New Haven: Artifacts of Plans from a Cultural Past
The Semblance, Convergence and Resilience of the Nine Squares
New Haven, founded in 1638, was planned around a 3×3 grid of squares — one of the earliest planned urban layouts in North America. As the city developed outward from this core, the original grid morphed into different typologies: curvilinear suburban streets, irregular industrial networks, and highway-era superblocks. The question is whether — and where — the memory of the original grid persists in the modern street network.
This project measures how much each part of the modern city "remembers" that original grid by imposing an expanded grid derived from the nine squares and computing street pattern similarity per cell.
New Haven Neighborhood Boundaries Imposed on Cellular Analysis
acquisition
Acquire the full street network for New Haven from OpenStreetMap via osmnx. The network comprises 2,054 nodes and 5,305 edges with highway classification tags, projected to EPSG:32618 (UTM Zone 18N).
identification
Locate the spatial pattern of the nine-square grid by snapping to the 16 street intersections that define the 3×3 array (Grove, Elm, Chapel, and George streets crossed by York, College, Church, and State streets). The grid orientation is derived from street bearings: 24.6° / 117.4° off north. Average cell dimensions: 286m wide × 324m tall.
imposition
Using each unit cell derived from the nine squares as a module, impose an arbitrary expanded grid that lengthens outward from the nine-square center onto the full basemap of New Haven. The grid preserves the 24.6° orientation and the ~286m × 324m cell dimensions of the original squares, tiling the entire city boundary. This produces 799 cells, of which 589 contain enough street network to score.
scoring
For each cell of the expanded grid, clip the street network edges that fall inside and evaluate whether the internal street patterns follow the same spatial logic as the original nine-square grid. Four metrics compose the similarity score:
Compute the bearing of every street segment within the cell. Measure the angular deviation from the two reference bearings of the nine-square grid (24.6° and 117.4°). Streets closely aligned to either axis score high; diagonal or curved streets score low. Reference value: 0.932.
An entropy-based measure of how ordered the distribution of street bearings is within the cell. A perfectly orthogonal grid has high order (Φ → 1.0); a network with streets pointing in many directions has low order (Φ → 0.0). Captures skewedness and directional consistency. Reference value: 0.728.
The ratio of actual street path length to straight-line distance between endpoints. A perfectly straight grid has circuity = 1.0; curving streets have circuity > 1.0. Inverted for scoring so that lower curvature = higher similarity to the rectilinear nine-square pattern. Reference value: 1.000.
Mean number of edges meeting at each intersection node. A regular grid produces 4-way intersections (degree = 4). Cul-de-sacs and T-intersections lower the average. Measures how grid-like the intersection pattern is. The score normalizes to the nine-square reference.
The four metrics are combined into a weighted composite similarity score per cell, ranging from 0 (no resemblance to the nine-square grid) to 1 (identical spatial pattern).
visualization
Each grid cell is infilled with a color on a gradient proportional to its composite similarity score. Cells with high similarity to the nine-square grid receive a warm/bright tone; cells with low similarity receive a cool/dark tone. The result is a continuous heatmap of "grid memory" across the city — revealing where the original 1638 spatial logic persists and where it has been overwritten by later development patterns.
overlay
The visualization layers traffic density onto the grid memory map. CTDOT AADT measurements (236 unique routes in New Haven, matched to 2,752 OSM edges via 30m spatial join buffer) provide measured volumes. For the remaining 2,553 unmatched edges — predominantly residential and service roads — AADT is estimated using FHWA functional classification guidelines.
| OSM Highway Class | Estimated AADT (vpd) |
|---|---|
| motorway | 80,000 |
| motorway_link | 30,000 |
| trunk | 35,000 |
| primary (532 edges) | 15,000 |
| secondary (1,009 edges) | 8,000 |
| tertiary (474 edges) | 4,000 |
| residential (3,128 edges) | 1,200 |
| service | 300 |
Street edges are rendered with line width (0.15–3.15pt) and color (gray → amber → yellow) proportional to AADT on a log scale. High-volume roads (>25,000 vpd) receive a glow effect. Grid cells are drawn at 40% opacity with neighborhood boundaries in neon green.
