Return of the original scale in a prediction function

Question

I created a simple function to perform prediction of tree heights through regression models built for this purpose (symmetric relation models). The function predicts the heights, and inserts them in a new column in the date.frame by merging them with the heights used to adjust the model, but some models of the symmetric ratio change the original scale of the heights (response variable)such as models that use log. Would have some way to make the function predict ( for example), return the values in the original scale, identifying in the model the scale change factor?

The function is as follows :

pp <-           
  function(model,data)
  { 
{
  vetor <- data[is.na(data$AltMed),]        
  vetor <- predict (model,vetor)    

  data$ALTMISTA <- data$AltMed          
  data$ALTMISTA[is.na(data$ALTMISTA)] <- vetor
}

return (data)       
}

The models I’m encountering are as follows::

modelo 1 = lm(log(Altura) ~ I(1/Dap) + log(AlturaD), data=data)

modelo 2 = nls(I(Altura/AlturaD) ~ a1*(1-a2*exp(-a3*I(Dap/Q))), data=data, start=list(a1=3.348296, a2=-11.675862, a3=1.88852))

These models change the scale of the response variable.