| 20 Aug 2026
RDycore-sediment v1.0: A two-dimensional sediment transport model for Earth system modeling
Abstract. Sediment transport links land-surface erosion, river and floodplain morphology, reservoir storage, water quality, and downstream material delivery. Yet scalable two-dimensional (2-D) sediment transport representation remains limited in large-scale Earth system modeling frameworks. We present RDycore-sediment v1.0, an extension of the River Dynamical Core version 1.0 (RDycore v1.0), which adds conservative multi-class suspended sediment transport and local water-bed sediment exchange to RDycore v1.0’s finite volume shallow water formulation. The model routes class-specific suspended sediment mass according to RDycore simulated hydrodynamics and accounts for bed erosion, deposition, active-layer exchange, and substrate-layer bed storage. In this first version, bed sediment mass is updated diagnostically, but bed elevation remains fixed and sediment dynamics does not feed back onto flow dynamics. Additionally, we extended the hydrodynamics model within RDycore to include hydrostatic reconstruction of the flow variables. RDycore-sediment is evaluated with both ideal and realistic benchmark cases, including 1) two conventional passive transport benchmarks, 2) a dam break erosion and deposition comparison with TELEMAC/GAIA, 3) a three-class method of manufactured solutions case, and 4) a Hurricane Harvey Houston benchmark driven by runoff and soil erosion inputs. In idealized tests, RDycore-sediment preserves passive transport and lake-at-rest states, reproduces TELEMAC/GAIA erosion and deposition in the idealized dam break, and achieves first-order convergence. For the Hurricane Harvey case, RDycore-sediment and TELEMAC/GAIA simulate comparable event-scale timing and total bed mass changes. The Hurricane Harvey case also highlights the importance of including coastal backwater effects and the use of well-balanced hydrodynamics for sediment applications in complex urban watersheds, as small errors in velocity and shear stress can bias the calculation of bed erosion and deposition. RDycore-sediment provides a scalable foundation for 2-D sediment routing in river and floodplain systems and supports future coupling with Earth system models across the land-river-ocean continuum.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Geoscientific Model Development.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
This manuscript presents RDycore-sediment v1.0, a two-dimensional, multi-class suspended-sediment transport extension of RDycore. The work is relevant to the development of scalable sediment-routing capabilities for large-domain hydrological and Earth-system applications. The implementation is supported by idealized verification cases, a method-of-manufactured-solutions test, and a Hurricane Harvey application. The manuscript is generally well structured and transparently acknowledges that the current version has fixed bed elevation and no sediment–morphology feedback. However, several aspects of the contribution, evaluation, and intended Earth-system application need clarification. I recommend minor revisions.
1 - The main advantage of RDycore-sediment relative to existing sediment models, particularly TELEMAC/GAIA, needs to be much more clearly articulated. The manuscript currently suggests that RDycore-sediment offers a scalable 2-D alternative suitable for large-domain Earth-system applications, but the precise application niche remains unclear. The authors should explain whether the intended advantage is computational efficiency, compatibility with E3SM software infrastructure, large-domain and ensemble capability, GPU portability, simplified process representation, or an ability to couple land-generated runoff and erosion with spatially explicit river–floodplain transport.
2 - The Hurricane Harvey application should be framed as a model intercomparison rather than validation of sediment dynamics. The model uses observed water levels for the downstream boundary and compares simulated stages with USGS observations, but no independent sediment observations are used for model verification. Agreement between RDycore-sediment and TELEMAC/GAIA does not establish that either model accurately represents the real sediment response. It only indicates that the two independently implemented models produce broadly similar behavior under aligned forcing and parameter assumptions.
3 - The claimed Earth-system coupling capability is not demonstrated in the present study. In the Harvey experiment, ELM is run separately using MRMS precipitation and NLDAS atmospheric forcing. Its outputs are subsequently read from external files by stand-alone RDycore-sediment. Likewise, the downstream coastal water level is prescribed from a tide gauge rather than supplied by a dynamically coupled ocean model. Therefore, the study demonstrates an offline land-model-forced river/floodplain simulation with a prescribed coastal boundary, not a E3SM land–river–ocean coupling. Figure 1 should be clearly labeled as a target architecture or future application pathway.
4 - The performance discussion is currently insufficient to support a general claim that RDycore-sediment is faster or more scalable than TELEMAC/GAIA. The reported runtimes are encouraging, but the comparison is not controlled. In particular, TELEMAC/GAIA includes bed-elevation evolution and morphology–flow feedback that are absent from RDycore-sediment v1.0, so its additional computational cost cannot be attributed solely to inferior scalability. The manuscript reports RDycore-sediment timings on two dual-socket AMD EPYC 7763 CPU nodes, whereas TELEMAC/GAIA timings are reported only for one and two “Perlmutter nodes.” It is unclear whether the two models were run on the same hardware (they should).