A machine learning emulator for Lagrangian particle dispersion model footprints: a case study using NAME

Abstract

Abstract. Lagrangian particle dispersion models (LPDMs) have been used extensively to calculate source-receptor relationships (“footprints”) for use in applications such as greenhouse gas (GHG) flux inversions. Because a single model simulation is required for each data point, LPDMs do not scale well to applications with large data sets such as flux inversions using satellite observations. Here, we develop a proof-of-concept machine learning emulator for LPDM footprints over a ~350 km by 230 km region around an observation point, and test it for a range of in situ measurement sites from around the world. As opposed to previous approaches to footprint approximation, it does not require the interpolation or smoothing of footprints produced by the LPDM. Instead, the footprint is emulated entirely from meteorological inputs. This is achieved by independently emulating the footprint magnitude at each grid cell in the domain using gradient-boosted regression trees (GBRTs) with a selection of meteorological variables as inputs. The emulator is trained based on footprints from the UK Met Office Numerical Atmospheric dispersion Modelling Environment (NAME) for 2014 and 2015, and the emulated footprints are evaluated against hourly NAME output from 2016 and 2020. When compared to CH4 concentration time series generated by NAME, we show that our emulator achieves a mean R-squared score of 0.69 across all sites investigated between 2016 and 2020. The emulator can predict a footprint in around 10 ms, compared to around 10 minutes for the 3D simulator. This simple and interpretable proof-of-concept emulator demonstrates the potential of machine learning for LPDM emulation.

A machine learning emulator for Lagrangian particle dispersion model footprints: a case study using NAME

“…Code used to train and evaluate the models is available as a free access repository at DOI https://doi.org/10.5281/zenodo.7254667 (Fillola, 2022b). Sample data to accompany the code, including the trained emulator for Mace Head and inputs/outputs to test it, can be found at DOI https://doi.org/10.5281/zenodo.7254330 (Fillola, 2022a). The NAME III v7.2 transport model is available from the UK Met Office under licence by contacting enquiries@metoffice.gov.uk.…”

Nitrate isotopes in catchment hydrology: Insights, ideas and implications for models

“…In the adjacent field of air quality and atmospheric pollution, machine learning methods were used inside an 'emulator' that mimics the behaviour of computationally demanding models, to find sources of pollution (Fillola et al, 2022). Many of the machine learning methods used for such models are not domain-specific and development of similar models for pollution in hydrological catchments could benefit from these existing models from other domains.…”

FootNet v1.0: Development of a machine learning emulator of atmospheric transport

“…Compared to STILT, the FootNet prediction has an IoU of 0.59 and the correlation is 0.72.There have been other methods developed to improve the efficiency of footprint calculations. For example,Roten et al (2021) uses nonlinear weighted averaging to interpolate footprints from locations near the receptors Fillola et al (2023). develops a similar footprint emulator based on gradient-boosted regression trees (GBRTs), at a coarse spatial resolution (20-30 km in mid-latitudes) and 10 grid cells around the measurement location.…”