This is one of those things that seems fairly obvious once pointed out: that the energy in light might be sufficient to directly break hydrogen bonds and separate water molecules from surfaces, especially in regions where water's absorption is low. This is a separate process from thermal gain (from any source, including light) heating and phase-changing the water.
I don't know if this is right but I'll put something forth to sanity check it:
0) say the energy required to break water's hydrogen bonds is 6.3 kJ/mol (random quora source)
1) 6.4 kJ/mol / 6.022 x 10^23 molecules/mol = 1.0461641 x 10^-20 J per bond
2) 1.0461641 x 10^-20 J * 6.242 x10^18 eV/J = 0.0653 eV per bond (convenient units for light)
3) widest band of visible light energies I found (per photon): red to violet is 1.63 to 3.26 eV
4) if this holds up then each violet photon could potentially break ~25 H bonds and still remain visible
5) this process might be detectable in visible light itself as a red shift, or maybe not because the red also shifts down (or because various frequencies participate more or less for reasons I don’t know about)