Weather forecasts are built from multiple forecast models, and those models can produce somewhat different results. Some models provide a single deterministic forecast, while ensemble systems generate multiple possible outcomes.
When the contributing forecasts agree closely, confidence in the forecast is generally higher. When they diverge, there is greater uncertainty about the expected outcome.
The API provides two measures to help describe this forecast consensus:
spread: element-specific disagreement between the contributing forecast modelsconfidence: a 0–100 score summarizing overall model consensus
High model agreement does not guarantee that a forecast will be correct, but stronger agreement is typically associated with a more stable and reliable forecast.
Why multiple forecast models matter
A weather forecast is rarely based on one source alone. The forecast system combines information from multiple sources, including global, regional, deterministic, and ensemble forecast models.
Ensemble systems are particularly useful for understanding uncertainty because they generate multiple forecasts using variations in initial conditions or model configuration. Together with differences between individual forecast models, this provides information about the range of plausible forecast outcomes.
When these forecasts converge on a similar solution, confidence in the forecast generally increases. When they diverge, the forecast becomes less certain.
This model agreement is the basis of the forecast consensus information provided by spread and confidence.
What is spread?
The spread object measures disagreement between forecast models for individual weather elements.
Each spread value is calculated as the weighted standard deviation of the contributing model forecasts around the weighted blended forecast.
Spread is currently provided for:
spread.temp— weighted standard deviation of temperature, in the configured temperature unitsspread.windspeed— weighted standard deviation of wind speed, in the configured wind-speed unitsspread.sealevelpressure— weighted standard deviation of sea-level pressure, in the configured pressure units
For example:
"spread": {
"temp": 1.2,
"windspeed": 2.8,
"sealevelpressure": 1.5,
"confidence": 86
}
A lower spread means that the contributing models are clustered more closely around the blended forecast. A higher spread means that the models disagree more substantially.
For example, if the API is returning temperature in degrees Celsius:
"temp": 1.4
means that the contributing model temperature forecasts differ from the weighted blend by approximately 1.4°C in weighted standard-deviation terms.
It does not mean that the temperature is 1.4% uncertain or that the forecast has a reliability score of 1.4.
Spread is an element-specific statistical measure rather than a probability, percentile, range, or normalized uncertainty score.
What is confidence?
spread.confidence is a separate 0–100 score that summarizes the overall level of forecast consensus.
Unlike the individual spread values, confidence is not expressed in weather units. It combines information including:
- element-level model spread
- the number of contributing forecast models
- forecast horizon
Higher values indicate stronger overall model consensus, while lower values indicate greater disagreement.
A general interpretation is:
- 80–100: strong forecast consensus
- 60–79: moderate forecast consensus
- 40–59: increased uncertainty
- below 40: weak forecast consensus
For example:
"spread": {
"confidence": 92
}
A value of 92 indicates very strong agreement among the contributing forecasts.
High confidence is typically associated with a more stable and reliable forecast, although model agreement does not guarantee that the forecast will ultimately be correct.
Spread vs. confidence
The simplest way to think about the two fields is:
spreadtells you where the models disagree.confidencetells you how strong the overall forecast consensus is.
For example, a forecast may have a high overall confidence score while still showing noticeably greater spread in wind speed than in temperature.
This lets applications use confidence as a quick summary while examining individual spread values when a particular weather element is important.
How to retrieve spread and confidence
To request forecast consensus information, include spread in the elements parameter.
For example:
/timeline?
key=YOUR_API_KEY&
location=NewYork&
elements=temp,windspeed,sealevelpressure,spread
The response includes the spread object within the applicable forecast records:
{
"days": [
{
"hours": [
{
"datetime": "2026-09-08T12:00:00",
"temp": 74,
"windspeed": 9,
"sealevelpressure": 1016,
"spread": {
"temp": 1.4,
"windspeed": 2.1,
"sealevelpressure": 1.8,
"confidence": 85
}
}
]
}
]
}
The primary weather values represent the blended forecast. The corresponding spread values show how tightly the contributing models agree around that forecast, while confidence summarizes the overall level of consensus.
Availability and compatibility
Spread information is currently available only in JSON-formatted API responses.
The following element-specific spread values are currently available:
tempwindspeedsealevelpressure
Additional spread elements may be added in the future. Applications should therefore treat the contents of the spread object as extensible and should not assume that it will always contain only these fields.
For example, clients parsing the response should safely ignore or handle additional spread properties that may appear in future API responses.
Using spread and confidence
Assess overall forecast stability
Use spread.confidence for a quick indication of how strongly the contributing forecasts agree.
A high confidence score generally indicates a more stable forecast. A low score indicates greater disagreement and therefore a greater possibility that the forecast may change as new model data becomes available.
Examine individual weather elements
Overall confidence can hide important differences between variables.
For example, models may agree closely on temperature while showing considerably greater variation in wind speed.
Element-level spread lets you identify where that disagreement is occurring.
Monitor forecasts over time
Spread and confidence are also useful when comparing successive forecasts.
If spread decreases and confidence increases, the models are generally converging toward a common solution. Increasing spread or falling confidence indicates that the range of predicted outcomes is becoming wider.
Build uncertainty into applications
Forecast consensus can be useful in:
- weather dashboards
- operational planning systems
- alerting systems
- forecast monitoring
- automated decision-support applications
For example, an application might highlight forecasts with low confidence or flag periods where the spread of a particularly important element exceeds a chosen threshold.
Summary
Forecast consensus provides additional information about how strongly the contributing forecast models support the blended forecast.
The API provides:
spread— the weighted standard deviation of individual weather elements across the contributing forecastsconfidence— an overall 0–100 model-consensus score derived from spread, model availability, and forecast horizon
Lower spread and higher confidence generally indicate a stronger and more stable forecast. Higher spread and lower confidence indicate greater disagreement and a wider range of plausible forecast outcomes.
These values help applications communicate not only what the forecast predicts, but also how strongly the contributing forecast models agree with that prediction.