Cokriging interpolation problem

Discussion created by caroline_qi on Nov 8, 2010
Latest reply on Nov 8, 2010 by EKrause-esristaff
I use monthly station temperature and elevation to interpolate U.S. national temperature. I want to create prediction map with Ordinary kriging. After the trend removal, I need to select the model of semivariagram. I find the cross-validation is better if I choose model 1(Exponential) + model 2(Exponetial) than only selecting model 1 (Exponential). What that means? Does it make sense if the the two models are same? I'm still confused about the model1, model 2 and model3.  I cannot find some exact exmple for multi-model selecton. Look forward to your response.