ConnectedSystems
commented
2 years ago HI @KRNA01
Could we make coralScenario() return just the raw data (i.e., what is currently kept inside all), and calculate the other aspects after the fact?
The current data structure is a little problematic for larger set of runs and increases file sizes when storing results to disk. One implication is larger transfers to/from a remote computer (e.g., workstation or HPC).
My current envisioned approach is shown in #58 but repeated here for convenience:
Here, Y would be equivalent to all.
% Run a set of simulations with sampled values of size N*D, where N is number of rows,
% with each row holding values for a simulation.
% As `nreps=8` here, this would produce a result set of N*8 in total
Y = ai.run(sampled_values, sampled_values=true, nreps=8)
% Collect a set of metrics by passing in an array of functions
a_struct_of_separate_metric_values = collectMetrics(Y, ...
[@some_metric_function, @another_metric_function, @yet_another_metric])The other concern I have currently is that there is a lot of hardcoded values. Are we absolutely definitely 100% sure we won't be adding/removing coral groups/species? Okay if we're not, but otherwise I would prefer that we generalize as much as possible.
KRNA01
Rosejoycrocker
Dear @ConnectedSystems , @Rosejoycrocker, @veroniquelago and @BarbaraRobson,
We now have a small but robust subset of metrics we can use to inform ecosystem service provision from ADRIA simulations.
Jointly, these four metrics are a significant step towards building a Reef Condition Index, which speaks, in part, to existence, tourism and fisheries values. The new coral function in ADRIA allows us to produce these performance metrics. The advantage is that our outputs can now be translated to benefits for reef and people to a greater degree than coral cover can achieve on its own.
I'm adding this issue here mainly as a heads up, but also to start a discussion about how far we take these metrics. Metrics for ReefMod are richer because there are more variables to play with (e.g. crown-of-thorns starfish, algae and rubble), but also more uncertainty from the expert elicitation of derived metrics (fish species richness etc). My suggestion is we keep translating ReefMod data using these derived metrics, but then keep it more robust for ADRIA model simulations - i.e. smaller but less uncertain set of metrics. Keen to discuss.
For your interest, here's an example output of these four metrics for a single_scenario.m run in the 40Corals branch using default values. The climate scenario is RCP 4.5, enhanced tabular Acropora and enhanced corymbose Acropora are seeded, but without assisted adaptation. Trajectories are for individual reefs. Note how coral evenness (a proxy for functional diversity) peaks mid-way for a subset of sites. Shelter volume is a proxy for structural complexity and scope for habitat provision (think trees in forests).
And here are the trajectories for the six coral groups making up the metrics in the first four panels. The dynamics of the six size classes are not shown here. Note that the enhanced groups (enhanced tabular and corymbose Acropora) only start with seeded juvenile (2-5 cm diameter) corals, whereas the other groups start with a mixture of size classes.
Looking forward to working with you be able to produce runs for a range of interventions, scenarios, reef settings and R&D + intervention assumptions :-)