Backward uncertainties Monte Carlo Propagation

Hello,

I am working on injection window sizing for a nanosatellite mission study (SSO ~400 km, 2-year lifetime) and would like feedback on whether my overall approach makes sense before investing more in it.

Goal

Given End-of-Life (EOL) orbit requirements after two years (altitude 350 ± 2 km, inclination ± 0.05°, …), I want to determine the required Beginning-of-Life (BOL) injection tolerances: what initial orbit do I need to guarantee ending up within spec?

Approach

Instead of a classical forward Monte Carlo (which requires guessing the injection window first, then iterating), I am running a backward Monte Carlo:

  1. Draw N final states uniformly from the EOL requirement box. The sampled parameters are:

Orbital elements: altitude, eccentricity, inclination, RAAN, argument of perigee
Spacecraft: mass, drag coefficient C_d, drag cross-section area
Environment: F10.7 solar flux index, sampled over the expected range for the mission epoch
I treat F10.7 and B* as the dominant drivers of altitude decay, but all parameters are varied simultaneously to capture cross-effects.

  1. Propagate each one backward for the mission duration using DSSTPropagator
  2. Collect the resulting initial states and study the relationship between final and initial uncertainties - try to find the initials orbits that verify the EOL specs.

Questions

  1. Is this backward MC methodology the right approach to size an injection window, or is there a more standard method in mission analysis?
  2. Is DSSTPropagator reliable for backward propagation under significant drag (350–425 km, F10.7 up to 150 SFU)?
  3. Since all parameters vary simultaneously across samples, simple pairwise correlation between an input and an output can be misleading — a third parameter may be driving both. What is the recommended way in Orekit workflows to isolate the independent contribution of each uncertain parameter?
  4. Monte Carlo simulations can be computationnaly costly, how could I avoid N propagations (2 years DSST) ?

Thank you for your help !