Development and evaluation of a fast chemistry solver for atmospheric modeling
- ENVIRON International Corp., Novato, CA (United States)
In most photochemical grid models, the majority of the CPU time is spent numerically integrating the time evolution of (solving) the chemistry. One approach to speeding up the chemistry calculations is to use parallel computers combined with a solver developed to exploit the parallel architecture. A potential drawback to this approach is that the resulting code may be tailored to a specific computer architecture and only run efficiently on a limited number of rather expensive computing platforms. Development of a fundamentally more efficient chemistry solver is an attractive alternative since the benefits will be realized on all computing platforms. The authors developed a highly efficient chemistry solver. Several numerical algorithms are built into the chemistry solver and during model runs an appropriate algorithm is selected for each {open_quotes}call{close_quotes} to the chemistry based on the chemical conditions for that call. The result is that a very fast algorithm is used most of the time but a slower, more robust algorithm is used in situations where the fast algorithm might become inaccurate. This solver has been implemented in the Urban Airshed Model (UAM), the model used for regulatory ozone modeling in the U.S. The adaptive-hybrid solver results in about a ten-fold speedup in the chemistry calculations and therefore an overall speedup in the model of 3 to 4 times. Model performance is very similar to the standard version of UAM. Most importantly, the response of ozone concentrations to changes in VOC and NO{sub x} emissions is almost identical between {open_quotes}standard{close_quotes} and {open_quotes}fast{close_quotes} versions of UAM. Use of the fast solver does not change any conclusions that would be drawn about model performance relative to observations or the effectiveness of emission control strategies. This paper will describe the methodology used in our adaptive-hybrid fast chemistry solver and present results of its implementation in UAM.
- OSTI ID:
- 466222
- Report Number(s):
- CONF-9606185-; TRN: 96:005931-0173
- Resource Relation:
- Conference: 5. international conference on atmospheric sciences and applications to air quality, Seattle, WA (United States), 18-20 Jun 1996; Other Information: PBD: 1996; Related Information: Is Part Of The 5th international atmospheric sciences and applications to air quality conference; PB: 322 p.
- Country of Publication:
- United States
- Language:
- English
Similar Records
Process analysis: A quantitative technique for evaluating the importance and functioning of chemical and physical processes in air quality models
Multi-node and Multi-core Performance Studies of a Monte Carlo code RMC