Experimentation, creation, and validation of the Hawaiian Koʻa Card occurred in two phases: Phase I: card development (Phase I) and card validation (Phase II) (Fig. 1).
Phase I: card development
Collection and acclimation
To assure the widest spatial coverage of ecologically important coral species represented on the Hawaiian Koʻa Card, coral species selection criteria was based on the Hawai‘i Coral Reef Assessment and Monitoring Program’s (CRAMP) determination of the species with the highest total percent cover throughout the Main Hawaiian Islands [25] and combined statewide data from other monitoring efforts [26]. Six of the most prominent coral species (n = colonies; Montipora capitata (n = 10), Montipora flabellata (n = 6), Pocillopora meandrina (n = 11), Porites compressa (n = 10), Porites evermanni (n = 8), Porites lobata (n = 12)) were collected from four locations on the east and north facing shores of Oʻahu, Hawaiʻi (Kāneʻohe Bay, Waimānalo, Punaluʻu, and Kualoa). Individual colonies (n) were fragmented into ten replicate fragments (~ 2–5 cm) and secured to ceramic plugs using cyanoacrylate gel to reduce variations in response due to genetic differences. Fragments were allowed to acclimate to mesocosm conditions in ambient seawater (23–24 °C) for 14 days prior to experimentation. All corals were kept in continuous flow-through 170 L3 mesocosms (n = 3) are under full solar radiation that provided a continual supply of unfiltered seawater pumped directly from Kāneʻohe Bay, Hawaiʻi. Therefore, conditions in the mesocosms are realistic and follow natural cycles on the adjacent reef flat. Flow was maintained at a rate of 4 L min−1 resulting in a full volume turnover rate of ~ 45 min. Environmental conditions during the acclimation period are shown in Table 1.
Table 1 Environmental conditions during the acclimation period during Phase I Experimental setup: laboratory bleaching
Flow-through mesocosm temperatures were gradually increased, using titanium heaters (Finnex 800 W), above winter ambient temperatures of 23–24 °C to a peak midday temperature of 32 °C over a four-week period (0.25–0.5 °C per day). Prior to heating, a fragment from each colony was selected as a baseline color to represent each colony from each of the six species (6 February 2017). Once heating commenced, coral fragments were monitored daily and selected based on visual changes in color by the same observer (K. Bahr) to maintain consistency. Selected fragments were measured for photosynthetic efficiency (quantum yield, Fv/FM) and photographed prior to flash freezing in liquid nitrogen for later processing for symbiont counts and chlorophyll a (chl a) analysis.
Coral health measurements: photosynthetic efficiency, symbiont density, and chlorophyll concentration
Rapid light curves (RLCs) using the saturating pulse method were conducted with a Diving-PAM (Heinz Walz GmbH, Germany) on each coral fragment to measure photosynthetic efficiency. Photosynthetically active radiation (PAR) is defined as the wavelengths of light (400–700 nm) utilized by plants for photosynthesis. During this experiment, PAR values were set at 0, 56, 95, 165, 225, 350, 480, 537, 585 µmol photons m−2 s−1 to allow for photosystem saturation in actinic light (PAR) to occur at the sixth or seventh measurement in all species. Once a maximum electron transport rate (ETR) value was reached, the increasing PAR values resulted in a plateau or decrease in ETR, commonly assumed to be a sign of photoinhibition.
The Diving-PAM was connected to a PC with WinControl-3.25 version software and fitted with a blue light-emitting probe (470 nm, LED, 0.05 µmol photons m−2 s−1, 5 Hz) set to intensity level 8, which is too low to induce fluorescence when used as a measuring light. System parameters of the Diving-PAM were evaluated for each species and set at the following consensus values: measuring light intensity = 8, signal damping = 2, gain = 6, saturating light pulse intensity = 8, saturating light pulse width = 1, actinic light intensity = 4, and actinic light width = 0:30. Rubber surgical tubing was attached to the probe end with a 1 mm protrusion beyond the probe’s measuring surface. This allowed consistent measurements at a fixed distance from the coral surface. The fluorescence offset was regularly reset in seawater with the Auto-Zero function.
