This deliverable outlines the initial development, testing, and implementation of digital shadow models designed to predict and simulate the thermal and psychological impacts of Nature-based Solutions (NbS). Standard urban planning struggles to quantify how introducing trees, green walls, or de-sealing concrete will change the microclimate.

D4.3 resolves this by developing a high-resolution 3D environmental simulation workflow centered around the Universal Thermal Climate Index (UTCI)—the international standard for assessing human outdoor thermal comfort.

The report documents the initial baseline modeling of three diverse pilot sites: Barcelona, Prato, and Helsinki. Additionally, it introduces the integration of unsupervised artificial intelligence and machine learning (AI/ML) analytics to process on-site sensor data and evaluate how these newly co-designed green spaces actively alleviate human stress.

Primary Audience:

• Urban Planners, Architects, and Landscape Designers

• GreenInCities Consortium Technical Partners

Secondary Audience:

• The 112 Climate-Neutral and Smart Cities network

• Environmental Scientists and Academic Researchers

The physics-based simulations and machine learning models were developed through a structured six-step workflow

• 3D geometry and asset acquisition: Constructing detailed 3D base geometry models of the pilot neighborhoods. In the case of missing municipal files or data gaps, remote-sensing techniques (including satellite imagery and drone photogrammetry) were deployed to capture physical geometry, building heights, and surface materials.

• Environmental parameter definition: Preparing essential environmental inputs, including surface albedo, sky view factors (SVF), and Typical Meteorological Year (TMY) weather files.

• Ray-tracing and shading simulation: Executing high-resolution solar radiation and shading simulations using IES's physics-based simulation engine (Apache) to map shadow depths across the site over seasonal cycles.

• UTCI equation solving: Ingesting wind vectors, radiant temperatures, humidity, and solar exposure data to calculate and compile localized UTCI metrics.

• Neurological AI/ML training: Running parallel research to calibrate machine-learning algorithms capable of linking spatial landscape metrics to physiological human stress responses (mEEG and heart-rate variability).

• Simulation validation: Testing initial baseline scenarios against historical city records to verify accuracy.

The physics-based simulation framework and machine learning analytics deliver highly localized, predictive climate data through several core technical components:

• An operational digital shadow model represents a high-resolution 3D base mesh enriched with one-way physics-based simulations to forecast microclimatic impacts before any physical work starts.

• The system calculates the Universal Thermal Climate Index (UTCI) across multiple seasonal profiles on any spatial plane, simplifying complex environmental physics to model outdoor thermal comfort.

• Three active, localized demonstration workflows have been deployed to simulate riverbank renaturalization in Barcelona, schoolyard micro-forest shading in Prato, and water-cooling canopy corridors in Helsinki.

• A strategic development roadmap outlines the integration of bidirectional thermal exchange between indoor building envelopes and outdoor nature-based solutions.

• An innovative machine learning module developed with NeuroLandscape processes physiological data from local participants to mathematically link spatial metrics with human stress reduction.

The practical deployment of these environmental digital shadow models is expected to advance the technical capabilities of climate-adaptive city planning through several key strategic outcomes:

• Cities can transition to evidence-based urban design by testing the precise cooling performance of nature-based solutions virtually, replacing qualitative estimates with empirical data.

• The state of the art in smart-city planning is advanced by creating a clear technical bridge between indoor energy simulations and outdoor microclimate models.

• Standardized open geospatial outputs are generated to allow these predictive models to easily plug into municipal 3D digital twins and web-GIS platforms.

The Figures include UTCI results per pilots and AI-rendered Indoor-Outdoor Bidirectional Modellings