Acid Sulfate Soils Classification and Prediction from Environmental Covariates using Extreme Learning Machines
Atsemegiorgis, Tamirat (2023)
Atsemegiorgis, Tamirat
2023
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2023053116799
https://urn.fi/URN:NBN:fi:amk-2023053116799
Tiivistelmä
Acid Sulfate Soil (ASS) mainly occurs because of naturally occurring phenomena. It is a sulfate-bearing sediment found in coastal areas around the globe. The highest concentration of Europe’s ASS is in Finland. It has been noted that ASS is a major environmental problem in Finland. Since the 1950s, researchers have been engaging in ASS local mapping projects, manually measuring the soil’s PH level to determine whether the sample soil is acid or not. This method is very laborious and time-consuming. In the last decade, GTK, with collaborators, has been working on systematically mapping ASS to alleviate the problem and make it accessible to the public. These days, Artificial Intelligence (AI) and Machine Learning (ML) methods are used widely to classify soil as ASS or not. This research thesis’s motivation is to assist in completing the compressive ASS digital map with high accuracy. The aim is to explore the performance of the Extreme Learning Machine (ELM) in an acid-sulfate soil classification task. The research database comes from Finland’s west coast region, containing point observations and environmental covariates datasets. The experimental results show the overall accuracy of ELM and Random Forest models is the same. However, ELM implementation is easy, fast, and requires minimal human intervention compared to conventional ML methods like Random Forest.