CHETNA-Residential Sector: High-Resolution GHG Emissions Analysis for Indian Cities using Building-Level Climate-Energy Modelling

India's rapid economic growth and unprecedented urbanization have led to a significant rise in energy consumption and greenhouse gas (GHG) emissions from residential buildings. Currently, the residential sector accounts for approximately 20% of India's total GHG emissions, with a few major cities contributing up to 42.8% of total urban emissions. Projections indicate an 8-fold increase in energy demand by 2050, driven by rising urban expansion, household incomes, and greater ownership of energy-intensive appliances, such as air conditioners and refrigerators. With India’s urban population expected to reach 50.3% by 2050 and over 70% of the country’s building stock yet to be constructed, cumulative emissions from the buildings sector between 2020 and 2070 could surpass 90 gigatonnes of CO₂e, exceeding the nation’s allocated carbon budget.

This study aims to create high-resolution CO₂e emissions datasets for 100 Indian cities to address the lack of reliable data necessary for effective urban GHG mitigation planning. We employed a semi-supervised learning approach to classify building types and estimate heights using an XGBoost model trained on Microsoft Building Footprint data, satellite imagery, OpenStreetMap features, and other urban datasets. These outputs were then integrated into a building-climate-energy model, which combines household survey data, climate variables, and derived building features to estimate residential energy consumption. The household survey provides detailed insights into appliance usage, energy consumption patterns, and variations across income classes.

The building characteristic prediction model achieved good performance, with an average F1-macro score exceeding 0.8 for type and height predictions on the testing set. Similarly, the energy prediction model demonstrated robust accuracy, with an R² > 0.6 on the testing set. Using explainable machine learning techniques, such as SHAP, we identified air humidity and income class as the most critical factors influencing residential energy consumption, highlighting the interconnected roles of climate and socioeconomic conditions in shaping residential energy demand. Finally, gridded emission map time series were developed for each city using the city population, building characteristics, and a simplified energy model that incorporates climate data and regional income classifications.

This work is part of the CHETNA project (City-wise High-resolution carbon Emissions Tracking and Nationwide Analysis), which leverages artificial intelligence and advanced datasets to provide high-resolution, near real-time CO₂e and air pollutant emissions data for over 100 Indian cities. By integrating building-climate-energy modelling, this study delivers spatially and temporally granular emissions datasets for the residential sector. These results empower the CHETNA project to support localized mitigation strategies, promote energy-efficient building practices, and inform sustainable urban planning tailored to India’s unique urban landscape.