Two-way ANOVA

I have data from two seasons and eight sites that I want to analyse using a two-way ANOVA. The problem I have is that the data do not have equal variances, or even close to. I know that two-way ANOVAs are robust to slight deviations, but this is more than slight! Is there an equivalent of Welch's F (and if so, can I get it in SPSS?)? Otherwise, what is the best non-parametric test to use? I thought of Friedman's, but that loses all of the variation, as I have to use means.

Thanks in advance for the help!
You might want to try some transformations on your data before switching over to Nonparametric statistics. A square root transformation may be useful for stabilizing variances.
Last edited:
Sorry, forgot to mention that, have tried transforming (square root, log and reciprocal, just for luck!) them with no joy.