Create treelists from synthetic plot network data and reference treelists

[grovduck]

The major goals of this project are to:

1. generate synthetic plot networks in feature space (i.e. #1) based on reference plot locations
2. use the k nearest neighbors and reference treelists to generate new treelists for the synthetic plots
3. derive stand-level attributes from the synthetic treelists
4. use sknnr-spatial to run a k=1 imputation of pixels from synthetic plots using the feature space.

This issue addresses design choices needed to implement #2 given: 1) a synthetic network created from #1; 2) a treelist file associated with the reference plots used to build the synthetic network; and 3) a few optional parameters:

• k: The number of reference neighbors used to build a treelist for each synthetic plot
• d: The power term in inverse distance weighting of neighbors, where d=0 implies equal weighting, d=1 implies inverse distance weighting, and d=2 implies squared inverse distance weighting. Note that d can be a floating-point value as well.
• r: A threshold associated with censoring tree records when their weighted density does not meet or exceed a reference density usually associated with the plot inventory design. With r=0, all tree records are kept, whereas with r at higher values, fewer records meet these reference thresholds are are excluded.

Much of the work for this issue has already been captured in a gist and private repo outside this project and has generated some key discussion points to consider when moving forward:

• In early testing, avoiding the use of pd.DataFrame.apply significantly improved performance. With 10,000 synthetic plots for which to generate treelists, @aazuspan was able to reduce running times from @grovduck's initial attempts from 45s to 4s.
• In continued testing, using Polars, a dataframe library built in Rust, @aazuspan further reduced run times from 4s to 0.7s. When this was scaled up to 100,000 synthetic plots, treelists were generated in 6.2s.

We will lean on this code in implementing this functionality.

[grovduck]

@aazuspan, just documenting the work you've already done in synthetic-knn-censoring-demo. I have a horrible habit of losing track of where we've been working between repos and gists!

[aazuspan]

Thanks @grovduck, so do I. Just to get all the links in one place, here's the original gist and the demo repo (both currently private, but I think it's safe to say they won't be of much interest to the public).

[grovduck]

Just to get all the links in one place, here's the original gist and the demo repo

I had considered this, but I think you're right that there is no danger in providing those links!