Quick answer: Starlink conjunction assessment does not ask only how close two objects will pass. It combines predicted miss distance with position uncertainty—often represented by covariance—to estimate collision risk and decide whether an avoidance maneuver is justified. A conjunction is a warning, not proof of an imminent collision.
Related context: Read how Starlink orbital shells are organized and why Starlink satellites are designed to reenter.
What Is a Conjunction?
A conjunction is a predicted close approach between two objects in space. The objects could be two active satellites. One could be an active satellite and the other could be debris. One could also be an old rocket stage or a dead satellite.
The key word is “predicted.” A conjunction is not the same as a collision. It means a model says the objects may pass near each other at a future time.
Think about two cars on different roads. If a map app predicts they will both reach the same intersection at noon, that is a possible conflict. But the prediction may be wrong. One car may slow down. The map may have old traffic data. The roads may not truly cross at the same height. A space conjunction has the same kind of prediction problem, but in three dimensions and at orbital speed.
Satellites move very fast. A small error in the predicted position can matter. If an object is a little early, late, higher, lower, left, or right, the close approach may become safer or more risky.
This is why conjunction assessment is about more than distance. Distance matters, but uncertainty matters too.
The Screening Problem
Screening means checking many possible cases and finding the few that deserve attention. For a large constellation, screening is not a small task.
Starlink satellites orbit Earth in large numbers. They must share space with active spacecraft and tracked objects that are no longer controlled. Public space-tracking systems can estimate where many objects are going. Those estimates can then be compared with the predicted paths of Starlink satellites.
The first screen asks a simple question: could any object pass close enough to care about during a future time window?
Many results from this first screen are not emergencies. They are candidates for more study. A candidate is a possible case, not a final answer. This is like sorting mail. Most letters are normal. A few need a closer look. A smaller number may need action.
Why Distance Alone Is Not Enough
It may sound easy to rank conjunctions by closest distance. If two objects are predicted to pass within a small distance, that must be bad. If they are predicted to pass far apart, that must be safe. Real life is not that simple.
The predicted distance is only the center of the guess. The real objects may not be exactly at those predicted points.
Imagine throwing two paper airplanes in a gym. If you know their exact paths, you can say whether they will touch. But if you only know each path roughly, the problem changes. One plane may be inside a wide “maybe zone.” The other may also be inside a “maybe zone.” If those maybe zones overlap, the risk can be more important than the center points alone.
In orbit, the “maybe zone” comes from measurement error, modeling limits, space weather, drag, and other sources. Engineers must ask, “How uncertain is this prediction?”
A close approach with very small uncertainty can be easier to judge. A farther approach with large uncertainty may still need attention. The details depend on the data quality and the type of objects involved.
Uncertainty in Orbit Prediction
Uncertainty means we do not know the exact answer. In space tracking, uncertainty is normal. It does not mean the system is careless. It means the world is measured and modeled with limits.
Ground sensors and other tracking sources observe objects in orbit. These observations are used to estimate an orbit. An orbit is the path an object follows around Earth. But each observation has limits. A sensor may measure angle better than distance. It may see an object only at certain times. A small object may be harder to track than a large one.
The future path also depends on forces. Gravity is the main force. But other forces matter too. In low Earth orbit, thin upper air creates drag. Drag is a slowing force caused by air. The air is very thin at orbital height, but it is not zero.
Space weather can change this upper air. Activity from the Sun can make the upper atmosphere expand. When that happens, a low-orbit satellite may feel more drag. More drag can change its future position.
An active satellite may also perform planned orbit control. Public conjunction assessment does not always know every future action by every satellite operator. This adds another kind of uncertainty.
Because of all this, an orbit prediction is not a perfect line. It is a best estimate with an error range.
Covariance: The Error Cloud
Covariance is a technical word, but the idea can be simple. Covariance describes the shape and direction of uncertainty in a prediction.
Think of a weather forecast map. The forecast may say a storm center will be near a city tomorrow. But the storm could be a little north, south, east, or west. The possible area is not always a perfect circle. It may be stretched like an oval.
For a satellite, covariance is like an invisible error cloud around the predicted position. The satellite is expected to be near the center of the cloud, but the real location may be somewhere inside the cloud.
The shape of the cloud matters. It may be longer along the direction of travel. It may be narrower across the path. It may be tilted. It may grow as the prediction looks farther into the future.
