eduKateSG · Why Science?
Understand tropical weather and climate across scales—connecting observations, convection, monsoons, numerical models, ensembles, seasonal outlooks, projections, sea level, extremes, uncertainty and resilience
A fact-checked guide to CCRS, Meteorological Service Singapore, tropical weather science, climate projections, high-resolution modelling and climate learning.
The Centre for Climate Research Singapore states that it was launched under Meteorological Service Singapore in March 2013 to become a regional centre of excellence in tropical climate and weather research.
Its official research areas cover climate projections, seasonal outlook, weather prediction and fundamental understanding of regional weather and climate processes, supported by numerical models and high-performance computing.
Did you know? Singapore’s thunderstorms can grow and change faster than many coarse global model grid cells can describe directly. Scientists therefore combine observations, physical equations, high-resolution models and parameterisations to study them.
CCRS and MSS control current research, forecasts and released projections; public agencies and communities control planning decisions. This guide explains evidence without replacing official advisories. Source status was checked 11 November 2026.
Chapter index
Chapters 1–9: weather and climate, observations, quality control, energy and water, convection, monsoons, oceans, land, numerical models
- 1. Separate weather from climate
- 2. Observe atmosphere, ocean and land
- 3. Quality-control environmental data
- 4. Follow the tropical energy and water cycles
- 5. Understand deep convection
- 6. Study monsoon variability
- 7. Connect oceans to regional climate
- 8. Represent land and cities
- 9. Build numerical weather models
Chapters 10–17: resolution, parameterisation, ensembles, seasonal outlooks, projections, extremes, sea level, communication and careers
Weather describes particular atmospheric states and events, while climate describes longer-term distributions, variability and change. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Match every claim to an explicit spatial and temporal scale. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
One hot day does not establish a climate trend, and a warming climate does not determine one exact day. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Separate weather from climate” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Sort ten statements into weather, seasonal climate and long-term climate categories. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 1 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
Tropical conditions emerge from exchanges of heat, water and momentum among air, ocean and land. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Record variable, instrument, location, height or depth, unit and sampling interval. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
No single station or satellite channel observes the complete coupled system. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Observe atmosphere, ocean and land” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Build a measurement table for temperature, humidity, rain, wind and sea level. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 2 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
Sensors drift, stations move, gaps occur and observing practices change. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Use flags, calibration records, neighbour comparisons and metadata before calculating trends. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
Cleaning should not erase real extremes simply because they look unusual. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Quality-control environmental data” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Inspect a fictional rainfall series containing one gap and one instrument shift. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 3 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
CHAPTER 4 OF 17 · Observe the tropical system
4. Follow the tropical energy and water cycles
Solar heating, evaporation, condensation, radiation and circulation connect energy with rainfall. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Draw conserved quantities and identify sources, sinks and phase changes. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
A local downpour cannot be explained from local humidity alone without circulation and stability. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Follow the tropical energy and water cycles” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Create a labelled water-and-energy pathway for an afternoon storm. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 4 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
| Evidence move | What to record | What it cannot prove alone |
|---|---|---|
| Observation question | Population, variable, unit and time window | A definition does not establish a result |
| Comparison | Baseline, comparator and uncertainty | Association alone does not establish cause |
| Decision boundary | Owner, trigger, review and escalation | A model does not replace accountable judgement |
Convection can organise rapidly when warm moist air rises, condenses and interacts with wind and stability. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Use soundings, radar, satellite and surface observations to test storm development. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
A necessary ingredient such as moisture may not be sufficient without lift and instability. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Understand deep convection” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Compare two fictional soundings and identify which ingredients differ. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 5 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
Monsoons involve seasonal shifts in winds and rainfall shaped by regional land-ocean contrast and larger circulation. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Analyse composites, onset definitions and year-to-year variation rather than one calendar date. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
A monsoon season contains wet and dry spells and does not guarantee uniform rain. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Study monsoon variability” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Build a monthly wind-rainfall chart and annotate uncertainty in onset. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 6 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
Sea-surface temperature and ocean dynamics influence moisture, circulation and modes of variability. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Compare observed anomalies with lagged atmospheric responses and model experiments. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
Correlation with an ocean index does not prove identical impacts every year. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Connect oceans to regional climate” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Plot a fictional ocean index beside rainfall and identify plausible lags. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 7 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
Surface materials, vegetation, moisture and urban form affect heat storage, evaporation and airflow. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Measure land cover and surface fluxes at scales that match the question. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
An urban average can hide neighbourhood differences and human exposure. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Represent land and cities” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Design a safe temperature transect using teacher-provided data. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 8 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
| Evidence move | What to record | What it cannot prove alone |
|---|---|---|
| Model question | Population, variable, unit and time window | A definition does not establish a result |
| Comparison | Baseline, comparator and uncertainty | Association alone does not establish cause |
| Decision boundary | Owner, trigger, review and escalation | A model does not replace accountable judgement |
CHAPTER 9 OF 17 · Interpret forecasts and projections
9. Build numerical weather models
Numerical models solve discretised physical equations from an analysed atmospheric starting state. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Document grid, vertical levels, time step, boundary conditions and data assimilation. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
Small initial errors can grow, and equations must be approximated on finite grids. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Build numerical weather models” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Sketch the path from observations to analysis, forecast and verification. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 9 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
CHAPTER 10 OF 17 · Interpret forecasts and projections
10. Understand resolution and parameterisation
