1. derive the input: globale mean delta (TS) corresponding to 108 qdp 

2. derive the ouput: qdp at 108 loctions (TS)

DeepONET: qdp = G1( delta(TS)) 

CNN:      qdp = G2( delta(TS))

training:  108, 97 training, 11 as testing 

extra step:  97 training x ? (increse smaple size)

derive global mean delta(TS) from climate model (furture - current)
G1(delta(Ts)) = qdp 

qdp(reference) = from climate model (furture - current)