Combining indicators for better decisions – Algorithms vs experts on lakes ecological status assessment
The results of ecological condition assessments of ecosystems are related to key decisions taken for the purpose of remedial measures or maintaining their current state. In the assessment process, experts come across extensive datasets, the quality, and completeness of which do not always allow for...
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
Elsevier
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/5c4f1588eb54443994b48fc004934d93 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | The results of ecological condition assessments of ecosystems are related to key decisions taken for the purpose of remedial measures or maintaining their current state. In the assessment process, experts come across extensive datasets, the quality, and completeness of which do not always allow for a reliable evaluation, especially if a single empirical approach is used. In this paper, results of machine learning algorithms are presented, with a focus on Self-Organizing Maps. In this context, measurements of component parameters for the assessment of the ecological state of lake ecosystems were subjected to the process of unsupervised machine learning with the aim to create an alternative assessment approach based on the capabilities of neural networks. Results are mapped and compared with expert evaluations, allowing to extend knowledge about sub-clusters present in the data. The primary target of this paper is the ecological assessment expert. At an early stage, information was obtained about the presence of ecological outliers that may be subject to separate monitoring or verification of environmental activities and objectives. In the back-mapping process, the presented technique of map construction and clustering with various versions of the division was referenced to a set of expert classification findings, revealing the underlying structure of the results when addressed with an unsupervised data comprehension. The approach introduced here does not intend to interfere with the format of an original assessment methodology. Rather it aims at obtaining useful additional information which may help in making better decisions. |
|---|