In this paper, a Sonabend-W et al. quantified the longwave radiative forcing of contrails at the top-of-atmosphere (TOA) based on GOES-16 satellite observations and ERA5 reanalysis, and estimated that the longwave radiative forcing of contrails is 46.9 GJ on average over the Americas. The authors apply causal inference to discern the effect of contrails while controlling for radiative and cloud confounders. The authors compared their results with CoCiP data, and illustrates how the longwave warming of contrails varies between day and night.

This method is plausible, but the results are unreasonable, so the authors should check the details to correct any potential mistakes. The paper might be accepted after addressing the following issues:

1. According to Fig. 10, the new method yields a longwave forcing of zero at noon, which is not reasonable. Unless there is no contrail at all at noon (which is untrue), the longwave forcing of contrails should not be zero. Furthermore, the surface temperature is higher at noon-time, so theoretically the longwave forcing of a contrail should be significant. Therefore, the CoCiP RF is more reasonable than that calculated by the new method.

Note: The authors added some discussions to address this issue, but it is hard to believe a zero longwave contrail forcing at noon. The unrealistic zero forcing at noon might be induced by issues in the regression process (see the next comment).

1. In Eq. (3), a simple linear regression is used to calculate the parameters in the equation. However, as the authors pointed out, the correlation between independent variables Ai and Aj is large, so linear regression is not valid in this case. If the authors keep using simple linear regression, then this equation should be rewritten.