Research Notebooks¶

The repository holds 91 Jupyter notebooks. Eighteen are rendered as pages (see Notebooks); the 73 listed here are not, because they are research in progress, need a GPU, a CyVerse data mount or private credentials to run, or are too heavy for a static site (one is 72 MB). They remain readable:

• GitHub shows a notebook with its stored outputs: https://github.com/tyson-swetnam/fractal-notebooks/blob/main/<path>
• nbviewer renders it, including Plotly and HTML outputs: https://nbviewer.org/github/tyson-swetnam/fractal-notebooks/blob/main/<path>
• Colab opens it for execution: https://colab.research.google.com/github/tyson-swetnam/fractal-notebooks/blob/main/<path>

Paths below are relative to docs/notebooks/ in the repository. The execution state is what the stored outputs show, as inventoried on 2026-10-24: "executed 9/28" means 9 of 28 code cells have saved outputs. The environments are described on the Notebooks page.

3DEP canopy height models¶

3dep/: canopy height models (CHM = DSM − DTM) from USGS 3D Elevation Program lidar for US forest research sites, and fractal analysis of the resulting surfaces. forest_sites.yaml defines 17 sites (bounding box, expected maximum height, priority) and the processing rules: 0.333 m resolution at 12 or more points per m², otherwise 0.5 m; DSM = maximum and DTM = minimum per cell, no smoothing; Cloud-Optimized GeoTIFF output. Point clouds are read anonymously from the public USGS Entwine Point Tile bucket on AWS, so no credentials are needed, but the notebooks write to CyVerse Data Store paths (/home/jovyan/data-store/...). Environment: 3dep (environments/3dep-environment.yml). The directory README documents the site list and the utility modules.

Workflow notebooks (modified)¶

Helper scripts in the same directory, formerly built as pages, are linked on GitHub only: generate_site_notebooks.py, utils/cog_utils.py, utils/validate_cogs.py and utils/update_notebooks_cog.py.

Upstream tutorials (original)¶

Reference copies of the seven OpenTopography OT_3DEP_Workflows notebooks; 05 is the basis of the modified CHM workflow. They expect Colab (condacolab) or the upstream 3dep conda environment, and call USGS National Map services as well as the EPT bucket. The upstream repository is published under the GNU GPL v3.0, so these copies keep that licence rather than this site's CC BY 4.0; whether they are still byte-identical to upstream has not been re-checked.

Smithsonian: Barro Colorado Island¶

smithsonian/: fractal analysis of the 0.5 m canopy height model of Barro Colorado Island, Panama, from the Smithsonian Tropical Research Institute's ALS Panama 2023 airborne laser scanning campaign. The notebooks read the 1.1 GB CHM GeoTIFF from a CyVerse ~/data-store mount (not included in the repository) and ran on CyVerse on 2025-12-24 and 2025-12-25. Environment: 3dep. The project plan is kept in the repository at notes/notebooks/smithsonian/PLAN.md.

Sierra Nevada Giant Forest¶

sierra-nevada/: planned analysis of the 2022 USGS 3DEP lidar (CA_SierraNevada_B22) over the Giant Forest in Sequoia National Park, following the BCI workflow but starting from raw LAZ tiles. Two of the five planned notebooks exist and neither has been executed (their download and processing cells are commented out, "uncomment to run"). Needs PDAL and GDAL (3dep environment). Plan: notes/notebooks/sierra-nevada/PLAN.md (status: not started).

Reviewer companions (Part II)¶

metabolic-scaling/review/: seven notebooks that accompany a critical review of the Part II manuscript, REVIEW_metabolic-scaling.md at the repository root. Unlike the tutorials below, they exist to test the paper's claims section by section. They share estimators, the data pull and the equivalence test through fractal_review_utils.py, so launch Jupyter from this folder. The empirical cells fetch the author's private CyVerse iRODS collection and need your own credentials (gocmd init); without access they fall back to deterministic synthetic objects and say so. The stored outputs show the synthetic path only. Environment: the scientific Python stack (the directory README points at fractal-foundations-gpu).

Metabolic scaling tutorials¶

metabolic-scaling/1_* to 5_*: five complete teaching notebooks (2025-12-24) on metabolic scaling theory. Every cell ran in order with no errors, and each has four to seven figures. They are not in the site navigation because the Part II methods and results notebooks superseded them as the manuscript's companions. Environment: the scientific Python stack (python3 kernel; any of the project environments).

GPU research notebooks¶

foundations/: DLA and growth-model experiments that need the fractal-foundations-gpu kernel (CUDA 12.2, CuPy, PyCUDA, Numba). Design notes for the CUDA work are kept in the repository under notes/notebooks/foundations/ (dla_cuda_plan.md, PLAN_lichen_growth_updates.md, lichen_dla_claude.md).

Blender point-cloud tools¶

foundations/blender/: 3D DLA simulated with Blender's Geometry Nodes and rendered with Cycles. Six Python scripts run inside Blender (scene setup, advanced features, export to NumPy, PLY, OBJ and CSV, batch and gallery rendering); the notebook analyses the exported point clouds. The user-facing guide is Blender DLA; the directory README documents the scripts and presets.

Scripts (GitHub): dla_blender_setup.py, dla_advanced_features.py, dla_export.py, dla_batch_render.py, dla_example_renders.py, dla_gallery_render.py.

Work in progress¶

foundations/: three notebooks that do not run cleanly and stay unpublished until fixed. Two notebooks that older links point at are gone: torch_test.ipynb (a PyTorch GPU smoke test) and old_dla.ipynb (superseded by the published DLA notebook) were deleted from the repository.