Grid resolution controls which processes are resolved explicitly and which need simplified representation. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Test results across resolutions and parameter choices while tracking computational cost. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
Finer grids do not automatically remove bias or uncertainty. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Understand resolution and parameterisation” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Compare how one island or storm appears on coarse and fine grids. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 10 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
CHAPTER 11 OF 17 · Interpret forecasts and projections
11. Use ensembles to sample uncertainty
Ensembles run multiple plausible initial states or model configurations to explore forecast spread. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Evaluate reliability, sharpness and event probabilities over many cases. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
A 30% probability does not mean the event occurs on 30% of the map or day. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Use ensembles to sample uncertainty” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Translate ensemble members into a probability statement with a clear event threshold. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 11 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
CHAPTER 12 OF 17 · Interpret forecasts and projections
12. Read seasonal outlooks probabilistically
Seasonal outlooks compare chances of broad categories over weeks or months with climatology. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
State baseline period, category thresholds, lead time and model skill. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
An outlook cannot schedule individual storms or dry days. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Read seasonal outlooks probabilistically” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Write two correct and two incorrect interpretations of a tercile outlook. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 12 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
| Evidence move | What to record | What it cannot prove alone |
|---|---|---|
| Projection question | Population, variable, unit and time window | A definition does not establish a result |
| Comparison | Baseline, comparator and uncertainty | Association alone does not establish cause |
| Decision boundary | Owner, trigger, review and escalation | A model does not replace accountable judgement |
Climate projections combine scenarios with models to explore long-term responses and ranges. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Compare ensembles, baselines and downscaling methods, and report scenario dependence. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
A projection is conditional evidence rather than a promise that society follows one pathway. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Construct climate projections” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Create a projection sentence that includes variable, period, baseline, scenario and range. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 13 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
Extreme rain, heat and wind require threshold definitions and careful statistics because observations are rare. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Use return levels, duration, compound-event analysis and uncertainty intervals. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
A return period is an average rate, not a timetable that resets after an event. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Evaluate extremes” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Explain a one-in-100-year event without implying it can occur only once per century. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 14 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
Local sea level combines global ocean change, regional circulation, ice contributions, land motion, tides and storm effects. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Separate mean change from episodic extremes and preserve vertical reference systems. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
Global mean sea-level rise does not translate identically to every coast. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Study sea-level change” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Draw components that contribute to a future coastal water level. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 15 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
CHAPTER 16 OF 17 · Translate science responsibly
16. Translate uncertainty for resilience
Climate services connect scientific evidence with decisions about infrastructure, health, water, biodiversity and emergency planning. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Co-design thresholds, update cycles and decision-relevant ranges with users. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
Uncertainty is not ignorance and should not be hidden behind one precise number. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Translate uncertainty for resilience” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Write a two-layer message: the robust finding first, then the range and conditions. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 16 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
The receipt connects variable, observation, model, baseline, scenario, scale, uncertainty and decision. For the Centre for Climate Research Singapore, the scientific task is to describe a tropical atmosphere that changes across minutes, seasons, decades and centuries. Weather and climate are related but not interchangeable: weather concerns particular states and events, while climate concerns distributions, patterns and long-term conditions. Keeping the timescale explicit prevents many common misunderstandings.
Archive datasets, code, model versions, evaluation and communication choices. Climate researchers combine observations, physical equations, numerical models, ensembles and statistical evaluation. A useful record names the station or grid, variable, unit, baseline, model resolution, scenario, lead time and uncertainty range. Comparison with withheld observations and other models helps reveal systematic bias and shows where a projection is robust or where confidence should remain limited.
A receipt cannot stop change but lets future reviewers understand why a conclusion was reasonable. Maintain the forecast boundary. A seasonal outlook is probabilistic rather than a daily timetable, and a climate projection is conditional on scenarios and model assumptions rather than a prediction of one exact afternoon. Local rainfall and convective storms are especially sensitive to scale, so high resolution can add detail while still leaving uncertainty that planners must manage.
Investigate “Keep a climate evidence receipt” with an open, non-sensitive dataset or a teacher-provided table. Plot one variable, annotate units and missing values, compare two time windows and write a claim with an uncertainty sentence. Build a one-page receipt for a rainfall or heat claim and list update triggers. The exercise should distinguish what the graph displays from why the pattern occurred and from what might happen under a different future scenario.
Did you know? A model can conserve physical quantities and still have regional bias because many processes occur below the grid scale. Chapter 17 connects physics, mathematics, geography, computing and public communication. Pathways can include atmospheric science, oceanography, environmental engineering, data science, geography, policy and resilience planning, subject to current programme and employer requirements.
| Evidence move | What to record | What it cannot prove alone |
|---|---|---|
| Climate question | Population, variable, unit and time window | A definition does not establish a result |
| Comparison | Baseline, comparator and uncertainty | Association alone does not establish cause |
| Decision boundary | Owner, trigger, review and escalation | A model does not replace accountable judgement |
Contents · Previous chapter · Continue in the Science Learning Hub
Primary sources and verification date
Source status for Why Science? | Centre for Climate Research Singapore was checked 11 November 2026. Research priorities, programmes, recruitment, facilities and participation routes can change; use the linked official owner for the current edition.
- CCRS Who We Are — Official mission, history and tropical research scope.
- CCRS Research Areas — Official climate projections, seasonal outlook, weather prediction and process research.
- Climate and Weather Research Programme Office — Current national research programmes and September 2025 update.
- CCRS Climate Projections — Official projection work and Third National Climate Change Study context.
Continue the evidence journey
Centre for Climate Research Singapore science becomes useful when a learner can match a claim to its scale, preserve model conditions and communicate uncertainty as part of responsible resilience planning. Families considering science tuition or a science tutor can use the same evidence questions: what will the learner practise, what feedback will they receive, and how will progress be shown?