- OpenStreetMap street network via osmnx — 2,054 nodes, 5,305 edges, EPSG:32618
- OSM city boundary — administrative boundary of New Haven
- Nine-square grid — bearing-derived, 24.6° orientation, 265m cells
- City of New Haven official neighborhood boundaries — 20 neighborhoods, ArcGIS Feature Service
- CTDOT AADT — 18,933 segments statewide, 236 unique routes in New Haven (2018–2024)
- CTDOT traffic monitoring — GeoJSON + GeoPackage, full statewide download
baseline
| Metric | Value |
|---|---|
| Bearing Alignment | 0.932 |
| Orientation Order (Φ) | 0.728 |
| Circuity | 1.000 |
| Grid Orientation | 24.6° / 117.4° |
| Avg Cell Width | 286m |
| Avg Cell Height | 324m |
| Total Grid Cells | 799 |
| Scored Cells | 589 |
& Spatial Culture
The nine-square grid — the standard American block structure organized around orthogonal arterials and expressways — is a spatial system that idealizes vehicular efficiency over all other modes of inhabitation. Its straightness and regularity optimize traffic throughput at the cost of the background network: the finer-grained, culturally-encoded circulation patterns that support pedestrian life, social interaction, and ecological function. [1]
The Federal-Aid Highway Act of 1956 authorized what was then the largest public works program in U.S. history. More than 475,000 households and over a million people were displaced by federal roadway construction. The neighborhoods destroyed were disproportionately Black and poor — by design. [2]
In Washington, D.C. alone, the construction of Interstates 395 and 695 consumed over 400 acres and displaced 23,500 people. In Miami, the expansion of I-95 destroyed 87 acres of housing in Overtown, reducing the population from approximately 40,000 to 8,000. [2,3]
The morphological consequence is fragmentation: previously connected neighborhoods severed into isolated cells. The highway acts as a hard boundary between morphological eras, destroying the spatial continuity that background networks depend on for cultural coherence. [4]
Boeing (2021) demonstrates that contemporary cities have seen declining entropy in street orientations as they developed car-oriented grids, replacing the organic distribution of streets and alleys of pre-industrial settlement. The nine-square grid is the spatial encoding of this decline. [5]
Consequential Cells
Salazar-Miranda et al. (2024) establish a causal link between 1930s redlining policies and present-day climate vulnerability: areas marked as less desirable for investment face disproportionately higher risks of flooding and extreme heat due to diminished environmental capital — reduced tree canopy and lower ground surface permeability. The morphological imprint persists for nearly a century. [1]
Hsu et al. (2021) demonstrate that the average person of color lives in a census tract with higher urban heat island intensity than non-Hispanic whites in all but 6 of the 175 largest U.S. urbanized areas. In 108 cities with historic redlining, 94% have higher land surface temperatures in formerly redlined areas. [2]
Nowak et al. (2022) quantify the disparity: tree cover averages 40.1% in historically graded-A areas and only 20.8% in graded-D areas; impervious cover is 30.6% in graded-A versus 53.0% in graded-D. The morphological cells that suffered the harshest deterioration are precisely those where highway construction and redlining overlapped — environments disrupted by road infrastructure, stripped of vegetation, and sealed with impervious surfaces. [3]
For each morphological cell: compute influence factor relative to adjacent cells (morphological distance); classify by formation driver (organic, policy-imposed, highway-severed, resistant); measure deterioration index (tree canopy loss, impervious surface increase, connectivity reduction, temperature differential); identify which cells resisted and retained cultural morphological signatures despite surrounding transformation.
