List view
This is the critical month. We will produce a proxy layer boundary (spatial data), review and make the GH repo public (which means tightening up the reproducibility), and produce a technical memo that summarizes 2022 Q1 work and charts next steps for 2022 Q2. At the outset, we should have all staged data in place, and EDAs underway to refine data cleaning, feature engineering, and model building. The effort here splits into two tracks. In one track, people are focused on getting the repo to a presentable state for the client and public, which may in include starting the technical memo, Dockerizing the repo, wrapping things into a `run.py` pipeline, and so on. The second track is focused more on prototyping and implementing the model for any remaining water system boundaries. Both efforts should align with a finalized preliminary model and writeup by the end of the month.
Overdue by 3 year(s)•Due by March 24, 2022•2/2 issues closedFebruary is the month to get our ducks in a row. There are many stories about the origin about the idiom of "getting one's ducks in a row" but my favorite is the shipbuilding interpretation: To develop the lines of the hull for building ship frames, a "spline" is used. Small, metal weights with a bill-shaped protrusion are used holds the spline in place, and these weights are called "ducks" after their duck-like shape. The ducks are lined up to delineate the smooth curvature of the ship's hull, and shipbuilders progressively build the ship around this row of ducks. Likewise, our ducks this month are the minimal set of code (e.g., downloaders, transformers, models) that create a pipeline with raw data on one end, and a proxy boundary on the other end. This is the goal out efforts should service. Let's aim to line up Downloaders (i.e., for all labeled/unlabeled data, and model features - predictors) and Transformers (for labeled/unlabeled data, and feature engineering like "road_density") for a minimal subset of states (11, including OK) to support EDA and model building prior to the March push where we finalize a preliminary model and outline next steps. Let's go wide and shallow in February to get everything in place for an end-to-end extract-transform-model pipeline, then go deeper in March. When deciding on what to prioritize, ask, "does this advance us closer to the full, minimal set of features in staging to inform EDA and build a prelim model?" If the answer is no (e.g., write another custom downloader for another state when we already have a minimal subset of state data with which to build a model), then punt this task to a running list of Phase 2 (2022 Q2 and beyond) objectives.
Overdue by 4 year(s)•Due by February 21, 2022•2/2 issues closed