Research

Mapping how efficiently cities use energy, from space

Buildings account for a large share of energy use and emissions, yet the evidence needed to plan retrofit is patchy, expensive to collect and quickly out of date. My work tests how far open satellite and geospatial data can fill that gap.

Doctoral project

Assessing energy efficiency of buildings: the role of remote sensing data and geospatial methods. PhD, School of Geography and Environmental Sciences, Ulster University (started September 2024), supervised by Dr Saad Bhatti, Dr Paul McKenzie and Dr Muhammad Irfan. Funded by a Vice-Chancellor’s Research Scholarship (VCRS), Ulster University.

Energy Performance Certificates exist for only some homes and date quickly. I am testing whether open satellite, mapping and climate data, available for every building, can fill those gaps well enough to support retrofit planning across whole cities. Greater London and Belfast are the two study areas.

Questions and objectives

  1. Review what geospatial and remote-sensing methods already achieve for building energy efficiency.
  2. Build reproducible building-level datasets for London and Belfast.
  3. Develop and test models that carry over to new areas, cities and years.
  4. Turn the estimates into maps that are useful for policy.

Data and methods

SourceWhat it describes
Domestic EPC register (England and Wales; Northern Ireland)Labels: the A–G band and assessment attributes
Sentinel-1 SARStructure, roughness and density of the built form
Sentinel-2 and Landsat optical imageryRoof and surface materials, vegetation, spectral built-up indices
Landsat and MODIS land surface temperatureThermal behaviour and surface urban heat
VIIRS night-time lightsActivity and occupancy
ERA5 reanalysisLocal climate and heating and cooling demand
OpenStreetMap and building footprints (Overture Maps)Building geometry, shared walls, density, streets and neighbourhood form
Airborne LiDAR (Environment Agency; OpenDataNI)Building height, storeys and roof form
Satellite embeddings (Google, annual)Learned summaries of the surroundings, compared with engineered features

Results will be added here as papers are published.

Atlas of results

The GBEM Atlas maps the energy performance of the housing stock in Greater London and Greater Belfast by 250 m grid cell, neighbourhood and street: certified ratings and related indicators. It shows aggregates of at least ten certificates per unit and no data on individual homes.

Open the atlas

Systematic review (in preparation)

A systematic and bibliometric review of remote sensing and geospatial methods for building energy efficiency. Details will follow on publication.

Open code and tools

GBEM: Geospatial Building Energy Mapping holds the project’s code (Apache 2.0, copyright Ulster University). It contains no building-level data.

AhmadAmmar/GBEM

General tools from the project: research-helpers (Python utilities for research workflows) and an unofficial Ulster University thesis template.

Talks and reports

  • Geospatial urban energy efficiency assessment. PhD confirmation report and presentation, Ulster University, October–November 2025.
  • GeoML-driven urban energy efficiency assessment. Research presentation (first-year PhD), June 2025.

Earlier research

GEDI LiDAR for mangrove canopy height and biomass, Indus Delta (MS, 2023)

MS dissertation at the Geospatial Research and Education Lab, Institute of Space Technology, Islamabad, supervised by Dr Hammad Gilani, with Dr Adeel Ahmad (University of the Punjab) as external examiner. I developed a wall-to-wall approach for canopy height and above-ground biomass density in the Indus Delta mangroves. Random forest models in Google Earth Engine predicted GEDI L2A RH98 heights and L4B biomass from features of ALOS PALSAR-2, Sentinel-1 SAR and Sentinel-2 optical imagery, and the continuous maps were validated against field plots. Abstract.

Deep-learning extraction of urban features from fused imagery (BS, 2019)

Final-year project at the Institute of Geo-Information and Earth Observation, PMAS Arid Agriculture University Rawalpindi, supervised by Dr Hasan Ali Baig, carried out with Muhammad Bilal. An end-to-end pipeline in Keras and fastai compared U-Net, SSD and Mask R-CNN for building extraction from fused multisource satellite and aerial data; U-Net performed best, with an F1 score of 0.89. The code and a ready-to-use dataset are on GitHub. Project page.

Shorter projects on air quality, urban growth, floods, snow, water quality and land cover are listed on the projects page, including snow cover and monsoon and Third Pole climate, and in the CV.