What shape the smooth is built from. tp is the default thin-plate spline and
suits almost everything. ts is the same with an extra shrinkage penalty, so
a term that earns nothing can be shrunk out of the model altogether rather
than left wiggling at one degree of freedom - the per-smooth equivalent of
select_features. cr is a cheaper cubic regression spline, worth it on a
large grid. cc is cyclic, which is the one that matters here: day of year
should join up at the end of December rather than being free to jump.
References
Marra G, Wood SN (2011). Practical variable selection for generalized
additive models. Computational Statistics & Data Analysis 55(7),
2372-2387. doi:10.1016/j.csda.2011.02.004
— the shrinkage penalty behind
ts, and behind select_features
Wood SN (2017). Generalized Additive Models: An Introduction with R, 2nd edition. Chapman and Hall/CRC. doi:10.1201/9781315370279 — every basis listed here