Cloud cover prediction: successful satellite launch

What if we put a camera in front of us to observe the part of the Earth that the satellite will fly over in 45 seconds? What if we use this data in real time to map cloud cover? Couldn't an algorithm be "trained" to predict the quality of high-resolution images that might be taken before they are even taken?  This is the innovative idea proposed by ONERA in collaboration with the company Loft Orbital.

This company offers a very original approach to satellite platform “sharing”, by taking charge of the integration of its clients' payloads and handing over the satellite “keys” to them when they need them. This concept enabled Onera and Loft Orbital to define and supply the cloud sniffer payload within a few weeks in mid-2020; it also enabled Loft Orbital to integrate it on its recently launched Yam3 satellite.

Although the technologies used are not new (using on-board optical sensors, developing prediction algorithms), it would not have been possible in a "traditional" approach to space to set up such an ambitious and innovative experiment, at such low costs and within such short timeframes. Thanks to these reduced costs and timeframes, the partners can take the risks inherent in innovation. Today, there are no databases of such image pairs in public literature that can be used to develop the algorithms and test this concept.

How will it work?

The images taken by the satellite will be used together: the images from the main payload in nadir view and those from the secondary payload (less spatially resolved but with a wider field of view and in forward view).
Objective: to evaluate the cloud cover on board a few seconds ahead, in order to trigger or not trigger a shot from the main camera, and/or to re-channel it autonomously to another request in the vicinity with more favourable weather.
The challenge: gaining 10% of usable images is like adding 10% more satellites to a constellation...