44°C Urban Cool Island Design Bureau

AHALab · 2026.07.18 · Urban Action-Based PBL

Build a Cool Island That Can Read the Weather

Students measure heat, observe heat, and change heat for people actually staying on campus. The project starts from one square meter, connecting the body, materials, city, El Niño, typhoon paths, and campus early warning actions.

Accept Assignment

Course Design Demo: Not yet connected to real-time weather. Weather probabilities, typhoon paths, and rankings are only for demonstrating the learning process; on-site projects still require measurement, construction, and re-measurement.

Campus Cool Island Section A cool island including shading, plants, sensors, seating, and drainage modules, capable of switching between high-temperature use, strong wind storage, and post-rain inspection states. Sensor LocationReal Stay TaskPlants and Water Cycle
"Choose a real location on campus, create a cool island of about one square meter, make a real user willing to stay, and prove with data and experience how it works."
Cycle
6 × 120 Minutes
Team
8–12 People Mixed-Age Group
Space
Approximately 1 m² Prototype
Public
Hot Clinic Open Day

From Skin to a Whole Pacific Ocean

Students move between four time scales. Each judgment records location, time period, indicators, evidence, and the next update time.

The Body Tells Us First

The back of the hand feels the chair hot, the side of the face feels the wind, shadows tell us where the sun is coming from. Students use body language to create the first thermal sensation map.

Evidence: thermal sensation words, postures, dwell time, on-site photos

Four age groups share one real artwork

Age differences become team capabilities. Each group shares one user, one location, one set of data, and one cooling island.

K1–3Experiential ObserverBody thermal sensation, shadow drawing, character interviews, icon testing, 30-second storytelling.
K4–6Materials and RecorderMeasure time, material quantity, water usage, photos, experiment order, and visitor feedback.
K7–9Data and Prototype EngineerSensor nodes, CSV, heat maps, structural connections, fault localization, and game rules.
K10–12System Design and ResearcherThermal balance, multi-objective trade-offs, AI models, meteorological data, budget, safety, and public defense.

Six actions to achieve one city change

Expand each lesson to view the 120-minute process, teacher follow-up questions, cross-age responsibilities, and stage evidence. The class progresses continuously through discovery, experimentation, making, and public presentation.

L01Where Does Heat Come FromPhysical Roaming · Real Commission

In the same campus, who is experiencing what kind of heat?

  1. Choice of Two ChairsTouch, observe shadows, feel the wind, and first write predictions.
  2. Campus Heat-Sensing RoamGroups of four, using stickers to record sunny, stuffy, scalding, windless, and waiting experiences.
  3. Meet Real UsersConduct short interviews with security guards, pick-up/drop-off staff, outdoor workers, or classmates.
  4. Heat Problem DictionaryConnect sensations to places, times, materials, and actions.
  5. Accept AssignmentChoose a user and a candidate location to complete a 30-second problem statement.
L02Make Heat VisibleTool Detective · Measurement Agreement

When instruments give different answers, how to find the meaning of each answer?

  1. Guess the Material TemperatureOrder: Black Cloth, White Cloth, Metal, Wood, and Grass
  2. Tool Identity InvestigationWhat evidence do air, surfaces, humidity, wind speed, and body heat each leave?
  3. Side-by-Side CalibrationReadings of the same type of instruments together, record the differences and time
  4. Sun/Shade PairingUnify height, time, distance, angle, and material
  5. Cross-Group RetestingAnother group reproduces according to the protocol and proposes revisions
L03Four Types of CoolnessShading · Reflection · Evaporation · Ventilation

Which combination can make the target users really willing to stay?

  1. Combination betChoose the most promising cooling combination for each group and explain the reasons.
  2. Four-station rotationUse the same-size model to complete A/B tests for shading, reflection, evaporation, and ventilation.
  3. Unexpected result consultationPick an unexpected data point, change one variable, and try again.
  4. Cross-age replicationSwap protocols and materials so that another group can complete the retest.
  5. Mechanism selectionChoose the modules entering the cool island and write down the weather conditions for suitability.
L04AI Partnering with the SkyScheme generation · Typhoon path · Early warning chain

How to turn a forecast into a timely, clear, and campus-appropriate action?

  1. Sky BriefingOn-site observation, official forecast, model updates, risks, and actions.
  2. El Niño InvestigationPlace sea temperature anomalies, seasonal probabilities, and local data on two timelines.
  3. Markov Weather StationGenerate a transition matrix from a state sequence, predict next-period probabilities, and wait for real-time scoring.
  4. Tracking a TyphoonConnect five points in time, predict the sixth, then compare with paths from multiple models.
  5. Pangu Path ExperimentRead public samples, track low-pressure centers, and calculate positional errors and lead times.
  6. Mazu Warning RelayObservers, modelers, forecasters, risk officers, and campus action officers complete the handover.
L05One Square Meter ConstructionMixed-age Construction · Three-State Stress Test

How to make each component come from experiments while also accommodating real usage actions?

  1. Ground LayoutMark one square meter, pathways, wheelchair turning and resting postures.
  2. Skeleton and CounterweightComplete connection, rounding, wobble test, and adult structure review.
  3. Four Mechanisms InstallationAttach corresponding experiment and user requirements to each module.
  4. Sensing and ExperienceSynchronize baseline/prototype nodes to let people of different heights and mobility try them.
  5. Three-State Stress TestDrill high-temperature unfolding, strong wind storage, post-rain inspection, and responsibility handover.
L06Cool Island OpeningPublic Trial Sitting · Hot Clinic · Maintenance Handover

How does the work enter a real day and continue to be used and cared for?

