Exploratory spatial association

Tree-pollen potential and socioeconomic conditions

A tract-level examination of whether NYC allergenic tree-pollen source potential varies with vulnerability, income, poverty, housing, age, race/ethnicity context, and historical HOLC grades.

The outcome is ln(1 + ATPI raw potential per land km²), not the capped map score. Associations are ecological and spatial; they are not causal effects or individual allergy risk.

—2020 analysis tracts

—individual trees spatially matched

—strongest Pearson r

—strongest Spearman ρ

Correlation overview

Pearson measures linear association; Spearman measures rank-order association. Intervals are tract bootstrap estimates and q-values use permutation tests with multiple-testing correction.

Pearson rSpearman ρline = 95% CI

Scatter & quadrant analysis

Median splits define four descriptive quadrants. Change the factor, outcome, or borough to test whether the pattern is stable.

FactorOutcomeBorough

Historical HOLC comparison

ATPI potential density is compared across A–D grades. Uncovered tracts and nonstandard grades are excluded from this panel.

Methods, exclusions & interpretation

Geography

All 2015 tree points were spatially reaggregated to 2020 Census Tracts before joining ACS 2023 and HOLC fields. Tracts with zero population or ≥90% water were excluded.

Statistics

Income and home value are log-transformed. Adjusted r removes borough means and the linear contribution of total tree density. This is a robustness check, not a causal model.

Limitations

Street trees omit parks and private land; ATPI uses transparent evidence weights and DBH proxies. Spatial autocorrelation, ecological fallacy, and HOLC coverage selection remain.