Sprint Plan: 04-08-2026

[msteckle]

Sprint Goal	Produce updated ILAMB-H2O results to display how H2O downscaled products perform compared to CMIP6 GCMs
Definition of Done	Tested and implemented new metrics & display in the ILAMB-H2O results Dashboard
Sprint Length	1 month
Attendees	Morgan S., Nathan C., Shih-Chieh K., Min X., Deeksha R.

Retrospective Recap

• Created ilamb3/compare library with neighborhood.py and base.py modules
  • neighborhood.py
    • extract_sites_closest(): Select the closest grid cell in the dataset to each site
    • extract_sites_mean(): Compute the mean grid cell in the dataset from a neighborhood around each site
    • extract_neighbors_by_window(): Return closest neighbors from the source dataset to each site in the target dataset
    • neighborhood_mean(): Reduce the neighborhood by taking a mean of the non-null cells
    • neighborhood_closest(): Reduce the neighborhood by selecting the closest non-null cell
    • match_label(): Apply a mask to the neighborhoods for cells that do not match the target values of label
  • base.py
    • Fixed site extraction logic in make_comparable()
    • Removed old extract_sites() function
• Tested the new neighborhood module
• Created ilamb3/transform that can be used in config label.py module that applies a label to the input dataset by NN interpolation
• Added default site extraction logic to ilamb3/config.py
  • "site_extraction": {"method": "closest", "window_half_width": 2.0, "window_shape": "circle", "scale_width_by_grid_resolution": 2.0, # Can also be None to disable}
• Created data contribution doc in ilamb3-data repository

Parent Issues

• More metrics for ILAMB-H2O #126

  • Data: ERA5, Daymet (hosted on OLCF, not reformatted for ILAMB), Runoff
  • Code:
  
  • Dashboard:
    • Consider adding new plots from Moet's paper

Notes

• This work is with the Hydropower and Hydrokinetic Office (H2O)
• One general goal is to ensure ILAMB can handle CMIP7 when they're released
• Implement all functions from Table 2 of Moet's Paper
• Goal of the study is to help users determine which GCMs to use
• H2O needs a working product by Summer

Action Items

• Pre-schedule the next Sprint Retrospective & Planning Meeting w/ Deeksha
• Temporarily move stand-ups from Slack to Microsoft Teams for accessibility
• Schedule a time to write Issues and allocate tasks in ILAMB Collective kanban

[msteckle]

20 Apr 2026: Follow-up meeting for reviewing Issues and allocating tasks

• Broke Min's More metrics for ILAMB-H2O #126 into sub-issues
  • Expand the annual cycle analysis to include amplitude #158 --> Add amplitude to existing annual cycle analysis
  • Add seasonal variants to bias, RMSE, and spatial distribution analyses #159 --> Break up bias, RMSE, and spatial distribution analyses by xarray built-in .season OR date range (user can specify which seasons)
    • Add new sub-issue to create a Seasons class similar to Regions. So any func using time dimension could accept a region object
    • Add a dispersion analysis module/transform #160 --> Add a dispersion analysis module/transform to compute IQR, skewness, ...
      • A means to select a time frequency #156 --> We need an analysis to be able to specify what kind of time frequency data it needs (e.g., daily); we want to avoid monthly and daily versions of a variable being loaded into one xarray
    • Add a transform for counts where a condition is true #161 --> Count the number of times that a condition is satisfied
    • A transform that computes quantiles #157
• Consider writing documentation on transforms while we're at it

Action Items:

• Nate: start working on A means to select a time frequency #156
• Morgan: start working on Add a transform for counts where a condition is true #161
• Min: start working on A transform that computes quantiles #157