MAGiC uncertainty training

This is the uncertainty component of the MAGiC system. It runs SIPNET at design points with the same meteorology, initial conditions, management events, and parameter priors as the ensemble workflow, and reports which of them the modeled soil carbon and nitrous oxide flux are most sensitive to.

Two analyses are run:

• A local (one-at-a-time) analysis moves each parameter in turn across its prior and reports each parameter’s share of the parameter variance.
• A global (Sobol) analysis varies parameters, initial conditions, meteorology, and management together and reports the share of output variance from each.

Each notebook is a self-study demo to work through beforehand, then discuss in the live session.

Topic	Audience	Objectives	Notes
Environment setup	Everyone	Conda, AWS, and repository setup	Needed before the demo
Uncertainty demo	MAGiC users	Run both analyses at two design points and read their output	Demo data covers two design points

Commands

Command	What it does
prepare	Stages the ensemble inputs into run_dir and builds the settings of both analyses.
local-sensitivity	Runs the one-at-a-time analysis, computes each parameter’s share of variance at each design point, and draws the figures.
global-sensitivity	Runs the Sobol analysis, computes the indices with bootstrap intervals at each design point, and draws the figures.

Data flow

flowchart LR
  ENS["ensemble inputs<br/>met, initial conditions,<br/>events, priors"] --> PREP["prepare"]
  PREP --> LOC["local-sensitivity"]
  PREP --> GLO["global-sensitivity"]

Audiences

Following the same roles as the MAGiC training index:

• Everyone: environment setup.
• MAGiC users: running the analyses and reading their output. Assumes the setup page is done and you are comfortable on the command line.