This happens because errors are connected. If an object is a little ahead of where we expected, it may also be at a slightly different altitude or speed. Covariance helps describe those connected errors.
In simple terms, covariance says: “Here is not only where we think the object will be, but also how the error is shaped.”
Why the Cloud Is Often Stretched
An object in orbit moves quickly along its path. A tiny timing error can look like a large position error along the direction of travel. If the object arrives a fraction of a second early or late, it can be many meters away from the predicted point.
This is why the uncertainty cloud may be longer along the track. Along-track means forward or backward along the orbit path.
The cloud may be smaller in other directions. Cross-track means side to side compared with the path. Radial means toward or away from Earth. The exact shape depends on tracking data and the orbit model.
This is important for conjunction assessment. Two objects may have center points that pass near each other, but their uncertainty clouds may overlap in different ways. The overlap affects the estimated probability.
Collision Probability in Plain English
Collision probability is an estimate of how likely it is that two objects will hit. It is often called probability of collision, or Pc for short.
Probability does not mean certainty. A high probability case may still miss. A low probability case may still hit. Probability is a way to reason under uncertainty.
Here is a simple analogy. Suppose you drop a marble onto a table, but you cannot see exactly where it will land. You draw a fuzzy circle showing the likely landing area. Now place a small cup on the table. The chance that the marble lands in the cup depends on the size of the cup, the size and shape of the fuzzy area, and where the cup sits inside that area.
For two satellites, the “cup” is the combined size of the objects and the unsafe space around them in the math model. The fuzzy area comes from the uncertainty in their relative positions. Relative means one object compared with the other.
If the error cloud is wide and the objects are near the middle of the possible overlap, the probability may be higher. If the predicted paths are well separated or the uncertainty cloud does not overlap much, the probability may be lower.
The result is still only an estimate. It depends on the quality of the tracking data, the assumptions in the model, and how the uncertainty is represented.
Thresholds as Decision Lines
Operators often use thresholds to help decide what to do. A threshold is a decision line. If a risk estimate is below the line, the case may be watched. If it is above the line, the case may get more attention.
This article will not give exact Starlink thresholds. Public sources do not provide a full private rulebook, and exact operational limits can depend on many details.
The concept is still useful. A threshold helps a team avoid guessing from scratch every time. It creates a repeatable way to sort risk.
But thresholds are not magic. A probability number is not the whole story. Operators may also care about data quality, time until closest approach, whether the other object can maneuver, the cost of changing orbit, mission impact, and whether new tracking data is expected soon.
A decision line is a tool. It is not a substitute for judgment, system design, and coordination.
The Maneuver Tradeoff
A collision avoidance maneuver is a planned change to a satellite’s path to reduce risk. For an active satellite, this can be possible. For debris or a dead satellite, it is not.
At a high level, a maneuver has benefits and costs. The benefit is that it may reduce the chance of collision. The costs can include fuel use, mission disruption, changed service coverage, schedule changes, and extra work to return to the planned orbit.
For Starlink, service quality matters because the satellites are part of a network. Moving one satellite may affect how the constellation is arranged. Constellation means a group of satellites working together.
There is also a timing problem. If a satellite moves too early, later tracking data might show the original risk was lower than first thought. If it waits too long, there may be less time to make a good decision. Real systems must balance these concerns.
This is why conjunction assessment is a decision problem, not only a math problem.
Why Not Maneuver Every Time?
If there is any possible risk, why not move every time? The answer is that unnecessary maneuvers have real costs.
First, fuel is limited. Many satellites use onboard propellant for orbit control. Using fuel for one event can reduce flexibility later.
Second, a maneuver changes the satellite’s future path. That new path must also be screened. A move that avoids one predicted close approach should not create a worse situation later.
Third, too many maneuvers can make operations more complex. A large constellation needs stable planning. It cannot treat every weak warning as a full emergency.
Fourth, false alarms happen. A false alarm is a warning that later turns out not to need action. False alarms are normal when systems are designed to catch possible risks early.
Good conjunction assessment tries to reduce real risk while avoiding needless action.
False Alarms and Missed Risks
Every warning system has two types of mistakes. It can warn too often, or it can miss something important.
If the system warns too often, operators may face many false alarms. This can waste time and fuel. It can also make it harder to see the truly serious cases.
