The PDFs of daily mean 2m temperature (T2m) in observational data have been characterized using the first three moments. For identifying the role of dynamical processes, studies focussed on the midlatitudes have analyzed temperature variability at 850 hPa, which represents the free troposphere. The observed skew could not be reproduced by linear theory of advection ([1]), but was owed to the covariance between anomalous winds and anomalous temperature ([2], [3]). Recently, frameworks have also been developed for studying the roles of different processes in driving temperature tendencies in different percentiles of temperature ([3]). However, most of the studies involving advection consider a purely meridional mixing process. Given the dynamical links between meridional and vertical advection, it is unclear if this is sufficient.
We turn focus to T2m, and consider 3D advection. We use the ERA5 reanalysis dataset to study the drivers of variability of T2m anomaly over the northwest Indian heatwave hotspot region during March and April, 1980-2022. We characterize the dry static energy (DSE) fluxes into this region, and develop a framework to identify quasilinear (QL) and nonlinear (NL) advective contributions to the temperature anomaly lifecycle.
Daily change in T2m was highly correlated with daily advection of DSE into a 600-900 hPa box over the region. Leveraging the decision tree framework to identify the dominant weather patterns explaining different terciles of advected DSE, we found that the zonal mean flow and anomalous vertical flow ([1], [2]) acted to reverse the effect of the anomalous meridional flow. Using regression, we established that an additive combination of QL terms involving these flow components served as the dominant mechanism acting throughout the distribution of net advection, with r2 > 0.65. The rest of the variability was almost entirely explained by the sum of NL terms. We saw that the NL sum acts to saturate the growth of the QL sum in its tails, supporting the observations made by [2]. Net advection peaked before the peak of the QL sum due to such a relationship, restricting the growth of net advection.
Furthermore, we study the patterns of advection in a phase space generated by the NL and QL terms. Regimes of advection were readily identified by identifying the NL terms dominating a particular region of the phase space.
We show how interpretable machine learning algorithms, like decision tree and regression, can be used to identify dominant circulation patterns and provide a mapping between magnitude of advection and eddy configurations with respect to the region of interest.
References
[1] Schneider, T., T. Bischoff, and H. P lotka, 2015: Physics of Changes in Synoptic Midlatitude Temperature Variability J. Climate, 28, 2312–2331.
[2] Garfinkel, C. I., and N. Harnik, 2017: The Non-Gaussianity and Spatial Asymmetry of Temperature Extremes Relative to the Storm Track: The Role of Horizontal Advection. J. Climate, 30, 445–464.
[3] Tamarin-Brodsky, T., K. Hodges, B. J. Hoskins, and T. G. Shepherd, 2019: A Dynamical Perspective on Atmospheric Temperature Variability and Its Response to Climate Change. J. Climate, 32, 1707–1724.