System and Method for Black Ice Risk Mapping Using Tire Pressure Monitoring System Temperature and Pressure Telemetry from Connected Vehicles

Abstract

Black ice is a thin, transparent layer of ice on a road surface. It is nearly invisible, it forms most often at night and at dawn, and the Federal Highway Administration attributes roughly 1,300 deaths a year to it in the United States. It forms only when the pavement surface drops to or below freezing while moisture is present, and it frequently forms while the air a few feet above the road is still above freezing, which is why drivers, forecasts, and air temperature sensors all miss it. Fixed road weather stations measure pavement temperature at a few thousand points across the country. The rest of the road network, millions of miles of it, is unmeasured.

This disclosure turns the tire pressure monitoring system already fitted to every new car into a moving road surface thermometer. Direct TPMS sensors sit inside each wheel and measure the pressure and the temperature of the air in the tire cavity. That cavity temperature is coupled to the road surface through the contact patch, and it is confounded by heat generated inside the rolling tire. A thermal correction model estimates the rolling heat from speed, load, and inflation pressure and subtracts it, leaving an estimate of the road surface temperature at the point where the tire touches the road. Fused with dew point and other moisture evidence, map-matched to road segments, and aggregated across a fleet of vehicles, these estimates produce a live, segment-level black ice risk map covering the roads cars actually drive, with no new hardware in the vehicle and no new sensors in the pavement.

Field of the Invention

This invention relates to automotive safety and road weather sensing: estimating road surface temperature and black ice formation risk from tire pressure monitoring system telemetry already present in connected vehicles, and aggregating those estimates across a fleet into a live road segment risk map.

Background

Road ice formation is governed by two conditions, and both must hold. First, the pavement surface temperature must be at or below the freezing point, 0 °C. Second, moisture must be present on or condensing onto the surface. A three-year study of Montana highway Road Weather Information System (RWIS) measurements found that ice formed only when the pavement skin was at or below freezing, while the air temperature two meters above the road could still be above freezing, and that pavement below freezing frequently did not produce ice, because the moisture condition was missing. Three mechanisms supply the moisture without snowfall: rain that pools during the day and freezes overnight, water vapor condensing directly onto a sub-freezing surface as hoarfrost, and snow that melts and refreezes. Where the surface temperature falls below the dew point, moisture from the air condenses onto the pavement, as liquid dew above freezing or as frost and black ice below it.

Existing defenses each cover only a fragment of the network:

• Fixed road weather stations (RWIS): Embedded pavement sensors and atmospheric stations measure pavement temperature, freezing point, and dew point at fixed locations. Iowa's Department of Transportation began with 22 RWIS sites and planned 47; the District of Columbia installed its first RWIS units on bridges, because bridges freeze before surface streets. RWIS gives precise data at a point and nothing between points. A state with 47 stations has 47 points on thousands of miles of road.
• Winter maintenance patrols: Some agencies mount infrared pavement thermometers on plows and patrol trucks. Coverage is limited to the patrol route and the patrol schedule, which is exactly when conditions are already suspected, not the hours before.
• Forecast models: Numerical road weather models predict pavement temperature from air temperature, radiation, and road properties. They run on coarse grids, they depend on road surface properties that vary by segment and are rarely surveyed, and they cannot see the cold pocket on one shaded bridge deck.
• Camera and friction sensing: Vision systems and acoustic friction estimation detect ice or low friction after a vehicle is already on it. They measure the hazard at the moment of encounter, not ahead of it, and they do not produce a map for the next driver.
• TPMS as a safety system: Direct tire pressure monitoring systems have been mandatory on new US passenger vehicles since 2007 under FMVSS 138. Each sensor measures pressure and temperature inside the tire and transmits both to the vehicle. That temperature channel is used today only to compensate pressure readings and to warn of overheating. No system known to the inventors uses it to infer the temperature of the surface the tire is rolling on.

The gap in the art is a system that measures road surface temperature continuously, everywhere vehicles already drive, using hardware already installed in every new car, and turns those measurements into a risk map before the next driver reaches the ice. The novelty claimed here is the use of TPMS cavity temperature, corrected for rolling heat, as a road surface temperature probe, fused with dew point and moisture evidence into a per-segment black ice risk score, and aggregated across a fleet with a confirmation loop from traction events.

