Vision#

Design principles#

Like MIKE IO and FMSkill, FMDAp should be:

  • Easy to use

  • Easy to install

  • Easy to get started

  • Open Source​

  • Easy to collaborate​

  • Reproducible

  • Easy access to new features

Functional objectives#

FMDAp should cover the following user stories:

Is my DA scheme performing well?#

As a user, I want to detect…

  • model instabilities

  • non-physical behavior

  • DA over-fitting (big difference between forecast and analysis)

  • DA under-fitting (little impact of assimilation)

  • non-Gaussian behavior (e.g. biases)

  • auto-correlation in innovation/increments

  • observation offset/biases

Visualization#

As a user, I want to plot…

  • observations on map showing the model domain

  • diagnostic positions on a map showing the model domain

  • multiple timeseries together, interactively

As a user, I want to

  • make all the static plots for my report/paper

  • make interactive plots for…

    • troubleshooting

    • DA calibration

    • DA assessment (exploratory; setup phase)

Input/settings (pre-processing)#

As a user, I want to…

  • create MEASUREMENTS section from xlsx file

  • create MODEL_ERROR section from xlsx file

  • check if measurements are inside domain, otherwise find nearest point inside

  • Estimate AR(1) half life from observations or forcing data

  • Estimate spatial correlation length scales

Automatic reporting#

After doing a DA simulation, I would like to get a report with:

  • summary of DA settings

  • map showing domain and observations

  • timeline showing simulation and observation coverage

  • summary skill table

  • summary bar chart

  • summary scatter plot

  • For each observation:

    • info/stats