| 22 Jul 2026
Interactive Land in E3SM Process-Level Models: Impacts of Land–Atmosphere Coupling in SCM and DP-SCREAM
Abstract. Process-level modeling is central to diagnosing and improving atmospheric parameterizations in Earth system models, yet continental single-column and convection-permitting process configurations are often run with prescribed surface fluxes, limiting realism and consistency with fully coupled global simulations. Here we develop, document, and demonstrate a reproducible workflow to enable fully interactive land–atmosphere coupling in both the E3SM single-column model (SCM) and the doubly periodic configuration of SCREAM (DP-SCREAM). The approach leverages a multi-year offline-forced E3SM/ELM integration to provide spun-up land initial conditions and introduces an offline preprocessing method that generates DP-SCREAM–compatible land restart files by extracting and replicating the appropriate ELM grid cell (including its landunit/column/PFT hierarchy) across the doubly periodic domain, avoiding intrusive modifications to ELM I/O. Using daily-initialized two-day hindcast integrations, we apply this framework to two years over GoAmazon and ten years over the Southern Great Plains, and to targeted short process-level cases (CASS and GoAmazon single/double pulse). Interactive land coupling is required to reproduce key near-surface thermodynamic biases characteristic of global coupled simulations, including warm 2-m temperature biases and the elimination of spurious persistent nocturnal fog layers that arise in prescribed-flux configurations. In contrast, prominent cloud and convection process-level biases identified previously (i.e. insufficient deepening of shallow convection, deficits of mid-level congestus, and weak convective organization) are largely insensitive to prescribed versus interactive surface flux treatments, indicating that they are driven primarily by atmospheric model physics and/or dynamics. These results provide both a community-ready capability for coupled land process-level simulations in E3SM and practical guidance on when interactive versus prescribed surface fluxes are appropriate for process-level model evaluation and development.
This study develops, documents, and demonstrates a reproducible workflow for enabling fully interactive land–atmosphere coupling in both E3SM SCM and DP-SCREAM. The authors performed multi-year offline-forced E3SM/ELM integrations and developed an offline Python preprocessing script to generate DP-SCREAM-compatible land restart files without modifying ELM's core I/O infrastructure, thereby obtaining realistic spun-up land initial conditions. Using long-term hindcast simulations over the GOAmazon and SGP campaigns, and short-term process-level case studies of shallow and deep convections, this paper find that interactive land coupling is essential for reproducing near-surface thermodynamic biases characteristic of global simulations, including warm 2-m temperature biases. Prominent cloud and convection biases (insufficient deepening of shallow convection, deficits of mid-level congestus, weak convective organization) are largely insensitive to the choice between prescribed fluxes and interactive land treatments, suggesting that they arise primarily from atmospheric model physics and dynamics.
The study provides practical guidance for the community on when interactive versus prescribed surface fluxes are appropriate for process-level model evaluation. The results are scientifically important and offer valuable insights for future process-level studies using E3SM SCM and DP-SCREAM. The technical workflow for generating spun-up land initial conditions is well documented and will be of value to the broader community. The manuscript is generally well-written, logically organized, and supported by extensive simulations. I have only minor comments for the authors to consider.
The first comment is about the comparison with interactive land options in other SCMs. As the authors mentioned, technical option for interactive land has existed in E3SM SCM. Also, I remember SCAM6 uses interactive land as its default settings. However, the current introduction gives the impression that all existing single-column models rely on prescribed surface fluxes. I would strongly recommend adding discussions of interactive land options in other SCMs within the introduction, along with a clear explanation of how the approach presented here differs from those existing implementations.
Another topic which may be outside the primary scope of this paper but I believe worth a brief discussion is air-sea interactions. The present study addresses land–atmosphere coupling for continental cases only. For oceanic cases, current SCMs similarly rely on prescribed surface fluxes. Some coupled atmosphere–ocean single-column models have been developed (e.g., EC-Earth; Hartung et al., 2018). What are the similarities and differences between air–sea interactions in oceanic cases and land–atmosphere interactions discussed here? Are there plans to extend this framework to include air–sea coupling in E3SM SCM in future work?
Some other minor comments:
A few abbreviations are used without their full names being provided, such as SHOC and SNICAR-AD. Please define these at first use.
Figure 5: The late-afternoon peak is not clearly illustrated in Figure 5. I recommend specifying that this late-afternoon peak corresponds to the 02 UTC time step, which represents approximately 20:00 LST.
Line 312-314: I don't quite understand this logical chain. why excessive boundary layer mixing may maintain overly warm near-surface temperature? Would the higher sensible heat flux and excessive mixing be associated with a stronger land-air temperature gradient, which would imply colder air temperature?
Line 356-357: This explanation provided here focuses on the contrast between interactive land and prescribed fluxes. How does this extend to the comparison between SCM and DP-SCREAM? Specifically, why does DP-SCREAM with interactive land fail to represent shallow cumulus as well as the SCM configuration?