Detailed Description

1. System Architecture

Each participating vehicle carries four direct TPMS sensors, one per wheel, mounted at the valve stem or banded to the wheel barrel. Each sensor measures the pressure and temperature of the air inside the tire cavity and transmits the pair by radio frequency or Bluetooth Low Energy to the vehicle's receiver, typically every 60 to 300 seconds while driving and on significant change. A vehicle gateway, the telematics unit, an aftermarket OBD dongle, or a fleet telematics box, reads the TPMS pair together with vehicle speed, ambient air temperature from the vehicle's own sensor, relative humidity where the vehicle carries a humidity sensor, wiper and rain sensor state, and GNSS position and heading. The gateway transmits a compact telemetry record to a cloud aggregator, which map-matches each record to a road segment, applies the thermal correction and the ice formation test, aggregates records across vehicles into per-segment risk scores, and publishes the resulting map to three consumers: drivers approaching a flagged segment, navigation and routing services, and winter maintenance dispatch. All scoring and aggregation happen server-side in the preferred embodiment; a vehicle-local embodiment computes its own segment risk from its own telemetry plus a downloaded risk map and issues alerts without connectivity.

2. TPMS Temperature Acquisition

Direct TPMS sensors measure the actual air inside the tire. Commercial sensor datasheets report temperature measurement accuracy of about plus or minus 3 °C over an operating range of minus 40 °C to plus 125 °C, with pressure accuracy of about plus or minus 0.3 bar. The temperature sensor exists because pressure must be temperature-compensated: for a fixed tire volume, pressure follows absolute temperature under the ideal gas law, so a tire at 35 psi gauge loses about 1 psi for every 10 °F (5.6 °C) drop in cavity temperature. That same temperature channel is the road probe in this disclosure. Indirect TPMS, which infers pressure from wheel speed differences through the anti-lock braking system, carries no temperature sensor and is excluded.

Four sensors per vehicle give four semi-independent probes of the same road segment, one per wheel track. The two front tires and the two rear tires sample slightly different lateral positions on the lane, which matters because ice forms first in the wheel tracks where water pools and in the center of the lane where it does not. Disagreement among the four sensors is itself a signal: a single cold tire suggests a sensor fault or a shaded patch, while all four tires cooling together across a segment boundary suggests the road itself changed temperature.

3. Thermal Correction: Separating Road Temperature from Rolling Heat

The cavity air temperature is not the road temperature. It is a balance of heat flows, and the correction model accounts for each one. Heat enters the tire from two sources: hysteresis, the heat generated by repeated deformation of the viscoelastic rubber as the tire rolls, which rises with speed, load, and low inflation pressure; and friction between the tread and the road. Heat leaves the tire by convection from the sidewalls to the ambient air, by conduction through the rim, and by conduction through the contact patch to the road itself. Finite element studies of rolling tires put the tread-to-road heat transfer coefficient near 12,000 W/m²·°C, against about 16 W/m²·°C for the tread and sidewall to air, which means the contact patch is the dominant thermal coupling to the road even though it is a small fraction of the tire surface. The same studies show the tire reaches thermal steady state roughly 15 minutes after the start of driving, and that the maximum tire temperature rises with load, speed, and low inflation pressure.

The correction model estimates the hysteresis heat from the vehicle's own signals: speed from the wheel speed sensors, load from the vehicle's rated axle load adjusted by measured suspension deflection or a static estimate, and inflation pressure from the TPMS reading itself. A lookup table or a learned regression maps those three inputs to an expected steady-state temperature rise above ambient, calibrated per vehicle model and tire size from fleet data. The model subtracts that rise, and the ambient air temperature, from the measured cavity temperature. What remains is the road contribution: an estimate of the road surface temperature at the contact patch.

Two calibration mechanisms keep the model honest. First, cold-start equilibrium: when a vehicle has been parked for two hours or more, the tire, the road beneath it, and the air have equilibrated, so the cavity temperature at key-on equals the local road and air temperature within sensor accuracy. That key-on reading anchors a per-tire offset that absorbs sensor bias and slow drift. Second, segment-to-segment differences: when a vehicle crosses from a warm segment to a cold one, the cavity temperature changes in the direction of the new road temperature with a time constant of minutes. The change cancels the steady hysteresis offset, which is common to both segments, so the difference between segments is measured more accurately than the absolute temperature of either one. Bridges are the sharpest case: a bridge deck loses heat to the air from below as well as above, so it cools faster than the road on either side, and a vehicle crossing onto a bridge shows a fast cavity temperature drop that the segment-difference method catches.

