated as areas of concern or, "hot spots," by managers and scientists.
CRAMP Long-term Monitoring Protocol
One of the major objectives of the CRAMP program during the first year was to establish a sampling protocol that could detect change in coral cover over time with sufficient statistical power (P>0.8). The first step involved the evaluation of historical methods to determine if any of these procedures could be incorporated into the CRAMP protocol. After careful analysis it was determined that only the fixed photoquadrats utilized by Dr. Steve Coles at Bishop Museum had sufficient power. The method, which samples a relatively small area, is suitable to address small-scale questions on coral growth, recruitment and mortality, but inference on general reef condition is difficult across broader sections of reef. The second step involved soliciting input from colleagues conducting coral reef monitoring programs in the Florida Keys and the Great Barrier Reef. Their general recommendation was to use digital video to sample coral cover over large areas of the reef. (This was prior to high-resolution digital cameras). Before we could implement their designs, however, we had to evaluate the appropriateness of these techniques for Hawai‘i. The following parameters in the sampling design were determined in the third step:
The results of this evaluation were presented at the National Coral Reef Institute Conference in Florida and are summarized by the CRAMP research team (Brown, et al. 1999). Repeatability and appropriate transect length were tested using photoquadrats on a transect line sampled over a short time interval. Shorter transects of 10m were found to have higher precision (Ability to replicate quadrats on a transect) than transects of 25m and 50m. Photoquadrats produced similar results to visual estimation techniques, regardless of observer, but neither method yielded satisfactory precision.
Digital video was evaluated at Hanauma Bay, O‘ahu over 2 time intervals separated by 84 days. It was assumed that overall coral cover would not change dramatically during this time period. Power curves were constructed using methods described by Zar (1999) for detecting a 10% change in coral cover across 2 time periods (Figure 1). Number of frames was more important in increasing power than number of points though the difference was not substantial. This is primarily due to the fact that more frames sample a larger portion of the habitat, which incorporates more of the heterogeneity of the substrate. A sample size of 10 transects per site appeared to be adequate for characterizing the coral cover using a power value of 0.8 set as a convention by Cohen (1988).
Figure 1: Power analysis for number of samples with different numbers of points/frame and different numbers of frames/transect (Zar, 1999). (Click for a larger view)
Fixed transects were chosen over random for several reasons. First, it is difficult to properly implement a randomized protocol for transect placement without a map of benthic habitats that is geo-referenced. When CRAMP first began in 1998, this did not exist for the state of Hawai‘i . Second, the majority of the historical data uses fixed transect locations so integrating the current protocol with previous work is simpler. Third, after the initial random setup, the fixed transects should be easier to resample, thus reducing preparation time and ultimately costs to generate the random grid for subsequent transect measurements (Green and Smith, 1997). Fourth, randomized sampling of transects will have difficulty in detecting change in coral cover if reefs change dramatically over time. This is because the random protocol measures inherent spatial variation at each sampling period, which adds variance associated with spatial heterogeneity of the reef rather than changes or patterns that are time-related (Green and Smith, 1997). Fifth, using a repeated measures ANOVA design with fixed transects can provide additional information on population and community structure that is difficult to obtain with random transects (Hughes, 1996; Connell et al. 1997). Sixth, the time and cost complications with random transects are not worth the broader inference about reef "condition" especially if the fixed transects are representative of habitat variation (Andy Taylor, personal communication). Finally, interpreting results from fixed transects is much easier for the general public and resource managers to comprehend than using a randomized sampling design.
monetary constraints were examined to determine the optimum sampling
protocol. The analysis revealed that digital video collected more data per
unit time than visual estimation, planar point intercept and photoquadrats.
It was the most expensive option considered at $5,500 for the system but
since field time underwater is the principal limiting factor then the
quantity of field data collected outweighs the expense. In addition, digital
video and photoquadrats also enable archiving of the data for later
re-analysis to address additional questions.
Kuulei Rodgers running a digital monopod survey. Photo by: Paul Jokiel
Based on the results from the evaluation procedure we selected 2 methods to address changes in overall coral cover and growth, recruitment and mortality of benthic organisms. Digital video was initially used to measure changes in coral cover by selecting at random, ten permanent (fixed) transects at 2 depths (3m and 10m). Each transect is 10m in length and analyzed using 20 randomly selected video frames with 50 randomly selected points per frame. Frequency of sampling will be once a year at each site. This is sufficient to detect a 10% change in coral cover over time with high statistical power (P> 0.8 at a £ .05) across of variety of habitats in Hawai‘i.
