01
Find
Find similar climates
Start with the target location and identify comparable climates from a global city dataset.
50,000 citiesClimate dataPythonSimilarity analysis
Project Michigan
Climate data as a starting point for architectural design.
Year2026StatusCompleteCategoryMaster's Project · Barcelona
Developed withBrooklyn Edsall
Data AnalysisMachine LearningComputational DesignParametric DesignOptimisationData VisualisationPython
A computational design workflow that finds places with similar climates, learns from their vernacular architecture, and translates those strategies into parameters for new climate-responsive designs.
Overview
How do you design for a place when there is no local vernacular architecture to learn from?
Project Michigan was developed as a computational workflow that searches globally for places with similar climates, learns from the architecture that evolved there, and translates those patterns into parameters for a new design. The project connects climate data, computational analysis and architectural optimisation to provide an evidence-based starting point for climate-responsive design.
Process
02
Match
Find the strongest analogues
Analyse the shortlisted cities to identify groups with genuinely similar climatic conditions.
PCAK-MeansUnsupervised LearningKNIME
03
Learn
Extract vernacular strategies
Study vernacular buildings from the matched locations and translate recurring strategies into measurable design parameters.
Precedent analysisParameter extractionArchitectural data
04
Test
Generate and test design alternatives
Translate the extracted parameters into controllable variables inside a parametric model and test alternatives against multiple performance criteria.
GrasshopperParametric DesignWallaceiMulti-objective Optimisation
05
Design
Select a climate-responsive typology
Compare the strongest alternatives and select a configuration that balances the different performance objectives.
DaylightVentilationThermal performanceDesign selection
Climate Engine Walkthrough
A walkthrough of the climate engine used to find comparable climates.
Climate Clustering Walkthrough
A walkthrough of the PCA and K-Means clustering process.
Design logic
Goal
Generate climate-responsive design guidance where useful local precedent is absent.
Decision
Which global climates and vernacular strategies are most relevant to the target site?
Output
Architectural parameters and an optimised building typology.
Model
Climate filtering -> PCA + K-Means -> precedent and parameter extraction -> parametric model -> multi-objective optimisation.
Data
Global city and climate data, climate variables, vernacular precedents, architectural parameters, and performance criteria.
Links