4. The Black Ice Formation Test

A segment is scored for black ice risk only when both formation conditions are evaluated. The first condition is temperature: the corrected road surface temperature estimate must be at or below 0 °C, with a safety margin of 1 °C to account for sensor error, so the test threshold is 1 °C. The second condition is moisture, evaluated from the best available evidence in descending order of reliability: a dew point at or above the estimated surface temperature, which means water vapor condenses onto the pavement; a rain sensor or wiper active state, or precipitation in the last 6 hours from a weather service; or a snow event followed by a temperature rise above freezing and a fall back below it, the melt-refreeze pattern. Where no moisture evidence exists, the segment is scored as cold and dry, not as icy.

Risk is expressed as a score R from 0 to 1, computed as a weighted combination of three terms:

R = 0.45 × T + 0.35 × M + 0.20 × C

T is the temperature term, defined as the estimated surface temperature's distance below the 1 °C test threshold, scaled over 4 °C: T = clamp((1 − RST_est) / 4, 0, 1), so a surface at 1 °C scores 0 and a surface at minus 3 °C or colder scores 1. M is the moisture term: 1 where condensation or recent precipitation is established, 0.5 where precipitation occurred in the last 6 hours without condensation, 0 where the road is known dry. C is the context term: 1 for bridge decks and overpasses, 0.5 for shaded cuts and low-lying sections near water, 0 otherwise. R maps to three bands: Low below 0.25, Watch from 0.25 to below 0.5, and High at 0.5 and above. Confirmed is not a score band: a segment becomes Confirmed only when a traction event occurs on it, as described in section 6. The weights are provisional until measured, and the formula is stated so the table below can be checked by hand. R is rounded to two decimals in the table.

Case	Surface temp. est.	Dew point	Moisture evidence	R	Band
Warm night, no frost risk	4 °C	-2 °C	None	0.00	Low
Cold dry road, no moisture	-2 °C	-8 °C	None	0.34	Watch
Hoarfrost night, dew point above surface	-1 °C	0 °C	Condensation	0.58	High
Bridge deck, rain then freeze	-2 °C	-1 °C	Rain, bridge	0.89	High
Same bridge, traction control slip reported	-2 °C	-1 °C	Rain, bridge, slip	0.89, confirmed	Confirmed

The table makes the central point visible: temperature alone never reaches High. A road at minus 2 °C with dry air scores Watch, because the Montana data show that pavement below freezing frequently carries no ice at all. It is the combination of a sub-freezing surface and moisture that moves a segment to High, and only a traction event moves it to Confirmed.

5. Fleet Aggregation and Segment Mapping

One vehicle with a plus or minus 3 °C sensor cannot resolve a 1 °C threshold. A fleet can. Each telemetry record is map-matched to a road segment of about 100 meters, and the aggregator maintains a per-segment estimate that combines all recent records with weights for sensor accuracy, the number of contributing vehicles, the time since each pass, and agreement among the four tires of each vehicle. The standard error of the segment mean falls with the square root of the number of independent passes, so ten vehicles through a segment in an hour reduce a 3 °C per-sensor error to roughly a 1 °C segment error. Segments with fewer than three independent passes in the last hour are marked low-confidence and are not promoted above Watch on temperature evidence alone.

Estimates decay with time. A segment estimate loses half its weight every 30 minutes without a new pass, and it is dropped entirely after 3 hours, because road temperature follows the air and the sun and a morning estimate is worthless by afternoon. Bridge and overpass segments carry a persistent context flag from map data, so they are scored with the context term even when the temperature estimate is stale, and they are the first segments a maintenance dispatcher sees when a cold front arrives.

6. Ground-Truth Confirmation and Model Learning

The system closes its own loop with signals the vehicle already produces. An anti-lock braking intervention, a traction control slip event, or a stability control correction on a segment flagged High is treated as ground truth that ice was present, promotes the segment to Confirmed, and is fed back to recalibrate the thermal model for that vehicle model and tire size. Conversely, a vehicle that crosses a High segment at normal speed with no slip event and no braking intervention is weak evidence against ice, and it lowers the segment score by a small step. Driver reports, a single button press or voice command for "ice here," are accepted as confirmation with a lower weight than a traction event, because drivers report what they felt and what they feared in the same press.

