Hobbyists and Researchers
Troubleshooting and FAQs
Understanding the Solcast’s Estimated Actuals and Forecast solar data
Let us walk you through these forecast data terminologies that can help you better analyse your solar irradiance data.
There is certainly a lot of specialised terminology in solar forecasting, and more broadly the fields of solar radiation and meteorology are teeming with acronyms than can make things fairly hard to understand. And when it comes to the hard work of deploying, operating and managing solar technologies in the grid, we'd hate for our own terminology to be too complicated!
So let's make things nice and clear for you.
Estimated Actuals are our 'live' and 'historical' data.
Live data refers to what is happening 'now' (real-time), reaching back through the past 7 days.
Historical data refers to all of our data beyond the previous 7 days, reaching back many years.
Why we call these data 'estimated actuals'
Many folks across the industry will refer to this type of data as 'measurements', but we are far from comfortable with this practice. Weather satellites cannot measure or directly sense the solar radiation which is arriving at the Earth's surface. It can only see the clouds, or see the heat coming back from the planet towards the satellite. So, any product which is providing historical solar radiation values at the Earth's surface over a given geographic region is actually giving you an 'estimate' of the 'actual' solar radiation at that location.
To produce our Estimated Actuals, Solcast's modelling team uses images of the clouds taken by weather satellites, along with a series of modelling steps in order to produce estimates of the historically available solar radiation at any given location. We're updating these values globally, with each new satellite scan, every 5-15 minutes through the use of our rapid update forecasting system. To reiterate, the Live or Historical API endpoints of our data products do not make any predictions in the future. It only uses ‘valid’ satellite scan, meaning the images of the clouds as they were at any given time.
Above: A validation of Solcast's Estimated Actuals as compared to a pyranometer at a solar farm. Here we are comparing the Estimated Actuals of Global Horizontal Irradiance (GHI) to what was recorded by an actual measurement device at that location.
Forecasts are predictions of what may occur. They contain uncertainty due to the challenging and somewhat chaotic nature of the atmosphere, and in the case of Solcast's products, are probabilistic to capture how much uncertainty is present for any given forecast. You'll notice in our forecast products that there is a P10, P90 and P50 forecast to capture the 10% and 90% probability around the median forecast.
Above: An example of one of Solcast's demo charts (available to all of our users). To the right of the vertical dotted line (current time) are forecasts of the total power output from the all the Solar PV systems installed in South Australia. Shading around the mid-line are the probability bounds on the forecast.
Solcast's forecasts reach out to 7 days ahead, with the next four hours from any given time being produce by satellite-based forecasting methods. With each new satellite scan (every 5-15 minutes), our forecasts are re-issued using the latest cloud imagery. With five weather satellites around the world, and an ensemble of model predictions made at each update cycle, we produce over 6 billion forecasts every hour!
Forecasts are not used to produce historical data, and are generally discarded after they become outdated. However, we do persist the 1-hour ahead forecast in our API, so that you can check out our accuracy at any location, whenever you would like.
Our data in point and time
The Solcast historical and live data APIs always provides the most up to date information that we have. As well using the satellite images, our irradiance estimates use inputs about meteorological conditions and aerosols from numerical weather prediction sources which are frequently updated. This can lead to revised values as new inputs become available. Any changes are generally only minor as the data from satellite imagery is the primary input, and this does not change. As the most recent API call will be using the most up-to-date information, you should treat this as the best data.
How to handle the fact that estimated actuals change as more updated data comes through?
If you are building a historical record from live data record, we would suggest making one API call for the previous day, as by then all satellite imagery, the primary driver of irradiance variability, will have safely been assimilated. If you need data earlier but want to ensure fewest number of changes we suggest making the request as late as possible.