“…Indeed, our training set contained only two congeners with a −CH 2 –CH 2 –CH 2 –CH 2 – fragment and none with a longer CH 2 chain. That said, the experimental K aw value for 1,2,9,10-C 10 Cl 4 could also be somewhat too high, as log K aw of -2 in combination with log K ow of 5 would result in log K oa of 7 (via K oa = K ow / K aw ), and this is even smaller than an experimentally measured log hexadecane–air partition coefficient ( L ) of 8.4 for 1,2,9,10-C 10 Cl 4 …”