Validation of AQ maps produced internally by EEA using machine learning and their comparison with reference RIMM maps

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00020699%3A_____%2F24%3AN0000168" target="_blank" >RIV/00020699:_____/24:N0000168 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Validation of AQ maps produced internally by EEA using machine learning and their comparison with reference RIMM maps

Popis výsledku v původním jazyce

The report validates and examines the Machine Learning (ML) mapping results produced by the European Environmental Agency (EEA) and compares them with the reference RIMM mapping results. Two EEA ML variants have been examined, i.e. those trained using the data up to 2020 and up to 2022. The validation has been performed against a validation set of the stations, which was not included in the mapping. The analysis has been carried out for PM10, PM2.5 and NO2 annual averages and for O3 indicator SOMO35, for the years 2021 and 2022. The validation results show that overall the EEA ML results are somewhat overestimated, especially in the urban, but also in the rural areas, for all pollutants. The RIMM maps provide more accurate results compared to the EEA ML ones, for both rural and the urban background area types. As expected, the ML results trained up to 2022 show better performance, compared to the ML results trained up to 2020 only.

Název v anglickém jazyce

Validation of AQ maps produced internally by EEA using machine learning and their comparison with reference RIMM maps

Popis výsledku anglicky

The report validates and examines the Machine Learning (ML) mapping results produced by the European Environmental Agency (EEA) and compares them with the reference RIMM mapping results. Two EEA ML variants have been examined, i.e. those trained using the data up to 2020 and up to 2022. The validation has been performed against a validation set of the stations, which was not included in the mapping. The analysis has been carried out for PM10, PM2.5 and NO2 annual averages and for O3 indicator SOMO35, for the years 2021 and 2022. The validation results show that overall the EEA ML results are somewhat overestimated, especially in the urban, but also in the rural areas, for all pollutants. The RIMM maps provide more accurate results compared to the EEA ML ones, for both rural and the urban background area types. As expected, the ML results trained up to 2022 show better performance, compared to the ML results trained up to 2020 only.

Druh

O - Ostatní výsledky

CEP obor

—

OECD FORD obor

10509 - Meteorology and atmospheric sciences

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů