Argus SunWatch

ARGUS SUNWATCH · HELP

Reading charts and data

Chart values, forecast probabilities, observation timestamps and model evaluation metrics.

Forecasts Live observations

Getting started

  1. Check current conditions. Open Live and check the observation timestamp to see how recent the available data is.
  2. Choose a forecast. Browse Products for the quantity you want to follow. Each product has its own charts, units, and API links.
  3. Read the uncertainty. Inspect the median and quantiles where available, or the probability of crossing a specified threshold.
  4. Check past performance. Open Metrics for that product and compare results at the lead time you care about.

Reading forecasts

A forecast describes possible future conditions. An observation describes recorded conditions. Keep the two separate when comparing charts.

Median & quantiles

The Median is the 50th percentile (q50), the model’s central estimate. Low and High are the 10th and 90th percentiles (q10 and q90), available in the chart tooltip.

The q10–q90 range represents the model’s central 80% prediction interval. It is not a guaranteed minimum or maximum; actual coverage depends on model calibration.

Threshold probabilities

Each heatmap row names an event, such as V ≥ 500 km/s. Each cell gives the model’s estimated probability for that event at a forecast lead time.

A value of 70% means a 70% estimated chance of meeting that threshold. It does not mean a 70% increase in speed or a 70% chance of infrastructure damage.

Issue time
The reference time from which a forecast is made.
Lead time
How far ahead a prediction looks, in hours after issue time. Lead hour 24 refers to one day ahead.
Valid time
The time a prediction applies to. The API exposes this as valid_time for each prediction.
Time zones
Live observations are labelled in UTC. Forecast charts also use UTC, with dates and times taken from the forecast. Check the timestamp and time zone when comparing data.

Forecast products

Products differ in the quantities and forecast types they provide. Some variables may not yet have a forecast available.

Public API

Solar Wind Speed

Quantile forecast and probabilities of high-speed solar wind.

Solar Wind Speed: km/s

Private API

Heliospheric Magnetic Field

Threshold probabilities for total field strength and southward Bz.

Total HMF: nT · Southward Bz: nT

Private API

Solar Radiation Indices

Joint forecast of the F10.7, S10, M10 and Y10 solar indices.

F10.7: sfu · S10: index · M10: index · Y10: index

Public API

Geomagnetic Activity

Kp threshold probabilities and the related Ap quantile forecast.

Kp Index: index · Ap Index: index

Public API

Dst Index

Quantile forecast of the storm-time disturbance index.

Dst Index: nT

Live data & freshness

The Live page shows solar wind measurements at L1, three-hour Kp and hourly Dst. It checks for updates every minute. All observation times are UTC.

Freshness
Solar wind is marked delayed after ten minutes. Kp and Dst use time since the observation interval ended: four hours for Kp and two for Dst. A recent page refresh can still return old data.
History resolution
Up to 24 hours uses minute solar wind samples; 3–7 days uses five-minute means; 30 days uses hourly means. Aggregate tooltips include min/max and coverage. Kp/Dst keep their original intervals.
Missing values
A dash means unavailable, not zero. Native observations are not filled. Missing intervals and provider-flagged values leave gaps in the charts.
Coverage
The percentage counts usable measurements. For aggregates it covers calculated windows only; uncalculated windows have unknown coverage. Values awaiting recalculation remain visible with a notice.
Collection status
Expand “Data collection” for source delays, request errors and overdue collection. If a page refresh fails, previously loaded data remain visible while it retries.
Additional hourly indices
This separate section uses normalized data, which may be filled. S10, M10 and Y10 are provisional daily estimates calibrated from GOES data.

Understanding model performance

The Metrics pages show evaluation results by forecast lead hour. Compare the same variable, threshold, and lead time: a single score does not describe every forecast situation.

Brier score
Measures the error of event probabilities against observed outcomes. Lower is better; zero is perfect.
ROC AUC
Measures how well the model ranks events above non-events. Higher is better; 0.5 corresponds to random ranking and 1 to perfect ranking. It does not measure probability calibration.
Precision chart
This chart uses average precision, which summarizes the precision–recall trade-off across thresholds. Higher is better; interpret it alongside how common the event is.
Reliability
Compares predicted probabilities with observed event frequencies. For a well-calibrated model, events assigned about 70% probability should occur about 70% of the time across many comparable predictions.
Continuous metrics
For quantile forecasts, the available metrics are shown separately for each variable. MAE and RMSE, when provided, measure prediction error in the variable’s units; lower values are better.

Historical evaluation describes performance on past observations. It does not guarantee the accuracy of an individual forecast.

API & support

Use the API documentation to explore endpoints and response schemas. Product pages link directly to their forecast and metrics JSON responses.

Current observations Solar wind history Kp/Dst history Solar wind forecast Solar wind metrics

API probabilities are numbers from 0 to 1; heatmaps display them as percentages. Forecast responses include the available variables and prediction times, so check these before using a value.

Questions or unexpected results?

Email krivenko.a.b@gmail.com or visit the GitHub repository. Include the page or endpoint, the relevant timestamp and time zone, and what you expected to see.