Selected coral fragments were removed from the outdoor seawater tables, placed in a 2-gallon aquarium and transported to a darkroom for acclimation (~ 40 min). Corals were aerated with battery-powered aerators and kept in total darkness for 20–30 min prior to measurements. Individual corals were moved from the acclimation tank to a measurement tank for each RLC to prevent photosystems in non-target fragments from reacting to the strong saturating light pulse and the increasing actinic light emitted from the PAM during measurements. RLCs were performed at three unique points at a 90° angle to each coral fragment. The first probe location was selected in the dark with subsequent locations as far from the previous as possible. This was necessary to avoid photosystem activation in areas adjacent to the probe. During the 90s from first saturating pulse to last, the probe was carefully held in position to avoid any changes in probe coverage. The mean value of Fv/FM for the three measurements were used for analysis.
Fragments were subsequently photographed, flash-frozen in liquid nitrogen, and stored at − 20 °C. Coral tissue was removed from the skeletons for symbiont and chlorophyll analyses. Symbiont density was determined for each fragment through replicate cell counts using a gridded hemocytometer. Chlorophyll concentration (a, c, and total) level was determined using a spectrophotometer and equations described in Jeffrey and Humphrey [27]. These measurements were standardized to coral skeletal surface area, which were calculated according to the wax dipping technique described in Stimson and Kinzie [28]. Changes in coral health parameters (i.e., symbiont density, chlorophyll a concentration, photosynthetic efficiency) due to elevated temperatures were analyzed using linear and curvilinear regression models (JMP Pro 13).
Photographic and color analysis
Photographing a coral specimen in the laboratory
To develop sets of relative color schemes for coral specimens for the Hawaiian Ko‘a Card, color changes and range of colors were documented as photographic images in a controlled environment. A digital camera (Canon G16) and an external flash mounted on a stationary stand were used to evenly illuminate and photograph a coral specimen while maintaining a fixed distance and angle. Camera parameters, including shutter speed, aperture, ISO, white balance, and flash settings, were manually set and remained consistent across all images. Coral fragments were placed in aquaria filled with clean seawater during photographing. A commercial underwater color reference card (DGK Color Tools WDKK Waterproof Color Chart) was placed parallel to the hind wall of the aquaria adjacent to the fragment for subsequent color balancing during image development. All images were preserved as raw image files in DNG format to retain full resolution captured by the camera sensor to minimize loss of information. Preserved DNG raw image files were white balanced (90% reflectance) and neutral balanced (gray, 18% reflectance) in Adobe Photoshop CS5 using the color chart captured in each image as reference. The histogram function allowed evaluation of the distribution of red, green, and blue (RGB) values, representing white and neutral, to maintain constant values while avoiding overexposure of images. Images were then converted into TIFF format for subsequent processes of color indexing and selection to establish relationships between colors in photographs and coral fragments.
Indexing coral colors
The preserved TIFF image files were used for indexing and selecting representative colors for each coral specimen using the Indexed Color function in Adobe Photoshop CS5. Initially, a representative surface area of each coral fragment in an image was isolated as a separate image layer, excluding shadows. The average surface average sampled was 47.5 ± 2.2 cm2 per colony but varied by species. The 10 most frequent colors appearing in a representative area of each fragment, from each image, were indexed and saved as a color table file with act extension, preserving associated numerical color information including RGB, CMYK (Cyan, Magenta, Yellow, Black), and HSB (hue, saturation, and brightness) values.
Swatch and color wheel production
Area of discoloration (i.e., shaded regions and coral tips) within each photograph were removed prior to color selection. Adobe Photoshop was used to select the ten most common colors per fragment to avoid potential biases in color selection. Color values (RGB, HSB, hex code, and CYMK) were recorded for each color (10 colors for each fragment) for all fragments (total of 5340 colors). All color sets were uploaded into Adobe Creative Cloud Photoshop CC version 19.1.8 and organized by coral species. Color palettes were sorted by species and grouped by sampling date and length of exposure to high temperature. The ten most common colors for each of these fragments were viewed and sorted from darkest to lightest hues (1–9) both visually and using RGB values. From these representative selections, colors were designated to one of four quadrats based on color group. CMYK and hex code values for each color were recorded and provided to the printer along with the finalized design. The Hawaiian Ko‘a Card was designed using Adobe Illustrator CC version 19.0. Colors displayed on the card were selected based on the highest frequency of occurrence within color pallets and representation for the common coral species found within Hawai‘i. A prototype version of the card was produced on waterproof high impact polystyrene sheet plastic in a limited quantity for field and laboratory validation. The final version of the card is printed on ecoplast biodegradable high impact polystyrene plastic (Fig. 2).
