[R] A comparison between span in lowess method and h in sm.regression

[FMH]

Dear All,

I have fitted a bivariate and an additive model on a monthly temperature from 1987-2007 via nonparametric regression and would like to compare between these models via F-test, but having difficulty in determining the possible values of span with lowess method.

For instance, these two models are defined in R as: 

######################################################
bivariate.model <- sm.regression(c(month,year), Temperature, h = c(0.2,0.8)) 
additive.model <- gam(Temperature~lo(month, span=A) + lo(year,span=B)) 
######################################################

However, i notice the sm.regression function uses h as a smoothing parameter  whilst the gam function uses span as a smoothing parameter, and these models can't just simply be compared straight  forward, but few adjustments should be made on the values of span in gam function, corresponding to the values of h in sm.regression function. 

Could anyone please advice me  the possible values of A and B in the gam function above? I'd would be very grateful for any idea and suggestion towards this issue.

Thank you
Fir