Migraine forecast · How it works
Migraine forecast: your daily risk estimate, explained
Short answer: CalmGrid's migraine forecast combines local weather and air-quality data with your personal signals — sleep, HRV, menstrual-cycle phase, recent attacks, and weekday patterns — into a daily score from 0 to 100, with the contributing reasons ranked so you can see what is driving it. It is an informational estimate, not a medical device or diagnosis, and it cannot say whether you will have an attack.
Last updated: 2026-10-16
What the forecast looks at
Each signal below shows the threshold CalmGrid uses and why it is included. Thresholds are the app's own operational cutoffs, not clinical diagnostic criteria — population studies support the direction of these associations, but the exact numbers at which an individual is affected vary from person to person.
Weather and air
Low barometric pressure
Flagged below 1007 hPa, stronger below 1000 hPa. A meta-analysis of 31 studies linked ambient pressure to migraine attacks, and reviewers consider low pressure a likely trigger — though studies measure it inconsistently. More on barometric pressure
[1] [2] [8]
Rapid pressure drop
Flagged when pressure falls 5 hPa or more in 24 hours, stronger at 8 hPa or more. A systematic review found that pressure drops or rapid fluctuations were tied to more frequent attacks in several studies, while effects on severity were unclear and no study linked pressure to attack duration.
[2] [8]
Temperature swing
Flagged when the 24-hour temperature range reaches 8 °C or more. Meta-analytic evidence links temperature to migraine attacks; CalmGrid operationalizes this as a day-to-day swing rather than an absolute temperature. More on weather changes
[1]
High humidity
Flagged at 85% or higher. The evidence here is the weakest of the weather signals: the largest meta-analysis found no significant association between humidity and migraine attacks, so treat this as a minor contributing signal, not a proven trigger.
[1]
Air quality (PM2.5)
Flagged above 35 µg/m³, stronger above 55 µg/m³. Higher PM2.5 is linked to more migraine-related clinical visits in meta-analysis, and a large Canadian case-crossover study tied wildfire-sourced PM2.5 to about 6% more emergency visits for migraine and other primary headaches — though background (non-wildfire) PM2.5 was not significant in that study.
[1] [3]
Pollen
Flagged at 50 grains/m³ or higher. The support is indirect: a meta-analysis of over 4.8 million people found allergic rhinitis associated with roughly 2–3 times the migraine risk, so pollen counts act as an allergy-season proxy rather than a directly proven migraine trigger.
[7]
Body and patterns
Short or reduced sleep
Flagged when you sleep 90+ minutes less than your recent baseline, or under 6 hours. A UK Biobank prospective cohort of over 480,000 people tied healthier sleep patterns to substantially lower migraine risk, and sleep disturbance is among the most commonly reported triggers. More on lack of sleep
[4] [8]
HRV drop
Flagged when heart-rate variability falls 20% or more below your baseline (iOS HealthKit, optional). A small wearable study found sleep-time HRV changes on pre-migraine nights, but with only 23 participants and large individual differences — promising as a personal signal, not established science.
[5]
Menstrual cycle phase
Weighted more strongly during menstruation than during estimated ovulation. Falling estrogen before menstruation is the leading explanation for perimenstrual attacks, though reviewers note the supporting evidence is still limited by small samples and inconsistent methods. More on menstruation · More on hormonal changes
[6] [8]
Recent attacks
Flagged when you logged an attack in the last 3 days, weighted higher for two or more. Attacks tend to cluster: the threshold model of migraine holds that triggers lower a threshold that is already depressed around an attack, and diary-based forecasting research uses recent headache days as predictive features.
[9] [11]
High-risk weekday
Flagged when a weekday carries at least 1.5× your average weekday attack count over the last 90 days (needs at least 8 logged attacks, so it cannot fire on noise). This is purely your own diary statistics — the same kind of personal history that diary-based forecasting models rely on.
[9]
Weekend timing
A small additional signal on Saturdays and Sundays. It is a rough proxy for the clinically described "letdown" effect, where attacks follow the release of stress — reviewers list letdown after stress among the verified migraine triggers. Routine changes (sleeping in, caffeine timing) may contribute too.
