ILAMB

1. Driver Script

1.1. Parallelization

1.1.1. mpi4py + model/confrontation pairs

1.1.1.1. Rework HTML output

1.1.1.2. Could avoid by storing in memory, need to think more closely about storing less, more confrontations means more memory

1.1.1.3. Move plots to 2 routines, 1 for inline, 1 for from files

1.1.2. reduce for correct plot limits

1.1.2.1. Not reduce, plots that can be done inline do, those that can't do from reading the netCDF4 files

1.2. Run from configure file

1.2.1. XML format?

2. GPP Confrontation

2.1. Restore functionality after Variable

2.1.1. Time integrals

2.1.2. Space integrals

2.1.3. Cycle

2.1.4. Phase

2.1.4.1. how to plot the std?

2.1.4.2. Fixed bug in my old phase computation

2.1.5. Variability

2.1.6. Problems

2.1.6.1. How to handle period means

2.1.6.2. Plot colorbars

2.2. Define a "layout" object which Html output can understand

2.2.1. Detect if no regions, don't add that bar

2.3. Precipitation

2.3.1. Add correlation routine/plot to Variable

2.3.1.1. returns two arrays and a correlation coefficient?

2.4. Todo

2.4.1. Fix units and/or period means

2.4.2. Fix plotting problems in maps

2.4.3. What metrics are missing?

2.4.4. Replace tooltips

2.4.5. Create bias legend

2.4.6. Add the benchmark back

2.4.7. Add links to the data files

2.4.7.1. What else can we add to these files?

2.4.8. Add section for all model plots, say, for annual cycle

2.4.9. Restore max time +std to cycle plots

3. Fix List

3.1. Variable.convert is fragile, improve

3.2. Weed out unused/unneeded functions

3.3. Docstrings are incomplete or wrong/missing

3.4. Make doc site update via a git push

3.5. Update image to include ILAMB.Variable

4. Open Questions

4.1. Is Basemape.interp faster than ilamblib.NearestNeighbor?

4.2. How do we handle/detect different time scale data?

4.3. How do we pass metrics to html output?

4.3.1. A: as variable objects in a dictionary

4.4. How can I include two time series in 1 netcdf file? Say, time and annual cycle?

4.5. Masks instead of regions is easy, what does the interface look like?

5. CO2 Confrontation

5.1. Can we make GPP general enough to replicate for Mauna Loa?

6. Tutorials

6.1. How to use Variable to do your own analysis

6.2. How to move your analysis into a confrontation

7. Design Principles

7.1. ilamblib should work on arrays and not depend on any objects we create. This way, users can work at whatever level they wish

7.2. We want to deliver a product that compares models to benchmarks, but to make this easier we really need to unify the way people write analysis. Makes things quicker to write, easier to debug