why is r square value low when data is differenced

[ray2013]

i have this table from an analysis i did on my data in eviews

model
1.AR(2)
AIC =13.57141
SIC = 13.61367
r-square=0.910038
DW= 2.023601

2.ARIMA(3,1,2)
AIC =13.52903
SIC = 13.59987
r-square=0.159654
DW= 1.954061

3.D(prod,1,12) ar(1) sma(12)
AIC =13.36215
SIC = 13.39133
r-square=0.619516
DW=1.995781

question 1. when comparing this 3 models can i pick AR(2) over the other 2 models because it has the highest r-square value eventhough the others had small AIC and SIC value compared to it.

question 2. why is it that when you difference your data and you fit a model to it, the Rsquare value becomes too low. how can this be explained

eg ARIMA(3,1,2) has a very small rsquare value when the data was difference compared to AR(2)