Over a winter season, the confirmation loop builds a per-segment history: which segments freeze first, at what air temperature, and under which moisture conditions. That history becomes a prior for the next cold night, so a bridge that froze at an air temperature of 2 °C last week is flagged Watch before any vehicle reports a cold reading this week. The prior never overrides live telemetry; it only sets the starting score.

7. Outputs and Consumers

• Driver alerts: A vehicle approaching a High or Confirmed segment within 500 meters receives an audio and visual alert naming the hazard and the distance, and, where the vehicle supports it, a recommended speed reduction. Alerts are suppressed for segments the driver has already crossed without incident in the last hour, to limit alert fatigue.
• Routing: Navigation services incorporate segment risk into route cost, offering an alternate route around Confirmed segments and warning on routes that cross them, in the same way they route around traffic.
• Winter maintenance dispatch: Road agencies receive a ranked list of segments by risk score, with bridge decks and high-traffic segments first, so salt trucks and plows are sent to the segments most likely to be icy rather than patrolling a fixed route. The same list supports anti-icing, applying brine before the freeze, which the FHWA notes takes less salt than melting ice after it forms.
• Open map: A public map shows segment risk at district level with a 15-minute delay, for drivers planning a trip and for researchers studying freeze patterns.

8. Privacy and Data Minimization

Risk is a property of a road segment, not of a driver, and the system is built to keep it that way. Telemetry records are aggregated at the segment level before storage, no record is retained with a persistent vehicle identifier, and segments with fewer than three contributing vehicles in an hour are not published, so a single car on a rural road at 3 a.m. cannot be tracked through the map. Raw TPMS and position records are retained for at most 24 hours for model calibration and then deleted. The public map carries a 15-minute delay and district-level aggregation. A vehicle owner opts in through the telematics settings, and opting out stops transmission entirely rather than degrading it.

9. Figures Description

• Figure 1: System block diagram: four TPMS sensors per vehicle feed a vehicle gateway with speed, ambient temperature, humidity, wiper, and GNSS inputs; the gateway transmits to a cloud aggregator; the aggregator outputs a segment risk map to driver alerts, navigation routing, and maintenance dispatch.
• Figure 2: Tire thermal balance diagram, showing heat in from hysteresis and tread friction, heat out by sidewall convection, rim conduction, and contact patch conduction to the road, with the contact patch as the dominant coupling.
• Figure 3: Time series from a single vehicle crossing a bridge deck on a freezing night: cavity temperature, corrected road surface temperature estimate, dew point, and the risk score, with the bridge crossing marked and the score rising from Watch to High across the deck.
• Figure 4: Segment risk map mockup for a highway interchange at dawn, with 100-meter segments colored Low, Watch, High, and Confirmed, bridge decks outlined, and a salt truck route overlaid on the highest-risk segments.
• Figure 5: Calibration curve: per-sensor temperature error against the number of independent vehicle passes per segment per hour, showing the segment mean error falling from about 3 °C for one vehicle to about 1 °C for ten vehicles.

Implementation Notes

Three deployment paths cover the fleet without new vehicle hardware. The preferred path is OEM telematics: the vehicle manufacturer exposes TPMS temperature and pressure through its connected vehicle platform, and the aggregator consumes the feed under the owner's opt-in. The second path is an aftermarket OBD dongle that reads TPMS data from the CAN bus, for vehicles whose manufacturer does not expose the feed. The third path is fleet telematics providers, which already collect TPMS data for tire management and can forward the temperature channel with the same feed. None of these paths requires a sensor change, a vehicle modification, or a roadside installation.

Calibration is per vehicle model and tire size, not per vehicle. The hysteresis lookup table is built from fleet data during the first winter of operation: cold-start key-on readings give the equilibrium anchor, and traction events give the ground truth. A new vehicle model with no history starts from a generic table for its tire size class and converges as its own data accumulates. Vehicles with a spare tire sensor or a fifth TPMS channel can use the spare as a control, since it experiences ambient air and no rolling heat, but the spare is not required.

A pressure-only fallback exists for vehicles whose TPMS transmits pressure but not temperature. Because pressure follows absolute temperature at fixed volume, a pressure change with no leak implies a temperature change: about 1 psi per 10 °F. The fallback is weak, because a pressure accuracy of plus or minus 0.3 bar corresponds to roughly plus or minus 24 °C of inferred temperature, and because slow leaks masquerade as cooling. It is disclosed as a degraded mode for coarse cold-versus-warm classification, not as a substitute for the temperature channel.

