Ontologies are very helpful for which scientific discipline?

Biology (biotechnology and child disciplines) is probably the discipline with a more extensive use of ontologies, there were traditionally many taxonomies in this discipline which provides a good ground for the creation and use of ontologies.

To name a few: Physics tends to the unification instead of classification, ontologies are not that interesting there. Medicine, specially pharmacology and biomedicine have some interesting use cases for ontologies, but there is a big mass of knowledge in medicine that is more related with diagnosis, and thus it is more suitable for expert systems than ontologies. Geography can provide some interesting ground for ontologies, but mereologies are better suited, the confusion between mereologies and ontologies is usual and not good.

I think that its important to point out that most people will be heavily biased on this point.

Its also important to make the distinction betweeen Linked Open Data and complex ontologies.

Clinical medicine definitely benefits from rich ontologies (personal bias) eg. coded diagnoses enable people to easily classify patients by diagnoses affecting their noses, injuries or nasal injuries. SNOMED CT and http://www.opengalen.org/ are fantastic examples of rich ontologies / controlled medical vocabularies.

Also to extend @database_animal 's example of PubMed's MeSH terms a search for asthma can include subcategories such as "Asthma, Aspirin-Induced" "Asthma, Exercise-Induced" "Asthma, Occupational" "Status Asthmaticus" and their synonyms also.

UMLS is a good source of medical vocabularies of varying complexity, each with important applications