The sea breeze that reaches this coast every afternoon is the largest free thermal asset any Indian metro owns. MADFLUX proposes to map it, unblock it, and steer it — using a validated digital twin of all 426 km² under Greater Chennai Corporation, ward by ward.
Between 1991 and 2016 the built-up share of Chennai rose from 1.46% to 22%, and mean land surface temperature rose from 33.3 °C to 38.9 °C across the same period. Heat is trapped between buildings, so the core now runs hotter than the periphery — the inverse of the pattern a coastal city should show.
The usual response is to plant trees and hope. That is not an engineering programme. A tree planted in a stagnation pocket cools a few square metres; the same tree planted on a breeze-alignment corridor cools a neighbourhood downwind of it. The difference is not horticulture. It is fluid dynamics, and it is computable.
MADFLUX has already computed it — for one campus, with CFD, live sensors, and a built result on the ground. This document sets out how the same method scales to a city of seven million.
Schematic east–west section through Greater Chennai. Not to scale. Flow behaviour illustrated from campus-scale CFD and published SUHI literature; city-scale flow field to be solved during Phase I.
Land-use based SUHI assessment of Chennai has identified 82 hotspots against just 18 cool spots — and residential land use has fallen furthest on the urban thermal index between 2006 and 2022.
Mean land surface temperature climbed from 33.3 °C to 38.9 °C as the urban share of land went from 1.46% to 22%. Chennai is projected to sprawl to roughly 2,376 km².
The Chennai Climate Action Plan sets a carbon-neutral roadmap across six priority areas, prepared with C40 Cities. Urban cooling is named but not yet instrumented. This programme supplies the instrument.
Phase 0 ran the full loop end to end on a live school campus in Chennai: GIS survey, a Virtual Physical Thermal Environment Replica, steady and transient CFD, physical build, then sensor validation against the model. Every step below is on the ground today.
| Stage | Window | Output |
|---|---|---|
| Baseline survey | 11–27 Sep | GIS map, manual traverse, procurement manifest |
| Static digital twin | 28 Sep–5 Oct | VPTER geometry, baseline CFD + thermal run |
| Weather station | 6–8 Oct | Multi-sensor mast, calibrated, MQTT/LoRa uplink |
| Green wind corridors | 9–15 Oct | Grassland + planted corridor, foggers, flow diversion |
| Hydroponic façades + EvapFog | 16–20 Oct | Vertical NFT modules, terracotta evaporative tiling |
| Community engagement | 10–20 Oct | Open blog, bug bounty on the twin, Green Data Day |
| Validation | 21–30 Oct | Measured vs modelled, RMSE, impact dossier |
Drag to compare. Left is the baseline CFD-derived surface field over the campus courtyard; right is the same field after corridor planting, EvapFog tiling and façade hydroponics. The hot spine down the centre of the courtyard is a stagnation pocket — it is removed not by shading it, but by reopening the flow path that feeds it.
Campus figures from the MADFLUX Phase 0 dossier. Heat surfaces below are renderings of the modelled field, not photographs.
The campus proved a closed loop: survey, twin, simulate, build, measure, correct. Scaling it is not a research problem — it is a logistics problem. The twin is the thing that makes the logistics tractable, because it lets the Corporation test an intervention before spending on it.
Critically, the loop does not end at build. Every sensor feeds the twin back. A model that disagrees with its ward by more than 1.0 °C RMSE is not permitted to advise on that ward's next intervention until it is recalibrated.
A ward twin must agree with its sensors within 1.0 °C before it can authorise spending in that ward.
LoRaWAN uplink to a GCC-hosted MQTT broker; no dependence on a vendor cloud.
Raw feeds, twin geometry and simulation inputs published under an open licence. A public bug bounty on the model, as run on campus.
One trained maintenance technician per ward, recruited locally, supported by a school-based student corps.
