Developer Notes¶
Last updated: 10 August 2020
Reproducibility Files¶
- Store information for reproducibility. This should include the following:
version information (git hash of commit)
forecast filename
evaluation catalog (including necessary information to recreate the filtering properties); maybe just md5
do we need calculation dates?
Evaluation Results¶
Each evaluation should return an evaluation result class that has an associated .plot() method.
We should be able to .plot() most everything that makes sense including forecasts and evaluation results.
How do we .plot() a catalog?
Should we .plot() a region?
For serialization, we can identify the appropriate class as a string in the class state and use that to create the correct object on load.
Forecast metadata information¶
Forecast should contain metadata information to identify properties of the forecast
Start and end date
Spatial region
Magnitude bins
Working with GriddedForecasts¶
Right now, we can only spatial counts over the entire magnitude range. What if we wanted to have some control over this?
Might want to plot above some magnitude threshold or within some incremental threshold.
Should be able to have a method that returns a new GriddedForecast with specified parameters such as min/max magnitude.
Region information¶
The region information will need to accommodate more complex spaces including 3D areas and those with non-regular grids (e.g., quadtrees or meshes)