Radiative and dynamic contributions to the observed temperature trends in the Arctic winter atmosphere

Abstract

The Arctic has been experiencing unprecedented changes in recent years due to anthropogenic greenhouse gas emissions leading to what is known as Arctic amplification, whereby the Arctic is warming faster than any other area of the planet. While the majority of research has focused on the near-surface level heating and radiative forcing, one under-explored area of research is the change in temperature and heating rates throughout the stratosphere and upper troposphere, particularly during the winter months (December–February). For instance, reanalysis data has revealed that the Arctic middle-lower stratosphere has been warming at a rate of approximately 0.5 K/decade over the 1980–2019 period in contrast to the prevailing cooling trends in the other regions of the stratosphere. To understand what is driving these non-negligible temperature trends, this work investigates the underlying dynamical and radiative heating rates over the same 40-year period. It is found that dynamics is the main driver of the stratospheric temperature change in the Arctic, with the middle-lower stratospheric warming being largely explained by trends in adiabatic motion. In comparison, besides the dynamical effects, the tropospheric warming trend is also driven radiatively by surface warming and thermodynamically by latent heating. Lastly, it is found that the stratospheric warming trend mainly occurs in the years with sudden stratospheric warming events, highlighting the influence of these events in driving the stratospheric temperature trends.

Appendices

Appendix 1: Comparison of temperature trends using other reanalysis data

To ensure that the temperature trend in the DJF Arctic atmosphere is not unique to only the ERA5 data set, we have also determined the temperature trends over the same period using the JRA-55 (Kobayashi et al. 2015) and MERRA-2 (Gelaro et al. 2017) reanalysis data sets, as seen in Fig. 13. From this figure we see that a similar patterns emerges among the reanalysis products, indicating that this feature is not data set specific.

Fig. 13

The December–February averaged temperature trends in the Arctic according to ERA5 reanalysis (a) which extends from 1980 to 2019, the JRA-55 reanalysis (b) extending from 1980 to 2019, and MERRA2 (c) which covers the 1981–2019 period. Units are in K/decade and the black dots indicate where the trends are significant at the 99% level

Appendix 2: Heating rate climatologies

To better assess the heating rate trends, we provide here, in Fig. 14, a break down of the climatological heating rates. This figure reveals an atmosphere largely in radiative-dynamic equilibrium whereby the heat supplied through dynamics is then cooled via radiation, as indicated by figures a and e. From a climatological perspective, the non-radiative parameterized physics and the total temperature tendency are secondary in terms of their magnitude and heating rate contributions during the polar winter.

Fig. 14

The December–February averaged climatological: radiative heating rate (a), heating rate from non-radiative parameterized physics (b), dynamical heating rate (c), residual heating (d), dynamical plus residual heating rates (e), and the total temperature tendency (f) in the Arctic

Appendix 3: Comparison of ERA5 and RRTMG radiative heating rates

The decision to use RRTMG-derived radiative heating rates to represent $\Delta \mathbf{H} _{SW}$ in Eq. (10) and $\Delta \mathbf{H} _{LW}$ in Eq. (11) as opposed to the long and shortwave heating rates provided by ERA5 was based on the fact that the ERA5 radiative heating rates use a climatological value of carbon dioxide and ozone, instead of one that varies in time. If this research was investigating climatological heating rates then using ERA5 data would be more justified given the relatively good agreement between the climatological ERA5 and RRTMG-derived radiative heating rates, as seen in Fig. 15. Discrepancies arise, however, when we look at the radiative heating rate trends over the 1980–2019 period. The impact on the residual radiative heating rate, determined using Eqs. (10) and (11), when we use the ERA5 data, as opposed to the RRTMG-derived values can be see in Fig. 16. While the residual appears to be smaller for the longwave heating rates when using ERA5 data, the same cannot be said of the shortwave heating rates. The residual that appears in Fig. 16d, when the ERA5 data has been used, provides a nearly identical pattern to the shortwave heating rate trend due to changes in ozone (see Fig. 4b). This means that this residual is due to the use of climatological ozone in the ERA5 radiative heating rates. It is for this reason that we do not use the short and longwave radiative heating rates provided by ERA5, except to determine the non-radiative parameterized physics.

Fig. 15

The December–February averaged climatological longwave radiative heating rates provided from ERA5 (a), RRTMG-derived (b), and the difference between the two (c). Also depicted is the climatological shortwave heating rates according to ERA5 (d), RRTMG (e), and their difference (f)

Fig. 16

The December–February 1980–2019 trends of the residual and higher order longwave heating rate as determined when ERA5 longwave heating rates are used to represent $\Delta \mathbf{H} _{LW}$ in Eq. (11) (a) and when the RRTMG-derived total longwave radiative heating rates are used instead (b). Panel c represents the difference between ERA5 and RRTMG residual longwave heating rate trends. Also depicted are the residual and higher order shortwave heating rates as determined using ERA5 shortwave heating rates to represent $\Delta \mathbf{H} _{SW}$ in Eq. (10) (d), when the RRTMG-derived total longwave radiative heating rates are used (e), and their difference (f)