Enhancing Data-Driven Urban Building Energy Model with Multi-Dimensional Data Resources:Microclimate and Demographic Perspectives

Introduction

The traditional Urban Building Energy Modeling (UBEM) approach predominantly relies on archetype-based simulations, which generalize building characteristics and operational patterns to estimate energy demand across urban areas. While effective for large-scale assessments, these methods often overlook the influence of microclimatic conditions—such as localized air temperature, humidity, solar radiation, and wind flow—which can vary significantly across urban environments due to morphological heterogeneity. This simplification limits the accuracy of energy predictions, particularly in dense and diverse urban contexts where microclimate dynamics play a critical role in shaping building energy performance. Recent studies have underscored the importance of integrating microclimate factors into UBEM, noting that urban heat island intensity, surface radiation balance, and convective conditions can substantially alter thermal loads at the building level. However, the lack of standardized methods for extracting and quantifying microclimatic variables at scale has constrained their application in mainstream UBEM workflows. To address this gap, this research develops a standardized data pipeline to systematically construct urban-scale datasets that incorporate both building attributes and localized microclimate indicators. Leveraging this dataset, statistical modeling techniques are employed to evaluate the relative influence of various microclimatic factors on building energy consumption. This work not only enhances our understanding of the interplay between urban form, climate, and energy use, but also offers a methodological foundation for future data-driven UBEM studies seeking to improve predictive accuracy and inform resilient urban energy planning.

Satellite data collection

• Method

The satellite data acquisition pipeline is adapted and enhanced from the methodological framework proposed in prior New York City studies (Dougherty & Jain, 2023). For each building, local environmental conditions are characterized within a 100 m radius to capture neighborhood-scale microclimatic influences. Specifically, the original building footprint is first simplified using a convex hull algorithm to reduce geometric complexity while preserving its spatial extent. A 100 m radial buffer is then generated around the convex hull to define the final sampling domain for microclimate feature extraction. Satellite-derived environmental variables with native spatial resolutions ranging from 10 m to 2.5 km are processed and resampled within Google Earth Engine to a unified 100 m spatial scale. This resampling strategy ensures spatial consistency across heterogeneous data sources and guarantees that each building-level buffer reliably captures representative pixel-level microclimate information.

• Microclimate information

We place particular emphasis on vegetation coverage in the surroundings of each building. In a previous study, we investigated the relationship between the urban tree canopy ratio and building energy consumption (Xu et al. 2025). In the present work, we adopt the Normalized Difference Vegetation Index (NDVI) as the primary indicator to characterize the greenness of the local built environment. Compared with canopy ratio metrics derived from land-cover classifications, NDVI offers a key advantage in its ability to explicitly capture seasonal variations in vegetation phenology, which is critical for subsequent analyses under representative summer and winter conditions. NDVI data are obtained from the ECOSTRESS Tiled Ancillary NDVI and Albedo L2 Global 70 m V002 product.

For integrated microclimatic conditions surrounding each building, including air temperature, humidity, and wind speed, we utilize data from RTMA (Real-Time Mesoscale Analysis), a high-resolution near-surface atmospheric analysis system provided by NOAA and generated through real-time data assimilation of surface observations and numerical weather prediction outputs. In addition, to characterize nocturnal urban activity and radiative intensity, which serve as indirect proxies for urban vitality and anthropogenic carbon emissions, we employ the VIIRS Stray Light Corrected Nighttime Day/Night Band Composites Version 1 product to extract nighttime radiance information.

Our completed dataset can be found here, and an introduction to each variable can be found here.

• Preliminary UTCI Observations

We conducted an initial exploration of the Universal Thermal Climate Index (UTCI) to combine air temperature, humidity, wind speed, and radiation into a single thermal stress metric. UTCI was calculated at the building-neighborhood scale using the same buffered satellite and reanalysis data. Early results show clear spatial variation in UTCI across nearby buildings, suggesting that localized microclimate effects may influence building energy use. This analysis is ongoing and motivates the inclusion of UTCI as an additional microclimate feature in later modeling.

• Sky view factor

Building facade data collection

Reference

Dougherty, T. R., & Jain, R. K. (2023). Invisible walls: Exploration of microclimate effects on building energy consumption in New York City. Sustainable Cities and Society, 90, 104364.

Xu, H., Li, C., Kastner, P., & Dogan, T. (2024) Understand Urban Building Energy Consumption with Explainable Machine Learning Approaches.

Presentation

Team