The rapid increase of in-orbit payloads in recent years, with mega-constellation exploitation such as Starlink, has made collision-avoidance analyses increasingly critical.

Conjunction-analyses, typically performed when the spacecraft is in the station-keeping position or during orbit acquisition, are only rarely performed pre-launch to assess the first hours after separation. This early phase is especially delicate, since the spacecraft typically cannot perform manoeuvres during the deployment of appendages, the initial  attitude acquisition,  the configuration of the propulsion and AOCS subsystems. This phase could last multiple hours, while the pre-launch screening by launch service providers usually covers the launcher ascent trajectory and few hours after separation (except for manned spacecrafts).

The growing congestion increases the need for accurate risk assessments, while improved debris tracking and reduced launcher dispersions enable better risk identification and management. That’s why EUMETSAT decided to start performing conjunction assessments pre-launch, for longer trajectories after separation from launcher, and to introduce this in the final procedure leading to selecting the final lift-off times. For this, EUMETSAT relied on EUSST for the provision of screening results, working together to establish a precursor service for pre-launch post-separation screenings.

The first launches to undergo this process were MTG-S1 (GEO), launched with Falcon 9 on 1st July 2025, and MetOp-SG A1 (LEO), launched with Ariane 6 on 13th August 2025. Both scenarios will be presented, involving different challenges:

• MTG-S1, on a Super-Synchronous-Transfer-Orbit, crossed Starlink altitude (~450km) shortly before separation but, due to its highly eccentric trajectory (300km/65000km perigee/apogee height), exited the LEO protected region quickly and re-entered it one orbit later, without being manoeuvrable in between.  Moreover, the evolution of the initial uncertainties along a high-elliptical orbit are difficult to model, as linear covariance propagation is not applicable and the associated ellipsoid starts to bend. A relatively wide launch-window (~150minutes) existed for launch day: EUMETSAT was to communicate the closed portions of the window, due to high-risk events.
• MetOp-SG A1 was injected at ~800km and did not cross Starlink altitude but it remained inside the LEO protected region for the entire period in which the spacecraft was not manoeuvrable. There was no launch window but only a single lift-off time per day. A small lift-off shift was however agreed to be implemented, in case of an anticipated (pre-launch) high risk post-separation.

With the support of EUSST, EUMETSAT assessed the safety of the post-separation trajectories for both spacecrafts, making use of newly developed tools for post-processing, implemented in Python/MATLAB and relying on Orekit as the main flight dynamics library.
For the nominal launch day and back-up dates, EUSST screened the ephemerides provided by EUMETSAT, using a wide screening volume. These involved multiple ephemeris per day, representing multiple lift-off times. The resulting conjunction data were then post-processed : the main idea was to classify each event according to the potential expected consequence of a collision, based on secondary object’s size (small/medium/big) and its status (active/inactive). For instance, a conjunction involving a large active spacecraft entails far more severe consequences (both in terms of orbital-environment degradation and in the implications for interactions with another operator) than one involving a small inactive object. This lead to the definition of different levels of acceptable risk based on the categories of the classification.

A major challenge was designing a methodology that could also cope with launcher dispersions and their rapid growth over time, particularly along-track.

For MTG-S1, the primary driver was geometric, using the normal miss-distance to the estimated orbital trajectory, considering the curvature of the covariance; this was modelled, together with along-track time evolution, based on propagates state vectors, from the launcher Monte Carlo dispersion analysis at separation.

For MetOp-SG A1, the decision drivers were the total miss-distance, the Probability-of-Collision(PoC) / Depth-of-Intrusion(DoI) and the cumulated PoC, together with the time of close approach (TCA) with respect to injection, as PoC and DoI are not representative when covariance ellipsoids are too large (dilution effect).

In parallel, EUMETSAT developed another tool to perform independent screening against a TLE catalogue (from SpaceTrack). These results were analysed and compared with those provided by EUSST, offering an additional layer of validation.

The actual operational experiences for the 2 launches, with the relevant results, will be detailed in this paper.