Key points are not available for this paper at this time.
Internet of Things (IoT)-enabled smart waste collection is typically evaluated based on operational savings, while the electricity use and emissions of the enabling digital infrastructure are rarely quantified. This article presents a unified, resource-aware evaluation framework that computes a net environmental balance by jointly accounting for operational impacts and end-to-end information and communication technology (ICT) energy use and C O 2 emissions across the device, network, edge, and cloud layers. To ground the assessment in realistic urban conditions, the daily atmospheric context was constructed from Sentinel-5P Level-3 time series over New Delhi from 2020 to 2025 using the UV Aerosol Index (UVAI), CO, and NO 2 . After QA-masked compositing and preprocessing, pollution regimes were identified via standardized multivariate clustering and evolved using a Markov scenario generator. Multi-horizon forecasting (7-day and 30-day horizons) was then compared across ARIMA, ridge regression, random forest, gradient boosting, and long short-term memory (LSTM) to estimate risk and drive a forecast-adaptive ICT policy that tunes sensing and communication rates. The results showed that ensemble learners provided the highest forecasting accuracy across pollutants, while the adaptive ICT design reduced average ICT emissions, showing measurable, regime-aware energy and CO 2 savings without sacrificing predictive capability for city-scale evaluation.
Bedi et al. (Wed,) studied this question.