The Trench

Reading the Landscape
Through Data Classification

A model built
on distance alone.

"They built the Cross Bronx Expressway right through the neighborhood in the 1950s and they never asked us anything. Now we live in the trench. The noise never stops. The air never clears. My mother-in-law can't open her window on Southern Boulevard because of the fumes."

West Farms resident, oral history interview 2025

Each extruded block on the map is one 100 m × 100 m grid cell. Height encodes the Random Forest model's predicted PM2.5 concentration — values were multiplied by 100 so that a 12 µg/m³ cell reads as 1,200 m tall. Height is pollution. The model was trained on 2,392 grid cells across the study area, using two input features: distance to the Cross Bronx centreline and STURLA land-use class.

The analysis reveals a critical insight: the trench geometry matters more than local land use. While STURLA classification helps us read the landscape, the Random Forest model shows that distance to the highway explains 100% of PM2.5 variation, while land cover type contributes virtually nothing (importance = 0.000). This suggests the concrete canyon's atmospheric trapping effect overwhelms local mitigation efforts.

Nine metres
below street level.

"You stand at the fence and look down and you realise — this isn't a road, it's a wound. They cut right through the fabric of place and left the edges to bleed. We've been breathing the bottom of that wound for sixty years."

West Farms resident, oral history interview 2025

The red glowing line traces the trench section where concrete canyon geometry prevents vertical dispersion. Diesel exhaust accumulates in this 4.8 km corridor, creating exposure levels 300% above EPA standards. The blue lines show elevated approaches—same trucks, different atmospheric consequences.

Construction between 1948 and 1963 required the demolition of 1,530 buildings and the displacement of approximately 60,000 residents. Robert Moses routed the expressway through the most densely populated neighbourhood in the Bronx because it was the cheapest land to acquire. The atmospheric consequences of that economic decision are still measurable sixty years later.

Three Data Streams,
One Reality

"The city built this and walked away. They put the highway where the people with the least power lived, and then they wondered why the neighbourhood collapsed. It wasn't negligence. It was a decision. The data just proves what we already knew."

Soundview resident, oral history interview 2025

The map now overlays three independent datasets on the PM2.5 model:

The convergence is not a coincidence. The clusters of 311 complaints, the FloodNet flood events, and the highest-urgency testimony nodes all stack onto the same grid cells as the tallest PM2.5 extrusions — precisely where the model predicts maximum exposure. The infrastructure is the evidence. The data just proves what the community already knew.