indexing should now be correct. Intermediate dummy variables are not allowed by JAGS apparently, so the code is still difficult to read, but should work.
I added the identifiable RE and intercept terms, now Estar and Bstar in the model. The mean of the REs is subtracted from each RE and added to the intercept.
Due to small group size (10 stations), I changed the prior from a tau.eps ~ dgamma(0.01, 0.01) to a sig.eps ~ dt(0, 10, 2). The folded t is recommended by Gelman (2006) as weakly informative and appropriate for hierarchical model variance terms.
EstarandBstarin the model. The mean of the REs is subtracted from each RE and added to the intercept.