Chennai is not one thermal problem. A port ward, a 1930s legacy core, a planned grid suburb and a glass IT corridor fail in different ways and must be fixed in different ways. Colour the map by heat, canopy, breeze role, priority or investment.
Zone boundaries are schematic representations of the 15 Greater Chennai Corporation zones, drawn for navigation rather than survey. Surface temperature, canopy and risk values are modelled indicative estimates derived from published Landsat-based SUHI studies of Chennai and from the campus-scale model; they are stated here to show the method and must be replaced with ground-truthed values during the Phase I survey before any works order is raised.
| Zone | Typology | Breeze role | Area km² | Pop. | LST °C | Canopy % | Risk | Tier | Capex ₹Cr | Modelled ΔT |
|---|
Six of these are already standing on the pilot campus. The rest are city-scale extensions of the same physics. Each entry states its mechanism, its unit cost, where it works, where it does not, and who maintains it.
Unit rates are planning-stage estimates at 2026 prices, inclusive of installation and first-year maintenance, to be firmed against GCC schedule of rates at detailed design. Cooling figures are modelled at the stated scale of application.
This is a transparent linear approximation of the full CFD model, intended for budget conversations rather than for design. Every coefficient it uses is printed below the output, so any reviewer can check the arithmetic.
Derived from the campus CFD run, calibrated against published urban-cooling literature for tropical coastal cities. Zone-specific multipliers apply for breeze role and baseline canopy. This linear surrogate is not a substitute for the full solve; it is a budgeting aid.
The programme is deliberately structured so that it can be stopped after any phase with the previous phase's value already banked. Nothing depends on the whole thing being funded at once.
Spread over six years, the full programme costs roughly ₹300 per Chennai resident per year — about the price of two bus fares. Select a segment to see what it buys.
No single source carries this. The financing plan deliberately blends existing sanctioned envelopes with new instruments, so that the programme does not create a new recurring demand on the GCC general fund.
Every claim in this proposal is written as a falsifiable target with a named data source and an independent verifier. Ward scorecards publish monthly whether the target was met or missed.
| Indicator | Baseline | Phase III target | Source | Verified by | Cadence |
|---|---|---|---|---|---|
| Mean daytime land surface temperature | 38.9 °C | −1.9 °C | Landsat 8/9 TIRS + ward sensor mesh | Anna University / IIT Madras | Seasonal |
| Peak surface temperature at treated sites | up to 52 °C | −8 °C | Handheld IR traverse + fixed nodes | GCC zone engineer | Monthly |
| Universal Thermal Climate Index, 14:00–16:00 | Strong heat stress | Moderate heat stress | Globe temp + RH + wind nodes | TN Dept. of Environment & Climate Change | Daily |
| Tree canopy cover, GCC area | 8.4% | 16.0% | Sentinel-2 NDVI + ground census | TN Forest Dept. | Annual |
| Heat-related emergency presentations | To be set | −25% | GCC urban health centres, Apr–Jul | Directorate of Public Health | Weekly, summer |
| Evening peak electricity demand, treated wards | To be set | −4% | TANGEDCO feeder telemetry | TANGEDCO | Monthly |
| Stormwater ponding duration, treated catchments | To be set | −30% | Depth loggers + citizen reports | GCC Storm Water Drain dept. | Per event |
| NOx at photocatalytic junctions | To be set | −12% | Paired reference / treated monitors | TNPCB | Continuous |
| Twin agreement with sensors (RMSE) | 0.74 °C (campus) | < 1.0 °C all wards | Model vs. measured, held-out nodes | Independent academic panel | Monthly |
| Recycled water share of cooling demand | — | 100% | CMWSSB tertiary treatment meters | CMWSSB | Monthly |
Raw sensor feeds, twin geometry, simulation inputs and every works order published on a public portal under an open licence, as the campus pilot did.
Anyone who demonstrates a material error in the twin is credited and paid. This ran on campus and found real problems. It is cheaper than being wrong at city scale.
