Temperature prediction markets on Kalshi offer binary contracts on whether a city's official daily high or low temperature will reach specific thresholds, typically structured in 2-5°F brackets. These markets settle against the official NWS Automated Surface Observing System (ASOS) readings as recorded in the daily Climate Report (CLI), with contracts resolving YES if the temperature meets or exceeds (for highs) or falls to or below (for lows) the specified threshold. Record temperature markets, available during extreme heat or cold events, settle based on whether a location breaks its historical daily temperature record as maintained in the NWS Cooperative Observer Program database.
Temperature markets represent some of Kalshi's most liquid weather contracts, with daily trading volumes regularly exceeding $50,000 during summer heat waves and winter cold snaps across major cities. Market depth typically concentrates in the 5-7 days before expiry when ensemble forecast spreads narrow, though volatility spikes occur 24-48 hours out when high-resolution models like the NAM and HRRR begin showing boundary layer temperature solutions that diverge from earlier GFS/ECMWF guidance. Record temperature contracts see explosive volume growth when deterministic models place cities within 3-5°F of historical thresholds, with contracts often trading at 15-30% implied probability even when historical base rates suggest 1-3% annual occurrence.
The informational edge in temperature markets derives from understanding mesoscale processes that operational NWS forecast offices struggle to capture in automated guidance: urban heat island effects that add 3-7°F to official readings in downtown stations, downslope wind patterns that can spike temperatures 10-15°F above model solutions in cities like Denver and Reno, cold air damming in the Appalachians that keeps cities like Charlotte and Greenville 8-12°F below guidance, and marine layer persistence that creates systematic cool bias in coastal California summer forecasts. Traders who monitor 00z and 12z model runs alongside mesoanalysis products like the RAP model's 13z initialization can identify when deterministic forecasts lag developing atmospheric conditions, particularly in shoulder seasons when diurnal heating patterns shift rapidly.
Kalshi temperature contracts specify an exact temperature threshold and settlement date, with the contract resolving YES if the official ASOS daily maximum temperature (for high temperature contracts) meets or exceeds the threshold, or if the daily minimum temperature (for low temperature contracts) falls to or below the threshold. Settlement uses the daily maximum and minimum temperatures as recorded in the NWS CLI report for the specified primary ASOS station—for New York this is KLGA (LaGuardia), for Chicago it's KORD (O'Hare), for Los Angeles it's KLAX. The CLI report aggregates the highest and lowest temperatures recorded during the 24-hour period from midnight to midnight local time, rounded to the nearest whole degree Fahrenheit. Contracts typically expire at 11:59 PM ET on the settlement date, with official settlement occurring the following morning after the CLI report is published, usually between 1:00-3:00 AM ET.
Temperature measurement at NWS ASOS stations employs platinum resistance thermometers housed in aspirated radiation shields, with sensors positioned at a standard height of 1.5 meters above ground level. The ASOS system samples temperature every 10 seconds and computes one-minute averages, reporting observations hourly in METAR format while continuously tracking daily maximum and minimum values. These thermistors have an accuracy specification of ±0.5°F under optimal conditions, though measurement uncertainty increases during precipitation events, equipment icing, or aspiration fan malfunctions. The recorded daily high represents the peak one-minute average temperature observed during the 24-hour period, not an instantaneous reading, which creates important edge cases when temperatures briefly spike during advective warming events or downdraft pulses.
Settlement edge cases arise most frequently when daily extremes occur near the contract threshold and measurement uncertainty becomes material. A recorded temperature of 90°F might represent a true atmospheric temperature anywhere between 89.5°F and 90.5°F given instrumentation tolerances, but NWS reporting standards round to the nearest whole degree, making 90°F the official value regardless of fractional components. Station malfunctions or periods of missing data create ambiguity: if an ASOS station experiences a 2-4 hour observation gap during the afternoon heating peak, the CLI report may substitute data from backup sensors or nearby cooperative observer stations, occasionally producing temperatures that diverge from standard ASOS readings. During extreme cold events below -40°F or extreme heat above 115°F, platinum resistance thermometers approach the edge of their calibration range, introducing additional measurement drift that can affect record temperature settlement determinations.