[8]
How the score works
Every signal present adds points to a base score, producing a 0–100 estimate. Reasons are ranked by their point contribution, so the first reason is the largest driver of that day's number.
| Band | Score | Meaning |
|---|---|---|
| Low | 0–29 | Few or no risk signals active. |
| Moderate | 30–59 | Some signals present; worth glancing at the reasons. |
| High | 60–79 | Several signals stacking up; notifications can fire at this level. |
| Very high | 80–100 | Many signals combined; the strongest warning level. |
Today vs tomorrow: today's estimate uses the full picture — current weather and air quality plus sleep vs your baseline, HRV vs baseline, cycle phase, recent attacks, and calendar patterns. Tomorrow's estimate uses forecast weather, the projected cycle phase, recent attack history, and day-of-week patterns only, because sleep and HRV cannot be known in advance.
Notifications: if you opt in and grant permission, CalmGrid can notify you at your chosen time when today or tomorrow reaches high or very high risk, naming the top contributing reasons.
Free vs Pro: free users see the band and the top reason. The numeric score, all ranked reasons with their point values, signal readings, and the 7-day history chart (the selected day plus the six previous saved days — a look back, not a 7-day forward forecast) are Pro features.
What a forecast can and can't do
Association is not causation. Population studies link weather, sleep, and air quality with migraine — but a signal being active does not mean it caused your attack. Early migraine symptoms (fatigue, neck stiffness, light sensitivity) are easily mistaken for triggers, so even careful diary patterns can mislead. Researchers describe triggers and premonitory symptoms as intertwined, with the hypothalamus integrating both. [8] [11]
Individuals vary widely. The HRV prediction study emphasized large individual differences in physiological responses, and trigger reviews stress that natural triggers are far less predictable than laboratory ones and differ between people. Your own logged history matters more than any population average. [5] [8]
Missing data lowers precision. Weather needs location permission; sleep, HRV, and cycle inputs need optional Apple Health access. When signals are missing, the score is computed from whatever remains — and tomorrow's estimate never has sleep or HRV.
It is not a diagnosis. The forecast cannot tell you whether a headache is a migraine, and it is not a medical device. If your pattern changes, attacks become more frequent, or treatment stops working, consult a doctor.
Can you really predict migraines?
The honest answer is sometimes, partially, and never with certainty. Attacks often begin 24–48 hours before head pain, with brain-network changes producing premonitory symptoms such as fatigue, neck stiffness, or sensory sensitivity. In a 3-month electronic-diary study, symptoms like nausea, light and sound sensitivity, and difficulty concentrating raised the odds of head pain within about 24 hours — but performance was modest, and the authors concluded these symptoms should not be stand-alone predictors. [10] [11]
Weather sensitivity has real support: across 31 studies, people reporting weather as a trigger had substantially more attacks, with temperature and pressure significantly associated — while humidity was not. Barometric-pressure reviews point the same way but flag inconsistent methods and small, regional samples. [1] [2]
Purpose-built forecasting models exist but remain early-stage: a prospective study of 18 patients combining phone diaries with heart rate, skin temperature, and muscle tension reached an AUC of 0.62 for next-day headache — better than chance, far from reliable. CalmGrid takes the transparent route instead of a black box: a rule-based score whose every reason you can inspect, grounded in published associations. [9]
Frequently Asked Questions
Can an app really predict a migraine?
No app can predict an individual migraine with certainty. Research shows that premonitory symptoms can raise the odds of head pain within about 24 hours, but their predictive performance is modest and they should not be used as stand-alone predictors. Diary-and-wearable forecasting models have reached only moderate accuracy in small studies. CalmGrid gives an informational daily risk estimate with the reasons shown, not a prediction that an attack will happen.
What raises my migraine forecast score in CalmGrid?
The largest single contributions come from a steep 24-hour pressure drop, very low pressure, the menstrual phase, and recent attacks, with smaller contributions from air quality, pollen, temperature swings, humidity, short sleep, low HRV, and weekday or weekend timing. Each reason is ranked by its point contribution so you can see what matters most that day.