The neighborhoods that resisted — maintaining density, connectivity, and cultural spatial patterns despite highway construction, redlining, and grid imposition — are the most consequential cells for understanding morphological memory. Their survival strategies encode spatial intelligence: the background network adaptations, building-to-street relationships, and social-use patterns that proved resilient against systematic erasure. [4]
Spatial Format
If morphological memory is the embedded spatial logic of a cultural era persisting in urban form, then its destruction — through highway construction, grid imposition, or policy-driven disinvestment — is measurable through the metrics established in the morphological memory framework: connectivity loss, block geometry disruption, and land-use pattern homogenization.
The question is whether morphological memory can be reinvented — not by rebuilding the demolished street network or restoring the original block geometry, but by introducing a new spatial format that recodes the deteriorated landscape. Tree canopy is proposed as this format.
Increasing tree canopy lowers urban air temperature by up to 1.5°C in heat-prone areas (Nature, 2025). The cooling efficacy of trees is determined by three factors: tree traits, urban morphology, and background climate — meaning that planting strategies must be morphologically informed, not uniformly applied. [1]
Street canyon geometry — the same metric central to the Favelas 4D framework — directly determines the thermal performance of tree plantings. Urban canyon width-to-height ratio and street orientation affect pedestrian thermal comfort, with spreading canopies planted at spacings allowing continuous cover producing measurably different microclimates. [2]
A greater number of street trees and presence of flowers increases rated restoration likelihood. Vegetated streets are rated as significantly better at reducing stress than non-vegetated streets. Visual diversity and colorful vegetation positively affect aesthetic preferences and perceived restorative potential. [3]
Tended greenways show better restorative effects than untended ones, indicating that perceived deterioration in the urban landscape is not merely a function of physical degradation but of spatial legibility — whether the environment reads as intentional, ordered, and cared for. Tree canopy provides this legibility. [4]
In 92% of U.S. urbanized areas, low-income blocks have 15.2% less tree cover and are 1.5°C hotter than high-income blocks. Street trees increase local biodiversity and biomass, but disproportionately in higher-income, denser neighborhoods — the morphological cells that already retained their spatial integrity. [5,6]
The disparity is not accidental. It maps directly onto the morphological cells that were most deteriorated by highway construction and redlining. Vegetation planting strategies targeted at these cells could simultaneously address thermal inequity, perceptual deterioration, and morphological memory loss.
Valid Spatial Format?
The proposition that tree canopy can serve as a morphological element — reconstituting spatial memory through a non-built format — is testable through the same quantitative framework used in the morphological memory project.
For cells that have undergone significant canopy planting programs: compute the morphological distance (Mahalanobis) between the cell's current profile and its historical profile (pre-deterioration), with and without canopy metrics included. If canopy-inclusive profiles show reduced morphological distance to the historical baseline, canopy is partially reconstituting the spatial character.
Compute perceived spatial enclosure, continuity, and legibility from street-level imagery (Google Street View time series) using the Urban Visual Intelligence framework (Zhang et al., 2024). Compare cells with high canopy cover to adjacent cells with equivalent built morphology but low canopy cover. If canopy-rich cells score higher on perceptual morphology indices despite identical built environments, vegetation functions as a spatial format.
Using Landsat thermal bands and GHSL temporal layers: compare the thermal signatures of deteriorated cells before and after canopy intervention. If canopy planting shifts the thermal profile toward the signature of the pre-deterioration era (or toward the signature of cells that resisted), canopy is recoding the microclimate dimension of morphological memory.
Measure whether canopy planting in deteriorated cells alters the morphological influence factor between adjacent cells. If canopy softens the hard boundaries created by highways — reducing the morphological distance between severed neighborhoods — it is functioning as connective spatial tissue, analogous to the background network it replaces.
// The validity test is whether tree canopy behaves morphologically — whether it produces measurable spatial effects at scales and in patterns that parallel the effects of built form. If it does, it is not merely an amenity or an ecological intervention but a genuine spatial format capable of re-encoding morphological memory.