English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

. Atmospheric chemistry models are a central tool to study the impactof chemical constituents on the environment, vegetation and human health.These models are numerically intense, and previous attempts to reduce thenumerical cost of chemistry solvers have not delivered transformative change. We show here the potential of a machine learning (in this case random forestregression) replacement for the gas-phase chemistry in atmospheric chemistrytransport models. Our training data consist of 1 month (July 2013) ofoutput of chemical conditions together with the model physical state,produced from the GEOS-Chem chemistry model v10. From this data set we trainrandom forest regression models to predict the concentration of eachtransported species after the integrator, based on the physical and chemicalconditions before the integrator. The choice of prediction type has a strongimpact on the skill of the regression model. We find best results frompredicting the change in concentration for long-lived species and theabsolute concentration for short-lived species. We also find improvementsfrom a simple implementation of chemical families(NOx = NO + NO2). We then implement the trained random forest predictors back into GEOS-Chem toreplace the numerical integrator. The machine-learning-driven GEOS-Chem modelcompares well to the standard simulation. For ozone (O3), errors from using therandom forests (compared to the reference simulation) grow slowly and after5 days the normalized mean bias (NMB), root mean square error (RMSE) andR2 are 4.2 %, 35 % and 0.9, respectively; after 30 days the errorsincrease to 13 %, 67 % and 0.75, respectively. The biases become largestin remote areas such as the tropical Pacific where errors in the chemistrycan accumulate with little balancing influence from emissions or deposition.Over polluted regions the model error is less than 10 % and has significantfidelity in following the time series of the full model. ModelledNOx shows similar features, with the most significant errorsoccurring in remote locations far from recent emissions. For other speciessuch as inorganic bromine species and short-lived nitrogen species, errorsbecome large, with NMB, RMSE and R2 reaching &gt;2100 % &gt;400 % and&lt;0.1, respectively. This proof-of-concept implementation takes 1.8 times more time than thedirect integration of the differential equations, but optimization andsoftware engineering should allow substantial increases in speed. We discusspotential improvements in the implementation, some of its advantages fromboth a software and hardware perspective, its limitations, and itsapplicability to operational air quality activities.

Application of random forest regression to the calculation of gas-phase chemistry within the GEOS-Chem chemistry model v10