Public Readiness
Modified Tropical Cyclone Warning System
Updated Tropical Cyclone Categories
Tropical Depression
(TD)
61 km/h or Less
Wind Signal 1
Tropical Storm
(TS)
62–88 km/h
Wind Signal 2
Severe Tropical Storm
(STS)
89–117 km/h
Wind Signal 3
Typhoon
(TY)
118–220 km/h
Wind Signals 3 & 4
Super Typhoon
(STY)
More than 220 km/h
Wind Signal 5
Tropical Depression
(TD)
61 km/h or Less
Wind Signal 1
Tropical Storm
(TS)
62–88 km/h
Wind Signal 2
Severe Tropical Storm
(STS)
89–117 km/h
Wind Signal 3
Typhoon
(TY)
118–184 km/h
Wind Signal 4
Super Typhoon
(STY)
185 km/h or Higher
Wind Signal 5
What is a Tropical Cyclone Wind Signal (TCWS)?
A TCWS is a plain text warning for a land area that may experience winds of at least strong breeze (39 km/h or higher) within 36 hours from the time the signal is issued during a tropical cyclone passage.
- Wind signals warn about general wind strength – local winds near buildings or mountains may be stronger or weaker.
- Signals are issued at the province or city level. Metro Manila gets one shared signal level.
- Signals can skip levels – Signal 1 can jump to Signal 3 if conditions change fast.
- Current system in use since March 23, 2022.
What is a Gale?
A gale is a very strong wind of 62 to 88 km/h: much stronger than an everyday breeze, but weaker than a storm or typhoon. How to tell them apart:
- BreezeBelow 39 km/hLeaves and small branches move; flags flutter.
- Strong wind39 to 61 km/hBig branches and whole trees sway; umbrellas are hard to use. (Wind Signal No. 1)
- Gale62 to 88 km/hTwigs and small branches snap off; walking against the wind is hard; loose roofing, signs and tarpaulins can fly. (Wind Signal No. 2)
- Storm or typhoon89 km/h and upTrees fall and roofs are torn off. (Wind Signal No. 3 and higher)
GALE-CAST counts a day as a gale day when the wind reaches 62 km/h.
How GALE-CAST Risk Levels Relate to Wind Signals
GALE-CAST rates each area by forecast wind speed. PAGASA’s signals remain the official warnings; this shows roughly where each GALE-CAST level sits.
LowBelow 39 km/hBelow Signal 1Normal conditions. No wind precautions needed.
Moderate39–61 km/hAbout Signal 1Secure loose outdoor items; take care on motorcycles and bicycles.
High62 km/h and Above (Gale)Signal 2 and AboveStay indoors during peak winds and follow PAGASA and LGU advisories.
Individual Wind Signal Levels
Click a card to see its potential impacts and what to do. Click it again to close it.
Tap a card to see its potential impacts and what to do. Tap it again to close it.
Signal
Tropical Depression (TD)
39–61 km/h
22–33 kt
Show impacts and what to do
Potential Impacts:
- Houses of poor construction (wood frame, bamboo) may suffer minimal to minor damage.
- Some banana plants tilt; twigs of small trees may sway. Rice crops in flowering/ripening stages slightly damaged.
- Minimal disruption to public transportation.
What to do
- Secure loose outdoor items such as pots, signs and laundry.
- Check your roof, windows and nearby tree branches.
- Prepare a go-bag: water, food, flashlight, medicines and a charged phone.
- Monitor PAGASA advisories and local announcements.
Signal
Tropical Storm (TS)
62–88 km/h
34–47 kt
Show impacts and what to do
Potential Impacts:
- Minor to moderate damage to makeshift or old structures. Homes may have minor roof damage.
- Unsecured lightweight items may become projectiles. Some electrical wires blown down.
- Minor to moderate disruption to public transportation.
What to do
- Bring loose items inside and tie down light roofing.
- Avoid unnecessary travel, especially by motorcycle, bicycle or small boat.
- Keep water, food and medicines for at least 3 days; charge phones and power banks.
- If your house is weak or near the coast, know your nearest evacuation center.
Signal
Severe Tropical Storm (STS)
89–117 km/h
48–63 kt
Show impacts and what to do
Potential Impacts:
- Makeshift structures suffer substantial damage. Poor/average homes: considerable roof damage.
- Many areas experience power outages: downed power lines and posts.
- Heavy damage to rice crops. Some small trees and most banana plants downed.
What to do
- Stay indoors, away from windows and glass doors.
- Evacuate early if your LGU says so, or if your home is weak or in a flood-prone area.
- Turn off the main power and gas if strong winds or flooding reach your area.
- Keep the emergency hotlines ready.
Signal
Typhoon (TY)
118–184 km/h
64–99 kt
Show impacts and what to do
Potential Impacts:
- Severe damage to all light structures. Poor/average homes: complete roof failure, possible wall collapse.
- Considerable airborne debris – risk of injury or death. Near-total loss of power and telecommunications.
- Banana plants mostly destroyed. Almost total loss of rice crops.
What to do
- Evacuate when your LGU tells you to. Do not wait for the strongest winds.
- If you stay, shelter in the strongest inner room, away from windows.
- Stay away from power lines, trees and the coast.