Costs are the cost of software and data, not of hardware. The vehicle side is a telematics configuration change. The cloud side is map matching, aggregation, and a risk map, comparable in scale to existing traffic services. The expensive part is the first winter of calibration, which is a data collection exercise, not a capital program.

Claims

1. A system for mapping black ice risk on road segments, comprising: a plurality of vehicles, each carrying direct tire pressure monitoring system sensors configured to measure the pressure and temperature of the air inside each tire cavity; a vehicle gateway configured to collect the pressure and temperature measurements together with vehicle speed and position; a thermal correction module configured to estimate hysteresis heat generation from the vehicle speed, load, and inflation pressure and to subtract the estimated hysteresis heat and ambient air temperature from the measured cavity temperature to produce a road surface temperature estimate for the road segment under the vehicle; and an aggregator configured to combine road surface temperature estimates from multiple vehicles into a per-segment black ice risk score.
2. The system of claim 1, wherein the thermal correction module calibrates a per-tire offset from a cold-start equilibrium reading taken when the vehicle has been parked for at least two hours, wherein the cavity temperature at key-on is taken as equal to the local road and air temperature.
3. The system of claim 1, wherein the thermal correction module computes segment-to-segment differences in cavity temperature, wherein a change in cavity temperature across a segment boundary cancels the steady hysteresis offset common to both segments.
4. The system of claim 1, further comprising a moisture evaluation module configured to determine moisture presence from at least one of: a dew point at or above the estimated road surface temperature, a rain sensor or wiper active state, precipitation in the preceding 6 hours, or a snowmelt and refreeze pattern.
5. The system of claim 4, wherein the aggregator computes the black ice risk score from a temperature term, a moisture term, and a context term, and wherein the score cannot reach a High band on the temperature term alone.
6. The system of claim 1, wherein the aggregator applies a context term that increases the risk score for bridge decks, overpasses, shaded road cuts, and low-lying segments near water, identified from map data.
7. The system of claim 1, wherein the aggregator weights each road surface temperature estimate by sensor accuracy, the number of contributing vehicles, and the time since the pass, and wherein a segment estimate decays with a half-life of 30 minutes and is dropped after 3 hours without a new pass.
8. The system of claim 1, wherein the aggregator requires at least three independent vehicle passes in the last hour before promoting a segment above a Watch band on temperature evidence alone.
9. The system of claim 1, further comprising a confirmation module configured to promote a segment to a Confirmed band on an anti-lock braking intervention, a traction control slip event, or a stability control correction occurring on that segment, and to feed the confirmation back to recalibrate the thermal correction module.
10. The system of claim 1, further comprising a driver alert module configured to issue an audio and visual alert to a vehicle approaching a High or Confirmed segment within 500 meters, and to suppress the alert for a segment the driver has already crossed without incident in the last hour.
11. The system of claim 1, further comprising a maintenance dispatch module configured to rank road segments by risk score and to direct winter maintenance vehicles to the highest-risk segments, with bridge decks ranked first.
12. The system of claim 1, wherein the aggregator stores risk as a property of a road segment with no persistent vehicle identifier, publishes no segment with fewer than three contributing vehicles in an hour, and deletes raw telemetry records within 24 hours.
13. A method for mapping black ice risk on road segments, comprising: measuring, with direct tire pressure monitoring system sensors, the pressure and temperature of the air inside each tire cavity of a plurality of vehicles; collecting the measurements with vehicle speed and position; estimating hysteresis heat generation from the vehicle speed, load, and inflation pressure and subtracting the estimated hysteresis heat and ambient air temperature from the measured cavity temperature to produce a road surface temperature estimate; evaluating moisture presence from dew point, rain sensor state, or recent precipitation; combining the road surface temperature estimate and the moisture evaluation into a per-segment black ice risk score; and publishing the risk score to driver alerts, navigation routing, or winter maintenance dispatch.
14. The method of claim 13, further comprising promoting a segment to a Confirmed band on a traction control slip event or stability control correction on that segment, and lowering the segment score on a crossing at normal speed with no slip event.