  1. Open PreparationWeather guidelines, diversion, drinking water, sensors, and visitor tasks in place.
  2. Hot Clinic Trial SittingBaseline experience, cool island tasks, thermal passport, and synchronized data.
  3. 30-second Feynman explanationExplain coolness using shadows, cloth, water, wind paths, or a curve.
  4. Design Bureau DefensePresent predictions, surprises, evidence, version changes, and next steps.
  5. Maintenance and SkillCleaning, hydration, storage, data updates, and personal methods for handover.

Four experimental stations, four visible mechanisms

Students place bets first, then change one variable. The experimental results are reproduced by another age group before entering the shared cool island.

Make the sun avoid the body

Change height, direction, porosity, and projection to draw the path of shadow movement over time.

Make the material absorb less heat

Compare color, surface, and texture while observing glare and the experience of surrounding people.

Let water carry away heat

Record water volume, humidity, duration, and touch to find suitable weather conditions.

Let air find its way

Use ribbons to see the wind, change openings, guide plates, and model density.

The probabilities and paths in this area are for teaching demonstration, not the current forecast; seasonal data should be read according to the release date.

Weather and AI Laboratory

Climate background changes probabilities, weather forecasts provide lead time, campus actions respond to specific locations and people.

El Niño Evidence Card

WMO · 2026-06-02

The probability of forming El Niño from June to August 2026 is 80%. Students write the probability, intensity range, forecast period, and update time on the same card.

Graphs are used for classroom chart reading exercises; official courses connect to the latest WMO/NOAA data.

Tracking a Typhoon

Path = forecast · dot = time
Three candidate typhoon paths Three models start from the Northwest Pacific and provide different predictions for the typhoon's future path.
Pangu exampleNumerical modelEnsemble members

The team first draws their sixth point, then compares model divergences, lead time, and the official best path. After real conditions arrive, calculate errors.

Markov Weather Station

K10–12 · Transparent probability baseline
P(i→j) · Each row sums to 1
Current\NextSunny and hotCloudyThunderstormStrong wind
Sunny and hot0.500.300.150.05
Cloudy0.250.400.250.10
Thunderstorm0.200.350.350.10
Strong wind0.250.250.200.30

In class, first calculate a row manually using paper cards, then run the local Python program to process the complete state sequence.

Waiting for Actual
12-Step Weather Simulation

Cloudy → Cloudy → Thunderstorm → Cloudy → Sunny and hot

Interactive buttons display preset procedures and sample results. The complete course additionally uses historical data, student submissions, and a grader; model training is not run on this page.

Missing Day 31 Weather Record

Two teams face the same real weather record, the same time limit, and the same scoring rules. Day 31 is sealed in a time capsule and will be revealed only after predictions are frozen.

Public Demonstration LocationGuangzhou Grid Points
Historical Daily Records2,192
Training / Validation / Blind Test2,129 / 30 / 30
Shared TaskPredict · Explain · Act

RED TEAM · See Every Step

Make a Transparent Prediction Ruler from Yesterday's Weather

Callable

Persistence Baseline, Monthly Climate State, Markov Chain, Linear Regression, 5 AI Help Tickets

Action Roles

Data Detective, Rule Builder, Error Auditor, Campus Action Translator

Leave Evidence

State Transition Table, Training Log, Daily Predictions, Error Explanation, and a Model Revision

H01
Data Wilderness

Turn Scattered Weather into Traceable Clues

K1–3 Use Body Weather Symbols, K4–6 Perform Data Quality Check, K7–9 Identify Features, K10–12 Establish Time Slices and Data Contracts.

H02
Dual-Track Timed Challenge

Freeze Day 31 Predictions on Both Tracks Simultaneously

Teams Save Inputs, Prompts, Code, Versions, and Judgments, So Every Result Can Be Reproduced by Peers.

H03
Purple Team Convergence

Translate accuracy into campus cool-island action

Swap a method, rerun the model, and release action recommendations with location, time period, and update time to real users.

TIME CAPSULE · Classroom demonstration

Give the judgment first, then let the live data in

Freeze the prediction after choosing a route. At this moment, the team uses 30 seconds to explain: what confusion the prediction solves, what image comes to mind, and what to check when the live data arrives.

The time capsule waits for the team to submit predictions and reasons.

Early Warning Relay Station

The outer rain and wind of a typhoon are approaching the campus. Players choose one option each from observation, prediction, risk, and action to form an executable campus message.

Scenario 03 · Before school dismissal on Friday

Multiple paths still have differences. Strong winds and short-term heavy rainfall may occur on campus in the afternoon; low-lying East Gate, outdoor display boards, and shaded areas need to be arranged in advance.

0 / 4

Complete one selection at each of the four stations on the right; the system will generate a campus action message.

First Station · Choose Observation

Second Station · Read Prediction

Third Station · Identify Risks

Fourth Station · Issue Action

One square meter construction workbench

Each component can correspond to an experiment, a usage action, and a maintenance responsibility.

WorkstationStudent actionsRelease EvidenceCross-age handover
LayoutMark 1 m², pathways, and posture for stayingDimensions and accessibility photosK1–3 trial walk, K10–12 boundary revision
SkeletonConnection, counterweight, rounded corners, and shake testStructural inspection and adult reviewK4–6 pipe count, K7–9 pipe connections
MechanismInstall sunshades, ventilation, reflection, and evaporation candidatesModule corresponding experiment sourceExperimenter explains conditions to the constructor
SensingSynchronize baseline points and prototype pointsTime, location, height, and CSVLow-age screen reading, higher-grade metadata checking
Three StatesUnfold in high temperature, pack in strong wind, check after rainOperation time and responsible personIcon testing and logistics sign-off

Teacher's Lesson Platform

Checked states will be saved in the current browser. The checklist covers real assignments, weather safety, data, materials, and public display.