The model consists of using four environmental parameters to estimate 3 categorical responses. The ideal is that these four parameters are independent, which does not happen because of the correlation between two of them. Is there any transfomation or model that predicts one of the 3 responses without ruling out any parameters?