Presets
Playback
Climate, heatwave and coral biology
Camera
Layers
Quality
Model check
The science
What you are looking at
A reef patch about 20 m across, 4 to 6 m below the sea surface. Every colony is a Blender-modelled mound, branching or table coral with baked pore textures; the colour of each colony is the live symbiont density S of the model (golden-brown when S is near 1, chalk-white when S is near 0). The number of live colonies is the coral cover C times 120 slots; a colony that dies stays as a skeleton that slowly turns to algae-green. The time slider replays 20 years at about 0.7 years per second. Fish are illustrative: their count follows C and they have no population dynamics of their own.
Heat stress
NOAA Coral Reef Watch defines the HotSpot as sea-surface temperature minus the Maximum Monthly Mean (MMM, here 27 + 1.5 = 28.5 °C). Degree Heating Weeks add up the HotSpots of at least 1 °C over the previous 12 weeks (°C-weeks): DHW ≥ 4 means bleaching is likely, DHW ≥ 8 severe bleaching and mortality. The SST here is a seasonal sine (27 ± 1.5 °C), a linear warming trend and one Gaussian marine heatwave at year 10.
Biology (phenomenological)
Symbiont density: dS/dt = ρ(S+s₀)(1−S) − κ·H·S with H = max(0, SST − MMM − 1 − tolerance). Cover: dC/dt = r C(1−C) − m·max(0, 0.5−S)/0.5·C. Both are integrated with RK4 at a quarter-week step by exactly the same code as on the 2D page, copied verbatim, so the numbers on the two pages are identical. The bottom-right graph shows the steady state S*(H), the positive root of ρ(S+s₀)(1−S) = κH S; the real S lags behind it because recovery takes weeks.
Honest simplifications
One site and one coral guild; no light stress, acidification, disease or reef-scale connectivity; constants are rounded and not fitted to a real reef. The scenery (colony positions, fish paths, caustics, light shafts) is decoration driven by the model, not part of it. Nothing here is a forecast for a particular reef.
Checks on this page
(1) Four closed-form checks of the model (constant-hotspot DHW, analytic logistic cover, steady-state symbiont root, RK4 step halving). (2) The instance buffers actually sent to the GPU are read back at three times: the live-colony weights add up to C×120 and the symbiont fraction recovered from the drawn colours equals S. (3) Rays cast onto the Blender seabed mesh confirm the fish stay above the bottom.