Bundle review table

t03 — Wind farm power density is a packing problem

notes.md

# t03 — Wind farm power density is a packing problem

## Design

US Wind Turbine Database v9.1, continental-US turbines grouped into projects
of at least five turbines (n = 1,217 farms). Farm area is the convex hull of
turbine coordinates; nameplate power density is project capacity over area
(MW/km²); packing is turbines/km²; rating is MW per turbine; specific power is
capacity over swept area (W/m²). The key test (P3) decomposes log power
density, which equals log packing plus log rating exactly.

## What the data show

Packing dominates. The variance of log packing is 0.376 versus 0.050 for log
rating — a ratio of 7.5 (bootstrap 95% CI 6.0–9.3). Power density correlates
with packing at 0.879, with rating at −0.007 and with specific power at 0.462.
Nameplate density is monotonic in spacing (rho −0.572): there is no interior
peak.

## Meaning

The thesis of the posted claim — that power density is set by spacing rather
than by how powerful each turbine is — holds emphatically for real US wind
farms. Farms differ far more in how tightly they pack turbines than in per-
turbine rating. The "underperform" and "interior peak" consequences cannot be
tested with this source, because USWTDB records nameplate capacity, not
generated energy; those predictions need capacity-factor data and are reported
as inconclusive rather than failed.

## Caveats

Convex-hull area overstates footprint for sparse or multi-cluster projects.
Specific power is a geometric proxy, not measured aerodynamic efficiency.

Included code and results

• tests/t03/code.py
• tests/t03/results.json
• tests/t03/series.csv