Data synthesis and analysis
To determine color numbers for each fragment, the Hawaiian Koʻa Card was photographed in the same lighting as the coral fragments from the Phase 1 simulated bleaching event (n = 534 fragments). The image of the Hawaiian Koʻa Card was placed alongside the photograph of each fragment on the same computer screen to identify the color of each coral fragment post hoc as it related to the colors on the card. The coral color was matched to a number on the Hawaiian Ko‘a Card by four trained observers.
Mean (n = 10 colors per fragment) color values for hue (reflected color), saturation (proportion of grey in the hue), and brightness (relative lightness and darkness) were quantified for each coral fragment. Additionally, mean color values of RGB and CMYK were also calculated for each fragment. Changes in mean color values (HSB, RGB) during the simulated bleaching event were analyzed using linear regressions. Mean color values (HSB, RGB) for each fragment were pooled by Hawaiian Koʻa Card number and compared to the color values displayed on the Hawaiian Koʻa Card. Agreement between the expected color values (from the Hawaiian Koʻa Card) were compared with the mean and SD of the color values obtained from the photographs via Adobe Photoshop.
PHASE II: card validation
Lab validation: coral health measurements
The laboratory validation was conducted in the same experimental tanks and under similar environmental conditions with the identical species (n = 6) but with fewer coral colonies (n = 3) and fewer fragments (n = 6) than the initial experiment. Only four of the 108 fragments were not used in the health analyses due to mortality. No disease was observed during experimentation. Temperature exposure was similar between experiments. Heating experiments were conducted between 23 October to 14 Nov 2018. As in the previous experiment, temperatures were elevated using titanium heaters (Finnex 800 W) at an increase of 0.25–0.5 °C per day. Corals were assessed daily using the Hawaiian Ko‘a Card and removed for health measurements (i.e., PAM, symbiont density, and chlorophyll density), replicating the initial parameter set used in the initial card development.
Field validation
Field validation of the card colors was conducted on shallow reefs on the island of O‘ahu in December 2018 (Fig. 3). To assure the selection of colors were representative of the majority of common Hawaiian coral species and their color and health condition, sites facing all cardinal and intercardinal directions were selected. Site selection criteria were based on the presence of symbiotic corals and site accessibility. Colors of corals were assessed using the underwater field Hawaiian Ko‘a Card prototype. Eleven sites were surveyed on reefs facing northward (site 1–3), eastward (4–6), southward (7–10), and westward (11 and 12) (Fig. 3). The color assessment was conducted by a pair of snorkelers at each site in depths ranging from 0.75 to 4.5 meters. Each snorkeler identified species and the most common color of a colony by placing the card adjacent to the coral and recording the corresponding number on the card most similar to the coral color. This color assessment was repeated with multiple individual colonies for the 1h survey duration at each site. Additional data, including observer, date, time and percent cloud cover were recorded. Field validation data were used to re-evaluate the prototype selection of representative colors of corals on shallow reefs.
Observer variation
Field surveys using a range of groups determined the extent of observer bias and error. A group of 20 marine biologists and students used the Hawaiian Koʻa Card to assess the color of nine coral fragments collected from the reef flats of Oʻahu, Hawaiʻi. The assessments were conducted under natural sunlight with corals presented in a black raceway tank (1.5 m × 0.75 m × 0.15 m). The Hawaiian Koʻa Card was submerged underwater next to the coral of interest, while the observer remained above water. The depth of water in which the corals resided in was ~ 0.10 m, minimizing the potential of light and color distoration. While assessments were in progress, internal water flow and aeration to the raceway ceased. Each observer independently determined the majority color of each coral fragment, avoiding tips and cut margins. Observer variation was based on the comparison of the selected color value by each observer to the color value that was selected most frequently for the coral fragment assessed. This will hereafter be referred to as ‘observer precision’. To determine if observer precision is influenced by any extraneous factors, the following was recorded by each observer: age, gender, coral familiarity (never snorkel, snorkel, know coral common names, know coral scientific names), highest level of education (high school, bachelors, advanced degree), and ease of use (easy, medium, difficult). Statistical analyses were performed in R (version i386 3.5.1) using a non-parametric Spearman’s rank correlation.