Sample the original data with replacement and fit LOWESS curves accordingly.
boot.lowess(x, y = NULL, f = 2/3, iter = 3, line.col = "#FF000033", ...)
Yihui Xie
We keep on resampling the data and finally we will see several bootstrapped LOWESS curves, which may give us a rough idea about a ``confidence interval'' of the LOWESS fit.
Examples at https://yihui.org/animation/example/boot-lowess/