Global Effects of Superparameterization on Hydrothermal Land‐Atmosphere Coupling on Multiple Timescales

Abstract

Many conventional General Circulation Models (GCMs) in the Global Land‐Atmosphere Coupling Experiment (GLACE) tend to produce what is now recognized as overly strong land‐atmosphere (L‐A) coupling. We investigate the effects of cloud Superparameterization (SP) on L‐A coupling on timescales beyond diurnal where it has been recently shown to have a favorable muting effect hydrologically. Using the Community Atmosphere Model v3.5 (CAM3.5) and its Superparameterized counterpart SPCAM3.5, we conducted soil moisture interference experiments following the GLACE and Atmospheric Model Intercomparison Project (AMIP) protocols. The results show that, on weekly‐to‐subseasonal timescales, SP also mutes hydrologic L‐A coupling. This is detectable globally, and happens through the evapotranspiration‐precipitation segment. But on seasonal timescales, SP does not exhibit detectable effects on hydrologic L‐A coupling. Two robust regional effects of SP on thermal L‐A coupling have also been explored. Over the Arabian Peninsula, SP reduces thermal L‐A coupling through a straightforward control by mean rainfall reduction. More counterintuitively, over the Southwestern US and Northern Mexico, SP enhances the thermal L‐A coupling in a way that is independent of rainfall and soil moisture. This signal is associated with a systematic and previously unrecognized effect of SP that produces an amplified Bowen ratio, and is detectable in multiple SP model versions and experiment designs. In addition to amplifying the present‐day Bowen ratio, SP is found to amplify the climate sensitivity of Bowen ratio as well, which likely plays a role in influencing climate change predictions at the L‐A interface.

Effects of Explicit Convection on Land Surface Air Temperature and Land‐Atmosphere Coupling in the Thermal Feedback Pathway

“…Thus, changes in SM dynamics are unlikely play a major role in changes of the LSAT distribution in these grid points. This is consistent with the arguments of Qin et al () that a mostly atmospheric effect of SP impacting time mean precipitation, cloud cover, and mean SM dominates surface temperature changes over the Arabian Peninsula. Europe and Australia . There is no dominant relationship between EF and SM at these locations, but a small fraction of days fall in the SM‐limited regime in both models.…”

“…From the distribution of surface SM variability (Figures c and d), the mitigation of SM and temperature coupling signals over the Sahara and the Middle East is mainly due to nearly no variability of SM over those areas in SPCAM. This is consistent with previous findings using an earlier version of SPCAM that SP dramatically reduces precipitation variability in these regions (Sun & Pritchard, ) likely due to a reduction in time mean cloud and precipitation that shifts the area away from being an artificial transitional zone (Qin et al, ).…”

“…It is interesting that SP does not produce any reduction of the U.S. warm bias despite its ability to capture mesoscale convective systems (Pritchard et al, ). Elsewhere, Figure confirms that SPCAM improves the mean LSAT (cold bias) over the Sahel and the Arabian Peninsula dramatically (Figures a and b) for reasons highlighted in Qin et al (). Both models produce less LSAT variability than ERA‐Interim at daily time scales as shown in Figures c and d; this is to be expected due to the removal of interannual SST variability in the idealized annual cycle boundary conditions.…”

“…The Sahel is a well‐known transitional zone where SM plays a major role in land feedbacks to the atmosphere. Previous studies have also indicated that SP alters the precipitation distribution and land‐atmosphere coupling (in the water feedback pathway) strength in this region (Qin et al, ; Sun & Pritchard, ). Together, this suggests that the Sahel is one potentially interesting geographic action center to explore the relationship between SP, the surface temperature distribution, and land‐atmosphere coupling.…”

Could Machine Learning Break the Convection Parameterization Deadlock?

“…To alleviate this problem, an interesting approach has been to use cloud “superparameterization (SP),” which computes the subgrid vertical heating and moistening profiles within a GCM grid cell by sampling a curtain of an embedded 2‐D CRM that uses convective permitting resolution (Grabowski, ; Khairoutdinov et al, ). This has led to many successes such as the possibility to rectify the diurnal continental cycle, to improve the representation of the MJO, and to represent both some MCS propagation and some degree of aggregation, and reduce overly strong land‐atmosphere coupling (Benedict & Randall, 2009; Grabowski, ; Holloway et al, , ; Khairoutdinov et al, ; Kooperman et al, , ; Pritchard & Somerville, ; Pritchard et al, ; Qin et al, ; Randall, ; Sun & Pritchard, ).…”