Table A1. Assessment of computational speed, difficulty, and accuracy of exposure assessment methods.

The table reports whether for our data analysis the methods were executed in less than a few hours on a standard PC (“Fast”), whether they ranked near the best cross-validation error (“Accurate”), whether they were fit with standard software without many tuning parameters (“Easy”) and whether predictions were paired with measures of uncertainty (“Uncertainty”).

Method	Fast	Accurate	Easy	Uncertainty
OLS	✓	✕	✓	✓
IDW	✓	✓	✓	✕
Universal Kriging	✓	✓	✓	✓
Downscaler	✕	✓	✕	✓
Random forests	✓	✓	✓	✓
Support vector regression	✓	✕	✓	✕
Neural networks	✓	✕	✕	✕