A Fast ODE Solver for Atmospheric Chemistry: ROS3 with H211b Control and Retained Steps

Solvers that restart every host interval of an operator-split atmospheric model from a small step repeat initialization work that the state does not require. We describe a solver that combines the three-stage, third-order Rosenbrock method ROS3 with step sizes retained across intervals and rounded to divide them, Söderlind’s H211b step-size filter with its history retained, and an absolute tolerance equal to the relative tolerance times the concentration threshold of the accuracy assessment. We explain the method, its selection by structure scans, and its implementation, and weight operation counts by the mechanism’s sparse structure. On eleven MOZART-TS1 scenarios, measured against converged references for all species above 10−14 molm−3, a pooled RMS relative error of at most 0.1%, with every species at most 1%, requires about one third of the arithmetic of a restarting ROS3 solver; the advantage decreases at higher accuracy. Variants isolate the benefits of retained controller history and step rounding. Backward Euler, as the Community Atmosphere Model with chemistry uses it, requires substantially more arithmetic at comparable accuracy. In a compiled library, the restarting solver takes 3.5 times as long as ours at looser error limits (1% pooled, 10% per-species RMS relative error) and 3.0 times as long at tighter limits (0.1%, 1%); host-model integration and timing remain untested.

Authors

Institutions

Publication Details

Published
2026-11-13
DOI
https://doi.org/10.22541/essoar.15012147/v1
Primary Topic
Numerical Solvers for Ordinary Differential Equations
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

A Fast ODE Solver for Atmospheric Chemistry: ROS3 with H211b Control and Retained Steps

David Fillmore
2 links
Numerical Solvers for Ordinary Differential Equations
preprint

A Fast ODE Solver for Atmospheric Chemistry: ROS3 with H211b Control and Retained Steps

David Fillmore
preprint en

Abstract

Solvers that restart every host interval of an operator-split atmospheric model from a small step repeat initialization work that the state does not require. We describe a solver that combines the three-stage, third-order Rosenbrock method ROS3 with step sizes retained across intervals and rounded to divide them, Söderlind’s H211b step-size filter with its history retained, and an absolute tolerance equal to the relative tolerance times the concentration threshold of the accuracy assessment. We explain the method, its selection by structure scans, and its implementation, and weight operation counts by the mechanism’s sparse structure. On eleven MOZART-TS1 scenarios, measured against converged references for all species above 10−14 molm−3, a pooled RMS relative error of at most 0.1%, with every species at most 1%, requires about one third of the arithmetic of a restarting ROS3 solver; the advantage decreases at higher accuracy. Variants isolate the benefits of retained controller history and step rounding. Backward Euler, as the Community Atmosphere Model with chemistry uses it, requires substantially more arithmetic at comparable accuracy. In a compiled library, the restarting solver takes 3.5 times as long as ours at looser error limits (1% pooled, 10% per-species RMS relative error) and 3.0 times as long at tighter limits (0.1%, 1%); host-model integration and timing remain untested.

NSF National Center for Atmospheric Research (US), Langley Research Center (US), National Aeronautics and Space Administration (US)
National Science Foundation
Numerical Solvers for Ordinary Differential Equations
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon