The Climate Impact of COVID19 Induced Contrail Changes

. The COVID19 pandemic caused signiﬁcant economic disruption in 2020 and severely impacted air trafﬁc. We use a state of the art Earth System Model and ensembles of tightly constrained simulations to evaluate the effect of the reductions in aviation trafﬁc on contrail radiative forcing and climate in 2020. In the absence of any COVID19 pandemic caused reductions, the model simulates a contrail Effective Radiative Forcing (ERF) of 62 ± 59 mWm − 2 (2 standard deviations). The contrail ERF has complex spatial and seasonal patterns that combine the offsetting effect of shortwave (solar) cooling and longwave 5 (infrared) heating from contrails and contrail cirrus. Cooling is larger in June–August due to the preponderance of aviation in the N. Hemisphere, while warming occurs throughout the year. The spatial and seasonal forcing variations also map onto surface temperature variations. The net land surface temperature change due to contrails in a normal year is estimated at 0.13 ± 0.04 K (2 standard deviations) with some regions warming as much as 0.7K. The effect of COVID19 reductions in ﬂight trafﬁc decreased contrails. The unique timing of such reductions, which were maximum in N. Hemisphere spring and summer 10 when the largest contrail cooling occurs, means that cooling due to fewer contrails in boreal spring and fall was offset by warming due to fewer contrails in boreal summer to give no signiﬁcant annual averaged ERF from contrail changes in 2020. Despite no net signiﬁcant global ERF, because of the spatial and seasonal timing of contrail ERF, some land regions that would have cooled slightly (minimum -0.2K) but signiﬁcantly from contrail changes in 2020. The implications for future climate impacts of contrails are discussed.

emissions and contrails. We do not consider effects of aviation aerosols in this study. The ACCRI 2006 inventory was developed based on detailed flight track data. The distribution of flight level water vapor emissions is shown in Figure 1B. We make the assumption that air traffic has increased significantly since 2006, but that the flight locations and relative density have not changed drastically. In some rapidly developing regions of the planet such as China, this assumption will result in some additional uncertainty.
To estimate the 2020 emissions we estimate the growth in fuel use since 2006 as equal to the growth in total aircraft distance 80 traveled using data from Lee et al. (2021). Lee et al. (2021)  This data shows growth over the last few years of 9%/yr. We thus use a 9%/year increase over 2018-2020 (2 years) to generate a scaling from 2006 to 2020 of 1.88 (88% increase from 2006) in a scenario without any COVID19 induced reduction to 85 aviation.
In order to determine the perturbation due to COVID19 lockdowns, we use daily data for total flights for each day of 2020 provided by Flightradar24 (available at, https://www.flightradar24.com/data/statistics), illustrated in Figure 1A, and compare this to a scaled up average of previous years 2016-2019, which is 9% above 2019 ( Figure 1C). We use weekly averages since there is a strong weekly cycle in flights ( Figure 1A). The analysis yields a scaling value for every week of 2020 from our 90 reference (scaled up 2006 emissions), as illustrated in Figure 1D. The first few weeks of 2020 were normal, then reductions started in February 2020 due to restrictions in China and Asia, and then in March (around week 12), most nations began lockdowns and most commercial flights were halted. Total aviation declined by 2/3 from what would have been expected.
Recovery was rapid for about 10-15 weeks until the middle of 2020, and then recovery has slowed, reaching approximately 75-80% of the expected value by the end of 2020. Note that this is total flights, including commercial (passenger and cargo), 95 private and military (with transponders). The total load factor on passenger flights has decreased, so the total passenger miles flown is different than this. But it is total flights that is most relevant for water vapor emissions.
We then have a scaling factor for 2020 from 2006 (1.88) and weekly modifications to that factor for COVID19 impacted emissions. These aviation water vapor emissions are used in our simulations to initiate contrails. All other emissions come from the Shared Socioeconomic Pathway (SSP) 245 emissions for 2019-2020 and are the same for all simulations.

Simulations
Full aviation simulations with 10 ensemble members are launched from January 1, 2019 to December 31, 2020 (2 years), with a small temperature perturbation (10 −10 K). The initial perturbation results in a slightly different atmosphere evolution for each ensemble member. Nudging keeps the atmosphere in a similar 'weather' state. The perturbation samples random fluctuations within that state. Critically, this enables estimates of the statistical significance of differences. We compared 10 105 and 20 ensemble members, and found that 10 members did not change the standard deviation and significance levels for full aviation emissions. We define statistical significance for maps with the False Discovery Rate (FDR) method of Wilks (2006), which reduces patterned noise. We use the standard deviation across the ensembles to estimate uncertainty and variability for global averaged quantities. A similar methodology was used to examine non-aviation COVID19 related aerosol emissions perturbations by Gettelman et al. (2021).
We run simulations with full aviation water vapor emissions (Full Air) and no aviation water vapor emissions (No Air). We can analyze 2019 and 2020 effects with different meteorology in the 2 year simulations. As will be noted below, the land surface takes a few months to react to adjusted forcing ( Figure 2G), but the other variables adjust quickly (see section 3.1). We also run an ensemble of 20 members restarted January 1, 2020 with COVID19 reduced aviation water vapor emissions (COVID) for 2020. 20 ensemble members are used due to the smaller peturbation. Finally we also run a pair of 2020 ensembles with 115 temperature nudging (Full Air T Nudge, COVID T Nudge) to explore how the evolution of temperature may affect the results.

Results
First we analyze global mean results by month in Section 3.1. We focus on the differences between ensembles with and without aviation or COVID19 affected aviation for key climate parameters. Then we assess the spatial and seasonal distribution of these parameters (Section 3.2). This puts overall global values in important context for assessing contrail ERF and COVID19 reduc-120 tions to contrails. For clarity in dates we will refer to the COVID19 affected aviation simulations in the figures as 'COVID'.
Finally we look in more detail at cloud changes and the effects of temperature nudging on the climate response to aviation contrails (Section 3.3).  Figure 2B) has virtually no annual cycle. The COVID19 emissions changes should then be noted in the context of this annual cycle. The LW CRE changes due to COVID19 reductions (Fig-135 ure 2B, blue solid and red dashed) clearly shows differences that map directly to the temporal evolution of aviation reductions ( Figure 1D). The phase of the SW CRE ( Figure 2A) and LW CRE ( Figure 2B) for the COVID case do not exactly line up (peak SW in August, peak LW in April). This is because of the convolution between reductions ( Figure 1D) with the peak in the SW contrail effect (Figure 2A). The Ice Water Path (IWP) due to full contrail effects ( Figure 2C) has a small annual cycle, and is similar to LW CRE