There are two primary missions to our Coral Monitoring program at UWC-HK: first, to transform our students into skilled and environmentally-conscientious citizens, and second, to acquire longitudinal data of coral health at the Hoi Ha Wan (西貢海下灣) marine park.1 The compiled reports go to inform policies, locally at AFCD2 & WWF, and globally to Reefcheck and Coral Watch, eponymous3 non-profit organizations that compile global data on ocean health.
The reef-check protocol is an extensive operation, where divers sequentially
- Identify a site, an optimal band of coral growth
- Lay a transect line, where a measuring tape is drawn and temporarily anchored over 100 m
- Monitor fish population, including their species, sizes, and types
- Monitor invertebrate population, including their species, sizes, and types
- Monitoring the substrate (more details in this post)
- Acquire supporting video and photos
The underwater data is then fed into a standardized spreadsheet by the Reefcheck organization, which outputs some standard graphs to help understand what goes on in the water. The standard representations for substrate — what’s on the sea-bed at 50 cm intervals — is shown as follows:
Besides not being visually attractive, the standard graphs all focus on the average and omits all details of which the average is comprised of. As an example, the above graph is usually interpreted as sand (SD) present between gaps of coral (HC); it could also mean a perfect lawn of coral, but the transect line veered half way to deeper waters where there is nothing but sand.
There is a second, less conspicuous drawback, which we realized only when we were processing the dives after the term ends. The bar-graphs were simply no good at telling the story of what we saw. All the divers recognized clearly the difference between Site 5 and Site 2 underwater: the same people, however, were hard-pressed to match this impression to individual bar-graphs.
The essence of longitudinal monitoring is to observe changes. The graphs — with their ten categories, lengths, arbitrary colors, and uncertainties — are not very good for compare and contrast during the analysis, and not compelling when putting together a narrative. Consider the following bar-graph of Site 3: what difference would you pick out?
Considering that we want to be able to compare between different sites, and to monitor how these sites evolve over Fall and Spring across multiple years, it seems that something else is needed.4
A brief interlude: the team this year had some fantastic students. The following owes a large part to Nancy Hui (HK), Jolie Lau (HK), Tech Tana (Thailand), and Norbert Monti (Hungary).
Specifically, we wondered if it would be possible to lay out an “abstract transect”, where each 0.5 m is represented by a box, and the boxes stack side-by-side:
…clearly marked off at intervals by a length marker:
Each box can then be color-coded to represent a different kind of substrate, and the transect tells a story of what the divers saw.
Choosing the color scheme took us many iterations, and we are still not sure if we have picked out the most useful scheme. We thought we would mark off similar “things” with a similar key, such that sand resembles silt/clay, and rock resembles rubble; these inorganic substrates are represented additionally with stunted bars. Hard and coral coral resembles one another, and the blue — together with the difference in length — makes it stand out clearly. Sponges and algae, non-coral growth forms, were colored in green — in hindsight I suspect that others is less important than the bright red we assigned it, and nutrient-indicator algae (NIA) should really take on more importance.
How does this first iteration work? Pretty admirably I’d say — the following picture encapsulates all elements of our Fall dives, and it’s easy to pick out the similarity and differences between each site.
It is also possible, down the line, to align the data isometrically to “see the flow of time”.
So far all of these visualization are manually constructed, really tedious, and no fun at all. At some point I’d like to sit down to first automate this (which would allow the visual representations to be optimized), and then make it a browser tool such that one can simply dump a Google Spreadsheet with raw reefcheck data and it plops out a visual. That would be spiffy. What would be even more spiffy is if there can be a link directly to a transect photo upon mouse-over… or perhaps even time-code to a transect video. That would be really spiffy.
In the long run, there is probably some way of integrating Fish and Invertebrate data into the “abstract transects”. I don’t know what that is yet. If you have any ideas I’m all ears.