How do I use statistical weights when combining datasets?

I am taking three years of a survey and pooling the observations into a single dataset. The survey is based on complex sampling procedures (e.g. clustering and stratification) and requires the use of weights to accurately estimate the sizes of the standard errors when doing simple statistics or using regression techniques. Are the sampling probability weights for each year still valid when the observations are combined with other years? Do weights have to be recalculated somehow to reflect the composition of the new, pooled dataset?