The optimal trading window for temperature contracts begins at 12z (7 AM ET) when the GFS, NAM, and ECMWF models complete initialization with the morning radiosonde network data. The 12z model suite provides the first robust guidance incorporating overnight boundary layer evolution, which is critical for assessing whether nocturnal cold pools will persist through morning mixing or if radiational heating will proceed unimpeded. Traders should compare deterministic model point forecasts against ensemble spread products (GEFS, EPS) at the 48-72 hour lead time: when ensemble spreads exceed 8-10°F for daily highs or 6-8°F for lows, market odds frequently misprice tail scenarios, particularly if the deterministic operational forecast sits near one standard deviation from the ensemble mean. The 00z model run provides updated guidance incorporating afternoon maximum temperature observations, making the window between 8:00-10:00 PM ET valuable for identifying when diurnal heating exceeded or fell short of model expectations, signaling systematic bias for subsequent days.
Divergence between NWS forecast office discussion products and private sector forecasts (Weather.com, AccuWeather) creates exploitable mispricing in temperature markets, particularly in complex terrain or coastal zones. When NWS area forecast discussions explicitly mention mesoscale uncertainty—phrases like "considerable spread in boundary layer mixing depths" or "uncertain timing of frontal passage"—market odds often fail to reflect this forecast uncertainty, presenting value on both sides of threshold brackets. For summer high temperature markets in interior western cities (Phoenix, Las Vegas, Sacramento), fading the consensus when 12z models show 500mb ridging and 850mb temperatures above +20°C often provides edge, as downslope warming and urban heat island effects routinely add 3-5°F to model solutions. Conversely, winter low temperature markets in Great Plains cities (Minneapolis, Kansas City, Chicago) create value on colder outcomes when snow cover exceeds 6 inches and 00z models indicate radiational cooling under clear skies, as urban ASOS stations cool 2-4°F more than models resolve once nocturnal drainage flows establish.
Record temperature markets demand specialized strategy due to their low base rates and explosive implied volatility. These contracts become tradeable when deterministic models place a location within 5°F of the daily record with 3-5 days lead time. The key signal is ensemble clustering: if 30+ GEFS members show temperatures within 3°F of the record, the market typically underprices the probability at lead times beyond 48 hours, as casual traders anchor to the low historical base rate rather than the conditional probability given the synoptic pattern. Monitor the 18z NAM and HRRR runs at 36-48 hour lead times for these markets—high-resolution models that initialize closer to the event frequently show mesoscale temperature details (mixing depths, advection timing, cloud cover breaks) that shift peak temperatures 4-7°F relative to coarser GFS/ECMWF solutions, often moving the outcome from possible to probable or vice versa.
Temperature market liquidity follows a pronounced seasonal pattern with summer (June-August) and winter (December-February) generating 65-70% of annual trading volume, driven by extreme heat and cold events that produce clear threshold-crossing scenarios. Summer high temperature markets peak in late July and early August when persistent upper-level ridging creates multi-day heat waves across the southern and central U.S., with contracts on cities like Phoenix, Las Vegas, Dallas, and Houston seeing daily volumes 3-4x higher than spring or fall baseline. Winter low temperature markets concentrate in January and February, particularly during polar vortex disruptions that send Arctic air masses into the Great Plains and Midwest; contracts on Minneapolis, Chicago, and Kansas City routinely exceed $20,000 in daily volume during these events. Record temperature markets show even more extreme seasonality, with 80%+ of volume occurring during the dozen or so days per year when synoptic patterns support potential record breaks.
Predictability varies significantly by season and region, with summer temperature forecasts generally more reliable than winter due to higher amplitude diurnal signals and reduced frontal complexity. Summer high temperature forecasts for interior western cities achieve skill scores above 0.85 at 3-day lead times, as persistent 500mb ridge patterns create low forecast uncertainty, often making markets efficient beyond 72 hours out. Winter low temperature forecasts in continental climates show much higher uncertainty due to snow cover albedo effects, nocturnal boundary layer stability, and cold air damming, frequently maintaining ensemble spreads of 8-12°F at 48-hour lead times. Shoulder seasons (March-April, October-November) present the highest forecast uncertainty and greatest trading opportunity, as rapid transitions between air masses and variable cloud cover create 10-15°F swings in daily maximum temperatures that models struggle to time precisely, generating systematic mispricing in contracts expiring during these periods.