- Do not go out during a sudden calm; strong winds can quickly return.
Signal
Super Typhoon (STY)
185 km/h or Higher
100 kt+
Show impacts and what to do
Potential Impacts:
- Severe to catastrophic damage to ALL structures. Most industrial buildings destroyed.
- Electricity, water, and telecommunications unavailable for prolonged periods.
- Vast majority of trees broken, defoliated, or uprooted.
What to do
- Follow evacuation orders at once: this storm is life-threatening.
- Shelter in a sturdy building, in the innermost room or under a strong table.
- Keep your go-bag, flashlight and charged phone with you at all times.
- After the storm, wait for the all-clear and watch for fallen wires and debris.
Emergency Hotlines
For help during a gale or any emergency. On a phone, tap a number to call.
Gale Warnings on This Device
About GALE-CAST
GALE-CAST forecasts where strong, gale-force winds may hit Metro Manila, so residents and disaster officers can prepare before the wind arrives.
Research titleGALE-CAST: A Dual-Stream Physics-Informed ANN–LSTM–Attention Hybrid Algorithm for Localized Gale Hazard Mapping in Metro Manila
Built by Computer Science students at Our Lady of Fatima University. It uses a
physics-informed hybrid model that combines Artificial Neural Networks (ANN),
Long Short-Term Memory (LSTM) networks and an attention mechanism, trained on
historical PAGASA station observations.
Physics-Informed Inputs
Besides the raw daily readings from each PAGASA station, the model works with these physically meaningful quantities.
Wind vector (u, v)
Wind is represented by east–west and north–south components so changes in direction can be compared consistently.
How it works
u = −W·sin θ v = −W·cos θ
W is the wind speed and θ the direction the wind blows from, in degrees clockwise from north. u is the east–west part (positive: a westerly wind, blowing toward the east); v is the north–south part (positive: a southerly wind).
From u and v the direction is taken as a sine and cosine (θ = atan2(−u, −v)): the previous day's direction at every station is an input, and the next day's direction is one of the values the model forecasts.
Why: an angle jumps from 359° to 0° although the two winds are almost the same; the vector form keeps them close, so the model can follow a shift in wind regime, such as the southwest monsoon (habagat) setting in.
Pressure tendency (ΔP/Δt)
Day-to-day pressure changes help the model identify developing weather patterns.
How it works
ΔP/Δt = Ptoday − Pyesterday (hPa per day)
P is the station's mean sea-level pressure. A negative value means pressure is falling; the site shows its opposite as the pressure drop.
Why: a sustained fall in pressure marks an approaching low or tropical cyclone, and the steeper pressure gradient that follows drives stronger wind, so falling pressure often comes before gale-force winds. In the training labels, a fall of 6 hPa or more in a day counts as High risk even before the wind reaches 62 km/h, and 3–5 hPa as Moderate.
Change over time
Recent wind and pressure patterns give the model context for the next forecast.
How it works
The model also reads the day-to-day change in wind speed (wind acceleration), the wind speed of the two previous days, the previous day's pressure, and the 3-day average, maximum and spread of wind speed.
Why: gales build over days. The LSTM and attention layers use this sequence to judge whether the next day is heading toward gale conditions, not just how one day looks.
How Well the Model Performs
The figures below summarize the original model's historical performance using PAGASA station records from 2015–2024. The evaluation also helped guide model development.
—AccuracyDays called correctly, gale or no gale
—PrecisionGale forecasts that turned out right
Key for warnings—RecallReal gale days the model caught
—F1 ScoreBalance of precision and recall
—PR-AUCSkill on rare gale days (1.0 is perfect)
Forecast vs. What Happened
Confusion matrix of gale / no-gale forecasts in the validation period
Predicted: No Gale
Predicted: Gale
Actual:
No Gale
-Correct: no gale
-False alarm
Actual:
Gale
-Missed gale
-Correct: gale caught
Correct forecastError
This chart compares GALE-CAST's forecasts with what really happened. Green boxes are correct forecasts (true negatives and true positives). Amber boxes are mistakes: a false alarm (false positive) warned of a gale that never came, and a missed gale (false negative) gave no warning, the most serious error for safety.
How the scores above are calculated: precision = true positives ÷ all gale forecasts; recall = true positives ÷ all real gale days.
Forecast Error by Weather Variable
How far each forecast value was from the historical value (lower is better)
| Variable | MAEAverage error | RMSEWeights big misses | MASEvs. simple baseline |
|---|
| Wind Speedkm/h | - | - | - |
| Wind Directiondegrees | - | - | - |
| PressurehPa | - | - | - |
| Humidity% | - | - | - |
| Rainfallmm | - | - | - |
Within the team's target error (target shown under each value)Above target
MAE and RMSE are in each variable's own unit. MASE compares the model with a simple “same as yesterday” forecast: below 1 means GALE-CAST did better.
The Team Behind GALE-CAST
Computer Science students, Our Lady of Fatima University
Christine S. Tamayo
Project Manager / UI/UX Designer / Front-End Developer
Nicole Courtney S. Delas Alas
Back-End Developer / Machine Learning Developer
Lynyer D. Pascual
Support Team
Andrea Christein A. Huerte
System Architect / QA
