Architectural Environmental Research · Taiwan

Better decisions
begin with evidence.

Residential Site Analyzer (RSA) explores how environmental simulation can become more explainable, traceable and useful during early-stage residential design.

01 / Why RSA

Simulation is powerful.
Understanding it is harder.

RSA investigates a critical gap between specialist environmental computation and design decisions: how can evidence be communicated clearly while preserving methods, assumptions and uncertainty?

01 / SOLAR

Solar & shading

Time-based direct solar and shading patterns, with relationships to neighboring massing and building levels.

02 / WIND

Wind conditions

Exploratory visualization of prevailing wind and airflow relationships; not a validated CFD substitute.

03 / THERMAL

UTCI comfort

Outdoor thermal-comfort interpretation, including sun-exposed versus shaded conditions and documented inputs.

03 / Research workflow

From context
to evidence-led decisions.

RSA organizes environmental information into a traceable workflow, while separating quantitative computation from explanatory interpretation.

01Site and massing

Set up site geometry, adjacent buildings and design assumptions with source information.

02Environmental analysis

Investigate solar/shading, wind-related context and thermal comfort at selected spatial and temporal scales.

03Evidence translation

Present readable spatial views and data summaries alongside uncertainty and validation status.

04Decision research

Study whether clearer environmental information improves user interpretation and early-stage design choices.

04 / Technical research & validation

Evidence, with a clear
validation boundary.

Historical RSA development records contain measured milestones. Figures below are associated with earlier versions and are not represented as current-release certification or complete-system validation.

r42 · historical shadow comparison
0.9795

Recorded direct-shadow intersection-over-union against a Radiance reference; 3/3 recorded cases marked PASS.

r42 · historical shadow comparison
0.25 m

Recorded shadow-boundary Hausdorff distance, with an approximate area bias of −2.05% under that comparison setup.

r44 · historical regression record
321

Tests marked PASS in the r44 historical record; newer versions require fresh regression testing.

Technology

Python and Streamlit underpin the prototype’s integrated research interface. Development includes map-based site setup, building massing, environmental visualization, time controls and reporting-oriented evidence.

PythonStreamlit2D / 3DSolar geometryRadiance reference comparisonWind previewUTCI

Data & methods

Project documentation references OpenStreetMap, Taiwan NLSC / official 3D sources where applicable, climate references including CWA and ERA5, and manual audit processes. The demonstration UTCI view labels Open-Meteo historical archive data.

Data source presence in research documentation does not imply every API connection is live, authoritative or already fully validated.

Sources: internal RSA development handoff dated 2026-10-05, with historical r42/r44 milestones. Scientific reproducibility requires source data, test scripts, environmental assumptions, benchmark configurations and re-execution on the current codebase.
05 / Future integration

Claude as an explanation layer.
Not a simulation engine.

A planned Claude API integration would help people read environmental evidence without replacing the underlying numerical analysis, source records or professional judgment.

Planned capability / 01

Evidence-grounded interpretation

Convert structured model results, configuration settings and limitations into plain-language explanations. Distinguish computed values from AI-generated interpretation.

Planned capability / 02

Comparison & traceable reports

Summarize floor-level and scenario differences, flag data gaps, draft reviewable environmental reports and preserve references back to inputs and outputs.

Status: proposed, not yet deployed. Claude will not calculate validated solar, wind or UTCI results in place of specialist engines. Human review and explicit uncertainty reporting remain essential.
06 / Developer & academic context

Independent research.
Practical application.

Residential Site Analyzer is led and developed by Yu-Jyun Chiu (CHIU YU JYUN) in connection with graduate architectural studies at the National University of Kaohsiung, Taiwan.

RSA is currently an independently developed research prototype and is not an incorporated company. Academic affiliation does not imply formal university sponsorship, endorsement, or authorization to represent the institution.