Accurate and Efficient Numerical Simulation of Land Models Using SUMMA With SUNDIALS.

Area of Science:

• Earth System Science
• Computational Hydrology
• Land Surface Modeling

Background:

• Numerical simulations of land models often suffer from inaccuracies due to the lack of rigorous error control.
• Previous versions of the Structure for Unifying Multiple Modeling Alternatives (SUMMA) land model utilized a low-order backward Euler method without error estimation, limiting simulation reliability.

Purpose of the Study:

• To enhance the accuracy and computational efficiency of the SUMMA land model.
• To integrate advanced numerical solvers for improved simulation of land surface thermodynamics and hydrology.

Main Methods:

• Incorporation of the Suite of Nonlinear and Differential-Algebraic Equation Solvers (SUNDIALS) package into the SUMMA land model.
• Utilizing SUNDIALS' algorithmic features, including error estimation and adaptive order/step-size control.
• Conducting large-scale continental simulations over North America with over 500,000 spatial elements.

Main Results:

• The SUMMA-SUNDIALS model demonstrates substantially improved accuracy compared to the previous SUMMA model.
• Simulations achieved results orders of magnitude closer to converged solutions at a comparable computational cost.
• Enhanced relative computational efficiency was observed.

Conclusions:

• The integration of SUNDIALS provides a more accurate and computationally efficient land surface modeling tool.
• The SUMMA-SUNDIALS model offers a powerful capability for reliable simulations, advancing the understanding of the terrestrial component of the Earth System.