Since the initial methods comparison in 1998, CRAMP has improved its methodology to keep up with advances in technology, replacing video with digital stills. Unlike prior digital cameras, recent cameras have resolution superior to video and the card media can store close to 1,000 high quality images. The initial costs of the equipment are lower and the images can be archived. The valuable in situ time is shorter as well as the time spent processing the images. Frame-grabbing is completely eliminated. The video camera cannot keep an exact distance from the bottom while the still camera mounted on a simple monopod assures a constant distance. With a still camera, there are no oblique angles that can affect results since the camera is held completely vertical by the monopod. It is however important to use consistent methodology when comparing sites spatially and/or temporally. Yet as newer and better technology is introduced it is important to update and upgrade methods. CRAMP began by using video techniques and replaced this with digital stills in 2003.
Prior to the switch, the compatibility of the methods was assessed through intercalibration, using both methods (video and digital still images) at a large number of sites (30) that encompassed a wide range of coral cover. Once the methods proved compatible, all subsequent surveys were conducted with digital cameras. Non-overlapping digital stills are taken to assess the characteristics of benthic populations. High resolution digital images are taken along a 10 m transect using an Olympus 5050 zoom digital camera with an Olympus PT050 underwater housing. The camera is mounted to an aluminum monopod frame, 1.7 m from the substrate to provide a 50x69 cm image. A 6 cm bar provides a measurement scale. The software program PhotoGrid (Bird 2001) is used to quantify percent cover, richness and diversity of corals, algal functional groups and substrate cover. Images are downloaded and the 20 non-overlapping images from each 10 m transect are imported into PhotoGrid where 50 randomly selected points are projected onto each image for a total of 1,000 points per transect. These data are saved in a comma separated values (CSV) file, proofread in Excel and imported into Microsoft Access XP, a relational database. Access data can then be queried and exported to statistical programs for analyses.
The second method employs fixed photoquadrats to examine trends of individual organisms with regards to growth, recruitment and mortality. Five haphazardly selected photoquadrats at each depth contour will be established with 4 pins at each corner to ensure accurate repositioning of the frame. The frame dimension will sample 0.33 m2 of the substrate at a height of 0.5m from the bottom. Images of sessile organisms are traced and digitized for 2D estimates of aerial coverage. In the early years sampling was scheduled once a year at each site in concordance with the digital video surveys. Most sites are surveyed annually but due to funding issues not all sites are surveyed every year. Sites that show changes over time take priority in resurveys. We project that all sites will be resurveyed at the 10 year mark in 2009 to compare temporal changes in coral populations.
Eric Brown positioning the frame for a photoquadrat. Photo by: Paul Jokiel
Brown, E, E Cox, B Tissot, K Rodgers, and W Smith (1999). Evaluation of benthic sampling methods considered for the Coral Reef Assessment and Monitoring Program (CRAMP) in Hawai‘i. International Conference on Scientific Aspects of Coral Reef Assessment, Monitoring, and Restoration. April 14-16, Ft. Lauderdale, FL.
Cohen J (1988) Statistical power analysis for the behavioral sciences. Lawrence Erlbaum Associates, Hillsdale, New Jersey (567 pp).
Connell, J H, T P Hughes, C C Wallace (1997). A 30-year study of coral abundance, recruitment, and disturbance at several scales in space and time. Ecol. Mono. 67(4): 461-488.
Green, R H and S R Smith (1997). Sample program design and environmental impact assessment on coral reef. Proc 8th International Coral Reef Symposium. 2: 1459-1464.
Hughes TP (1996) Demographic approaches to community dynamics: a coral reef example. Ecology 77: 2256-2260.
Zar JH (1999) Biostatistical Analysis. Prentice Hall, New Jersey. 663 pp.
Last Update: 05/08/2008
By: Lea Hollingsworth
Hawai‘i Coral Reef Assessment & Monitoring Program
Hawai‘i Institute of Marine Biology
P.O. Box 1346
Kāne‘ohe, HI 96744