If the system warns too little, it may miss a real risk. That is worse for safety.
The hard part is finding a good balance. A large constellation cannot depend on human attention for every tiny case. It needs automated filters that find possible risks early, then better data and better analysis to narrow the list.
Think of a smoke alarm. You want it to warn you about fire. You do not want it to scream every time you make toast. But you also do not want it to stay quiet during a real fire. Space safety screening has a similar balance, but with different tools and much higher speeds.
Why Large Constellations Need Automation
Automation means using computers to perform repeated tasks. For conjunction assessment, automation is not just helpful. It is necessary.
A small number of satellites can be checked by a small operations team. A very large constellation creates many more possible pairings. Each satellite must be compared with many tracked objects. Each comparison may need updates as new tracking data arrives.
The number of possible checks can grow very quickly. If there are many satellites and many other objects, the system must sort huge numbers of possible future passes. Most are harmless. A few need more study.
Automation can do the first pass quickly. It can look for close approaches, estimate uncertainty, rank cases, and alert humans or higher-level systems when needed. It can also repeat the process as new data arrives.
This does not mean humans are unimportant. Humans design the rules, monitor the system, review unusual cases, coordinate with other operators, and improve the process over time. Automation handles the volume. Human responsibility remains important.
Data, Time, and Coordination
Conjunction assessment is only as good as its inputs. The most important inputs are the estimated positions and velocities of the objects. Velocity means speed and direction.
If an object is well tracked, its uncertainty may be smaller. If it is poorly tracked, its uncertainty may be larger. A larger uncertainty cloud can make risk harder to judge. Fresh data also matters. Old data can become less useful because orbit predictions drift over time.
Time before closest approach is important too. Closest approach means the moment when two objects are expected to be nearest to each other. Several days before that moment, there may be enough time to wait for more data. Later, the prediction may be clearer, but there is less time to act.
Space safety is also not only a single-company task. Satellites share orbital regions. At a public level, coordination can include sharing orbit information, understanding who can maneuver, and avoiding confused actions. Better shared information can reduce uncertainty and support better decisions.
Why Public Readers Should Be Careful With Risk Numbers
Public reports sometimes mention a probability number or a close approach distance. These numbers can sound exact. They are not always simple to interpret.
A probability number depends on the data source, the orbit estimate, the covariance, the assumed object sizes, and the calculation method. A distance number depends on the predicted center paths. Without uncertainty, distance alone can be misleading.
For Starlink, exact internal decision rules are not public. It is better to say, “This is how the concept works,” than to claim a private threshold or private algorithm.
Good public discussion should avoid pretending to know live operational details. It should also avoid turning a general risk concept into a command for how a satellite should move.
How This Fits Starlink
Starlink is a large satellite network. Its size makes conjunction assessment a core safety function. The system must watch many satellites, compare them with many tracked objects, update predictions, and decide which cases deserve attention.
At a public level, Starlink satellites are active spacecraft with orbit control capability. That means they can adjust their paths when needed. But the exact rules for when and how Starlink satellites maneuver are not fully public.
The important lesson is not a secret number. The important lesson is the decision structure.
First, predict future positions. Second, describe uncertainty. Third, estimate risk using probability and covariance. Fourth, compare the case with decision criteria. Fifth, consider the cost and benefit of action. Sixth, keep updating the assessment as new data arrives.
This structure is useful for understanding many satellite systems, not only Starlink.
Conclusion
Starlink conjunction assessment is not just a question of whether two satellites are close. It is a question of prediction under uncertainty.
The predicted path is the center of the story, but not the whole story. Covariance describes the error cloud around that path. Collision probability uses that uncertainty to estimate risk. Thresholds help sort cases, but they do not replace judgment. Maneuvers can reduce risk, but they have costs and can create new planning needs.
Large constellations make this work harder because there are many satellites and many possible close approaches to screen. Automation is needed to handle the volume, update cases, and focus attention where it matters.
For public readers, the safest way to understand Starlink conjunction assessment is to focus on the decision problem. Space operators must compare uncertain predictions, avoid overreacting to false alarms, avoid missing real risks, and keep updating the answer as better data arrives.
That is the heart of conjunction assessment: not a single magic number, but a careful process for making safer decisions in a busy orbital environment.
Leave a Reply