Why is tomorrow's forecast less detailed than today's?
Sleep and HRV cannot be known in advance, so tomorrow's estimate uses forecast weather, the projected cycle phase, recent attack history, and day-of-week patterns only. Today's estimate additionally compares last night's sleep and HRV against your personal baselines when that Health data is available.
Does CalmGrid diagnose migraine?
No. CalmGrid is a tracking and informational tool, not a medical device. The migraine forecast is an informational estimate, not a diagnosis, and it cannot tell you whether a specific headache is a migraine.
Is the migraine forecast free?
Free users see the risk band (low, moderate, high, or very high) and the top contributing reason. The numeric 0–100 score, all ranked reasons, signal values, and the 7-day history chart are Pro features. Forecast notifications for high or very high risk today or tomorrow can be enabled separately and require notification permission.
What should I do on a high-risk day?
Treat it as a prompt to keep a steady routine: regular sleep, meals, hydration, and stress management, and take medication only as your clinician has prescribed. A diary of what happened on high-risk days can make your next clinical visit more productive. If your pattern changes, becomes more frequent, or stops responding to treatment, talk to a doctor.
How is this different from checking a weather app?
A weather app shows the same sky for everyone. CalmGrid combines local weather and air quality with your personal signals — sleep and HRV baselines, cycle phase, your own attack history, and weekday patterns — into one 0–100 estimate, and ranks the reasons behind it so the number is inspectable.
Start Tracking Your Migraines Today
CalmGrid helps you identify patterns, track triggers, and share reports with your doctor.
Download CalmGridPrefer the App Store directly? View CalmGrid on the App Store.
Sources
- [1] Li S, Liu Q, Ma M, et al. Association between weather conditions and migraine: a systematic review and meta-analysis. J Neurol. 2025;272(5):346. PubMed
- [2] Farah A, Adam Y, Ahmed O, et al. Impact of barometric pressure changes on the severity, frequency, and duration of migraine attacks: a systematic review of the literature. Cureus. 2025;17(11):e96821. PubMed
- [3] Dales R, Paul N, Cleland SE, et al. Wildfire-sourced PM2.5 and emergency department visits for migraine and other primary headache. JAMA Netw Open. 2026;9(7):e2625015. PubMed
- [4] Zhang L, Xu Y, Lei Y, et al. Associations of sleep, sedentary behavior, and physical activity with incident migraine: a prospective cohort study. BMC Public Health. 2026;26(1):1365. PubMed
- [5] Jankeviciute R, Kapustynska V, Andruskevicius S, et al. Heart rate variability as a predictor of migraine: sleep-time data analysis of pre-migraine nights. Technol Health Care. 2026;34(3):422-434. PubMed
- [6] Raffaelli B, Do TP, Chaudhry BA, et al. Menstrual migraine is caused by estrogen withdrawal: revisiting the evidence. J Headache Pain. 2023;24(1):131. PubMed
- [7] Wu J, Wang C, Meng F, Wang H. A systematic review and meta-analysis on the risk of migraine in patients with allergic rhinitis. Front Med (Lausanne). 2026;13:1766176. PubMed
- [8] Marmura MJ. Triggers, protectors, and predictors in episodic migraine. Curr Pain Headache Rep. 2018;22(12):81. PubMed
- [9] Stubberud A, Ingvaldsen SH, Brenner E, et al. Forecasting migraine with machine learning based on mobile phone diary and wearable data. Cephalalgia. 2023;43(5):3331024231169244. PubMed
- [10] Lacritz M, Lipton RB, Metts C, Seng EK. The role of prodromal symptoms in predicting headache onset: a longitudinal electronic momentary assessment study. Headache. 2026;66(8):1767-1778. PubMed
- [11] Sebastianelli G, Atalar AC, Cetta I, et al. Insights from triggers and prodromal symptoms on how migraine attacks start: the threshold hypothesis. Cephalalgia. 2024;44(10):3331024241287224. PubMed