Prior Art References

1. FHWA, "How Do Weather Events Impact Roads?": Weather-related crash statistics, including black ice fatalities and wet pavement crash counts
2. Jin and McBroom, "Investigating Road Ice Formation Mechanisms Using Road Weather Information System (RWIS) Observations," Atmosphere (2024): Three years of Montana RWIS data: ice forms only when pavement is at or below freezing, moisture is the second condition, hoarfrost mechanism, bridges and lake-side roads freeze first
3. FHWA, "Iowa Gets a Jump on Storms with New Technology" (FHWA-SA-96-045): RWIS deployment: pavement temperature sensors, ice detectors, dew point, 22 sites expanding to 47
4. FHWA, "District of Columbia RWIS" (FHWA-SA-96-045): Bridges freeze before surface streets; first RWIS units installed on Chain Bridge and an overpass
5. British Columbia Ministry of Transportation, "Road Weather Information System: Data Definitions": If surface temperature is lower than the dew point, moisture condenses onto the pavement as frost or black ice below freezing
6. Bridgestone, "What is TPMS and How Does it Work?": Direct TPMS sensors may provide tire temperature readings; sensors mounted inside the tire at the valve stem
7. Schrader Electronics, "PF4 Tire Pressure Monitoring Sensor User Manual": Sensor mounted inside each tire periodically measures actual tire pressure and temperature
8. FMVSS 138, 49 CFR 571.138: Tire pressure monitoring systems mandated on new US passenger vehicles since 2007
9. US20140118134A1: Tire pressure monitoring system with pressure sensor and temperature sensor sensing air temperature inside the tire
10. Nguyen et al., "Rolling Tires on the Flat Road: Thermo-Investigation with Changing Conditions through Numerical Simulation," Applied Sciences 13(8), 4834 (2023): Tire temperature rises from hysteresis and tread friction; tread-to-road heat transfer coefficient near 12,000 W/m²·°C; maximum temperature rises with load, speed, and low inflation pressure
11. Namjoo et al., "Evaluating the Effect of Operating Conditions on Temperature Variation Rate of Inner Walls and Inside Inflated Air of Pneumatic Tires," Int. J. Advanced Design and Manufacturing Technology 11(3), 2018, pp. 45-53: Tire reaches thermal steady state roughly 15 minutes after the start of driving; cavity air temperature rises with speed
12. Trelleborg, "TPMS Internal Tire Pressure Sensor" datasheet: Measures pressure and temperature inside a tire; temperature accuracy plus or minus 3 °C; operating range minus 40 °C to plus 125 °C
13. 35 U.S.C. § 102(a)(1): Prior art; printed publications available to the public before the filing date

Limitations

The method in this disclosure has not been validated against measurements from real vehicles. The thermal correction model is a design, not a measured result: the hysteresis lookup table, the 0.45/0.35/0.20 risk weights, the 1 °C safety margin, the 30-minute decay half-life, and the three-pass minimum are engineering estimates stated so they can be tested. A per-sensor temperature accuracy of plus or minus 3 °C is coarse against a 1 °C threshold, and the fleet averaging argument depends on enough vehicles driving the same segments in the same hour, which holds on highways and fails on rural roads at 3 a.m., exactly where a lone driver most needs the warning.

Three failure modes bound what this system can do. First, correlated sensor error: all four tires on one vehicle share the same road and the same weather, so their errors are not independent, and the square-root averaging argument overstates the gain from one vehicle. The averaging works across vehicles, not within them, and a segment with one vehicle an hour is one noisy measurement no matter how many tires it has. Second, moisture evidence is indirect: dew point from a weather service is a grid value, not a road value, and a rain sensor says it rained, not that water is still on the road. A segment can score High on a wet road that drained, or score Watch on a dry-looking road where hoarfrost is forming invisibly. Third, the confirmation loop is biased: traction events confirm ice, but uneventful crossings only weakly disconfirm it, because a careful driver crosses ice without slipping. The map will therefore be slow to clear a segment after the ice melts, and a stale High will send salt trucks to dry roads.

The strongest case against this approach is that it measures the wrong thing well. TPMS cavity temperature is a lagging, filtered, heavily corrected proxy for road surface temperature, and a direct measurement would be better. Infrared pavement thermometers on maintenance trucks measure the surface directly, and fixed RWIS stations measure it continuously. The case for this disclosure is not that the proxy is better than a direct measurement, but that it is already deployed at a scale no direct measurement will ever reach: every new car carries four of these sensors, and the fleet drives every segment every day. A noisy measurement everywhere beats a precise measurement at 47 points, provided the noise is stated honestly and the map says where it is confident and where it is guessing. Where confidence is low, the system must say Watch, not High, and where there is no data at all, it must say nothing.