Temperature markets settle based on the official daily maximum or minimum temperature recorded in the NWS Climate Report (CLI) for the designated ASOS station. The contract resolves YES if the high temperature meets or exceeds the specified threshold (for high temperature contracts) or if the low temperature falls to or below the threshold (for low temperature contracts). Settlement occurs the morning after the contract date once the CLI report is published, typically between 1:00-3:00 AM ET. Temperatures are rounded to the nearest whole degree Fahrenheit in official NWS reports.
Each city's temperature markets settle against a specific primary ASOS station: New York uses KLGA (LaGuardia), Chicago uses KORD (O'Hare), Los Angeles uses KLAX, Miami uses KMIA, Phoenix uses KPHX (Sky Harbor), and Seattle uses KSEA (Sea-Tac). These are typically the major airport stations that serve as the official reporting location for each city's weather records. The exact station identifier is specified in the Kalshi contract details and matches the station used in NWS forecast products for that city.
The optimal trading window is after the 12z (7 AM ET) model runs when GFS, NAM, and ECMWF guidance incorporates morning radiosonde data and overnight boundary layer observations. The 48-72 hour lead time offers the best risk-reward as ensemble forecasts begin converging while market odds often lag model updates. A secondary window occurs after the 00z (7 PM ET) run when afternoon temperature observations reveal whether models over or underperformed, signaling systematic bias for subsequent days. Avoid trading beyond 5-day lead times when ensemble spread exceeds 12-15°F and forecast skill drops substantially.
Mispricing commonly occurs when mesoscale effects aren't captured in deterministic models: urban heat islands add 3-7°F to official readings at downtown stations, downslope winds spike temperatures 10-15°F in mountain-adjacent cities like Denver, cold air damming keeps Appalachian cities 8-12°F below model guidance, and marine layers create persistent cool bias in coastal California. Markets also misprice during shoulder seasons when rapid air mass transitions create timing uncertainty, and during extreme events when traders anchor to historical base rates rather than conditional probabilities given the current synoptic pattern. Divergence between NWS area forecast discussions and automated model output signals exploitable uncertainty.
Record temperature markets resolve YES only if the daily high or low breaks the historical record for that specific date at the designated station, as maintained in the NWS Cooperative Observer Program database. These markets appear only when forecast models show potential for record-breaking temperatures, typically when deterministic solutions place a city within 5°F of the daily record. They carry much lower base rates (1-3% annual probability) but higher implied volatility, with contracts often trading at 15-30% when models show temperatures within 3°F of records. Settlement requires the temperature to exceed the existing record, not merely tie it.
Consumer weather apps often display temperature from personal weather stations, smartphone sensors, or interpolated model data rather than official ASOS observations. Official ASOS stations use calibrated platinum resistance thermometers in aspirated radiation shields at standard 1.5-meter height, while personal stations may have poor siting (near buildings, pavement, or heat sources) and uncalibrated sensors. For Kalshi settlement purposes, only the official ASOS temperature recorded in the NWS CLI report matters—readings from weather apps, personal stations, or even nearby ASOS stations are irrelevant. Always verify contract settlement source before trading.
Snow cover creates strong radiational cooling enhancement that frequently causes official low temperatures to fall 3-6°F below model forecasts, particularly when snow depth exceeds 4-6 inches and clear skies allow unimpeded longwave radiation loss. The high albedo of snow cover (0.80-0.95) also suppresses daytime heating, allowing nocturnal cooling to begin from a lower baseline temperature. Models systematically underestimate this snow-albedo feedback in the 24-48 hours after fresh snowfall, creating value on colder outcomes in low temperature markets. This effect is strongest in urban ASOS stations where snow cover persists despite heat island effects.
Ensemble spread measures the range of temperature outcomes across multiple model runs with slightly varied initial conditions, quantifying forecast uncertainty. When the GEFS or EPS ensemble shows spread exceeding 8-10°F for daily highs at 72-hour lead time, it indicates low confidence in the forecast, meaning outcomes at the tails of the distribution are more probable than markets typically price. Conversely, ensemble spreads under 4-5°F at 48-hour lead time indicate high forecast confidence and efficient market pricing. Traders should compare where deterministic forecasts sit relative to ensemble percentiles—if the operational GFS forecast is at the 30th percentile of the GEFS ensemble, warmer outcomes are statistically more likely.
Want to track these markets in real time?
Open Dashboard →