Hey Karina,
Because the error pattern is unique to each run, it's recommended to denoise each run separately and then merge your feature tables.
This only works if the exact same region is sequenced on both runs and the same trimming settings are used! (See this forum thread and GitHub issue.)
After dada2, run
merge: Combine multiple tables — QIIME 2 2022.8.3 documentation and
merge-seqs: Combine collections of feature sequences — QIIME 2 2022.8.3 documentation
then continue on to tax classification and diversity analysis.
(When I merge Illumina runs, I add a column to my metadata like runNumber so that I test for run-to-run variation after I merge.)