The raw trips maps is now corrected!! Every arc in this map of Santiago is coloured by how much the raw mobile network data was wrong. Red means under-counted. Blue means over-represented. Width is corrected trip volume.

The striking thing is that the map is overwhelmingly red. Not just at the periphery, where sparse tower coverage is expected to introduce bias. Even the thick, dominant corridors are red.

The reason is likely geometry. The flows that rank highest visually are the ones that combine high volume with long distance, and long-distance commuting in Santiago almost always has at least one endpoint in a less-covered area outside the dense urban core. That is exactly where tower density drops, detection probability falls, and the network starts to miss trips.

This is the corrected origin-destination matrix for the Santiago metropolitan region (the raw one is here), derived from mobile phone (XDR) data and reweighted using inverse probability weighting (IPW) to account for heterogeneous tower density. The colour of each arc is the log-ratio of corrected to raw trips for that specific corridor, *not the trip count itself*.

We do this here:
– Ferres, L., & Elejalde, E. (2026). Systematic biases in mobile phone mobility data from heterogeneous tower density. Zenodo. https://doi.org/10.5281/zenodo.19484460

I’ll write more about the statistical fraamework we used, but the detaills are there and in the associated github.