Yes this is possible for same locations with several years data and across locations with several years of data. You can also use best linear unbiased prediction (BLUP) and best linear unbiased estimator (BLUE) to combine your data.
You can also do the association analysis with individual years and then compare across for marker trait association (MTAs). The same will also apply to locations. The more number of years or location you have for the association of same phenotypic traits then the greater will be your confidence of identifying a putative or potential major Quantitative trait loci (QTLs) for your traits of interest.
I request you to have a look at following literature's
Genome-wide association study (GWAS) of carbon isotope ratio (δ13C) in diverse soybean [Glycine max (L.) Merr.] genotypes
Yes, of course is possible and of course you should do it. Specially for quantitative characters. As a result of this you may end up with QTLs that are expressed in one year and not in others. You should have a good year characterization in order to explain different QTLs for diferent years.