What are the error messages shown? Theoretically, as long as the same primers are used, there shouldn't be any problem merging the two.

However, in fact you shouldn't use dada2 for denoising the combined dataset since each sequencing run has its own error model. A proper way to do this is to denoise each dataset and then combine the results. Or, you can simply use deblur to denoise the combined dataset.

Agreed with Man Kit Cheung. I am not very familiar with QIIME2 (although I plan to change that), but if you are still struggling with QIIME for this phase, an alternative could be to do this using Mothur, and then put the combined dataset back into QIIME to continue the curation.

If I may ask, since you are going to compare different datasets that were sequenced with different technologies (or with different amplified regions e.g. V3V4 with V4V5 or V4-V6), are you planning to make ASVs? or are you going to align and then using distancing to generate zOTUs or 0.03 OTUs?

Sorry, normally when I hear dada2 in a conversation, people are usually focusing on ASVs, hence my question.

As already mentioned by Man Kit Cheung, one can denoise each dataset separately by dada2 and then merge output files (feature table and rep-seqs), or use Deblur with combined dataset.

Just wanted to add that if you will decide to use dada2, each dataset should be treated with exactly the same parameters (such as trimming, truncation...).