If a zone misses its Phase-gate temperature target by more than 40%, that zone's next tranche is withheld pending an independent review. Written into the programme, not left to discretion.
This programme does not ask for a new parastatal. It asks for a small technical cell inside the Greater Chennai Corporation that owns the twin, and for existing departments to keep doing what they already do — to a specification the twin supplies.
| Risk | Severity | Mitigation built into the design |
|---|---|---|
| Water scarcity. Evaporative cooling in a city that ran dry in 2019 is politically and practically indefensible if it touches drinking water. | Critical | At peak-season full operation the programme draws about 5.2 MLD, roughly 5.8% of existing tertiary-treated reverse-osmosis capacity at Koyambedu and Kodungaiyur, and far less outside summer. Potable draw is prohibited by design; assets are plumbed only to recycled mains and shut down automatically on supply failure. |
| Cyclone and monsoon damage to sensors, façade modules and foggers. | High | All exterior assets rated to IP66 and 180 km/h; façade modules designed for seasonal strike-down; sensor masts on GCC streetlight poles with breakaway mounts. Northeast monsoon months are maintenance windows, not build windows. |
| Maintenance collapse after handover — the standard failure mode of Indian urban greening. | High | Maintenance is funded for seven years inside the capital line, not left to a future revenue budget. One paid ward technician, plus a student corps, plus a public monthly scorecard that names which ward's assets are dead. |
| Model overconfidence. A twin that is trusted more than it deserves misdirects real money. | High | Hard 1.0 °C RMSE gate; held-out validation nodes; independent academic verification; public bug bounty; every simulation input published. |
| Land availability for corridors and microforests in the dense core. | Medium | Tier-1 core zones are deliberately assigned vertical and surface measures — cool roofs, façades, EvapFog tiling — which need no land. Corridor work is concentrated where road widths already allow it. |
| Electoral cycles interrupting a six-year programme. | Medium | Phased so each tranche delivers standalone value; ward committees, not the state, hold the scorecard; the specification change to the schedule of rates survives any administration. |
| Vandalism and theft of sensor and irrigation hardware. | Moderate | Low-value commodity hardware, no resale market; nodes mounted above reach; local employment and student ownership are the real deterrent, as the campus pilot showed. |
| Baseline data quality. Published LST and canopy figures vary between studies. | Moderate | Phase I exists precisely to replace every modelled figure in this document with a ground-truthed one before any works order is raised. No number here is treated as final. |
We are asking for Phase I: three wards, one city twin, eighteen months, and the permission to put sensors on streetlight poles. Everything after that is contingent on Phase I meeting published targets.
Covers the base city twin, three reference ward builds, the sensor mesh, and eighteen months of the twin cell. Drawable against the existing AMRUT 2.0 and Challenge Fund envelopes without new appropriation.
One dense legacy core ward, one planned residential ward, one coastal or IT-corridor ward — so that the three dominant Chennai typologies are each proved before scale-up. We propose wards in Thiru Vi Ka Nagar, Anna Nagar and Perungudi.
Access to GCC building footprints, road and drain asset registers, ward GIS layers and available LiDAR, plus CMDA land-use data. Without these the twin has to be rebuilt from satellite imagery at four times the cost and half the accuracy.
A standing permission for the mesh to use streetlight poles, bus shelters and school compound walls, with a standard mounting detail agreed with the Electrical wing. This is the single biggest schedule risk in Phase I.
Commit to considering a breeze-corridor overlay in the next Master Plan revision and cool-surface line items in the next schedule of rates update. Costs nothing now; determines whether the next twenty years of construction helps or hurts.
A formal pathway for school and college students to hold ward-level data and maintenance roles, with credit recognised by the Directorate of School Education. This programme was designed by students; it should be inherited by them.
Phase I publishes measured temperature change in three wards, or it fails publicly and the programme stops.
1.4% of the full programme. The cost of finding out whether the method scales.
Everyone inside the 426 km² Corporation boundary, with the wider metropolitan area following in Phase IV.