How it works

The science behind the prediction.

MAPP reads fluorescent signals from the algae living inside coral tissue. Those signals reveal how well the cells are coping with light and heat stress. Our models turn them into a thermal-tolerance estimate for each colony.

01 · The instrument

MAPP: Multispectral Aquatic Photosynthesis Profiler

MAPP is a submersible fluorometer that non-destructively measures a coral colony's physiological health in minutes. It excites the coral's symbiotic algae at four wavelengths (~420, 448, 470 and 505 nm) and records the chlorophyll fluorescence they produce.

The production design is a compact, wrist-mounted unit with a flexible fibre-optic probe, built for use by divers and snorkelers in the field. Software and online training videos are included.

Reserve a unit from the first production run →

Rendering of the production MAPP unit: a blue wrist-mounted housing with a flexible fibre-optic probe
Five generations of the light-response profiling system, from a 2021 benchtop prototype to the final production design

From the 2021 prototype (A) through the 2022–2024 upgrades (B–D) to the production design (E).

Development

Five years of field-tested refinement

With research and development funded by the National Science Foundation, National Geographic Society, NOAA and CORDAP, the instrument has moved from a benchtop prototype to a field-ready, submersible tool built to be easy to use for researchers and restoration practitioners.

  • 2021

    Benchtop prototype proves the light-response method on coral fragments.

  • 2022–2024

    Successive upgrades take the system underwater and onto reefs in Fiji, Guam and Florida.

  • 2027

    First production units ship (June–August) in a submersible housing.

02 · The measurement

One standardised, five-minute light-response run

Every MAPP measurement follows the same protocol. The light source steps through a dark baseline, three light levels and a dark recovery, while MAPP continuously fires flashes and records the fluorescence response at each of its four excitation colours.

StagesWhat happens
1Dark-adapted baseline: the reference every other stage is compared to.
2Multi-turnover flash drives photosystem II fully closed.
3–9Light step L1, ~300 µmol m⁻² s⁻¹
10–15Light step L2, ~100 µmol m⁻² s⁻¹
16–22Light step L3, ~500 µmol m⁻² s⁻¹
23–27Recovery in the dark

At every stage, single-turnover and multi-turnover flashes are fitted as curves, giving quantum yield, antenna size, connectivity, relaxation kinetics, and photochemical and non-photochemical quenching.

Irradiance program over a five-minute run, and example chlorophyll fluorescence transients from single- and multi-turnover flashes

Irradiance program (top) and the chlorophyll a transients captured at a single sample time point (bottom).

03 · Data Hub

From raw scans to ~4,200 named metrics

Upload the raw files from the instrument to the cloud and processed, analysis-ready data comes back in about ten minutes. Each measurement expands into six families of documented metrics:

FamilyWhat it captures
Base metricsEach fitted quantity (yield, cross-section, quenching, kinetics…) at every stage and colour.
Light-stage summariesMean, min, max, range and slope of a metric across a whole light phase.
Light-response ratiosFractional change between two individual stage readings.
Stage-average ratiosFractional change between two whole-phase averages.
Induction–relaxation pointsTurning points of the raw fluorescence trace itself, independent of curve fits.
Channel ratiosMaximal fluorescence compared across excitation colours.

The Data Hub also hosts curated reference datasets from partner reefs, so your corals can be compared directly with published collections. Every metric name, with a plain-language definition and its exact formula, is documented in the app's Metric Reference ↗.

04 · Predictive models

Trained on real bleaching outcomes, tested on new reefs

Each model learns from MAPP scans taken before a thermal-stress event, paired with how those same corals actually responded, whether in an experiment or a natural heat wave.

Data prep

Align metric panels across reef sites; drop columns with excess missingness, no variance, or cross-site drift.

Feature selection

Boruta stability selection across repeated subsamples keeps only consistently informative metrics, then prunes redundant ones.

Training

A small, regularised neural network (Keras/TensorFlow) is fit to the training reef's data.

Validation

Settings are tuned by nested, leakage-aware leave-one-out cross-validation, with features re-selected inside every fold.

Cross-site test

The final model is scored once on a second reef site it never saw, the closest proxy for performance on a new reef.

05 · Analytics Studio

Turn predictions into restoration decisions

Browser-based tools that work on your own samples and on the curated reference datasets.

Bleaching predictions

Estimate a thermal-tolerance outcome for every uploaded colony, shown as a sortable ranking.

Functional analysis

Cluster colonies into groups that share similar photosynthetic strategies and light-stress handling.

Photochemical profiles

Compare light-response curves and derived metrics side by side across sample groups.

Group comparison

UMAP or NMDS ordination shows how similar nurseries, sites or genotypes are overall.

Try it with the reference datasets

The Data Hub and Analytics Studio are open in your browser. Allow up to a minute for the app to wake.

Open the app ↗

Want MAPP for your program?

Reserve a unit from the first production run. It's free, and nothing is charged until you confirm.