Synthetic Smart City IoT Dataset — High-Pollution Day, 672 Rows

Overview

Synthetic urban public-space IoT time series generated by the IoTSyn v4.4.0 reference engine (Smart City & Public Spaces Edition). A coupled physical–stochastic model produces meteorology (solar geometry, cloud/precipitation Markov chains, AR(1) air temperature with a nocturnal urban-heat-island term, and a Weibull wind field induced through a Gaussian copula), human presence (a state-modulated non-homogeneous Poisson arrival process with LogNormal dwell times and infinite-server occupancy), road traffic, species-resolved air quality from a well-mixed mass-balance box model (PM2.5 and NO2, with separate physical and sensor columns), environmental noise composed in the energy domain (LAeq), and UTCI thermal comfort. A three-way separation between physical truth, sensor observation, and decision output is preserved throughout, and a multi-hazard decision layer emits an ordinal urban risk vector (thermal, air quality, noise, crowding), operational action levels, risk episodes, and provenance-tagged behaviour-change strategy objects. All randomness derives from named substreams of a portable MT19937 generator; identical configuration and seed reproduce byte-identical output. Each dataset ships with machine-readable sidecars: normalized scenario, domain validation report, risk episodes, and the strategy file.

672 data points, 64 columns. Config: IoTSyn v4.4.0 engine; master_seed: 88; Capacity: 500; Controls: preset=high_pollution; EXTENDED (non-normative).

Column Schema

Mathematical Models

• Meteorology: AR(1) air temperature + sinusoidal diurnal cycle + nocturnal urban-heat-island (Oke); Magnus–WMO saturation vapour pressure; Weibull 10 m wind via an AR(1) Gaussian copula (Sklar), pedestrian wind through a log profile
• Thermal comfort: UTCI operational polynomial (Bröde et al. 2012) from air temperature, mean radiant temperature, 10 m wind and vapour pressure; clamp-and-flag domain policy; 10 scientific categories plus a coarser 4-level operational band
• Air quality: Single well-mixed box mass balance with ventilation exchange, deposition and chemical first-order loss and a traffic-driven source; exact exponential one-step update; mean-preserving log-normal sensor observation model; 24 h rolling means
• Presence: Discrete-time Markov-modulated NHPP arrivals with thermal/rain/wind deterrence; exact per-interval Poisson counts with sorted uniform offsets; LogNormal dwell; infinite-server occupancy reconstructed from events; largest-remainder audience allocation
• Noise: A-weighted equivalent continuous level (LAeq) over a declared window, summed in the energy (intensity) domain per ISO 1996
• Risk & strategy: Conservative-max hazard vector with compound-risk indicators; compound priority score H·log10(1 + person-minutes / E_ref); ex-post, provenance-tagged behaviour-change strategies

Use Cases

• Thermal-comfort (UTCI) and heat-risk analytics
• Air-quality (PM2.5 / NO2) exposure and sensor calibration / drift research
• Occupancy and arrivals (NHPP) estimation benchmarks with exact ground truth
• Environmental-noise (LAeq) exposure analysis
• Multi-hazard urban risk aggregation and decision-support research

Reproducibility

Seed: 88. Same seed + parameters = identical output. IoTSyn v3.1.