Air pollution is the main environmental health issues that affects all the regions and causes millions premature deaths every year. In order to take any preventive measures, we need the ability to predict pollution level and air quality. This task is conventionally solved using deterministic models. However, those models fail to capture complex non-linear dependencies in erratic data. Lately machine learning models gained popularity as a very promising alternative to deterministic models. The purpose of this thesis is to conduct a comparative study between ChemicalTransport Model (WRF-Chem) and a Statistical Model built from machine learning algorithms in order to understand which one is advantageous predicting the air quality and the meteorological conditions using data from Cuenca, Ecuador. The study aims to compare the two methods and conclude on which of them is better in forecasting the concentration of fine particulate matter (PM2.5) in an urban complex terrain. I concluded that even though WRF-Chem has the biggest advantage of forecasting all the data of interest for broader time horizon machine learning algorithms provide better accuracy for middle-term period. Machine learning models also require much less computational power but lack ability to predict meteorological conditions along with pollution level.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:du-34435 |
| Date | January 2020 |
| Creators | Kudryashov, Andrey |
| Publisher | Högskolan Dalarna, Mikrodataanalys |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.002 seconds