Hi Pham,
I used ARMA model to forecast precipitation (yearly) and consequently river inflow. but this model considers a stationary conditions in variable time series that is not matchs with climate change concepts.a more complicated techniques is SSA(singular spectrum analysis),artificial nural network,wave theory and ... . also you can see USGS hydroCLIMATE software.

Thank you so much for your suggestion
I would like to make sure something.
if I use ARMA model, I just got yearly data, right?
But I need monthly data to input into WEAP model.
Because my study area is about 2200 km2. So, After use some kind of model to project precipitation and temperature, I also have to downscalling.
Pham Quoc

Dear Pham,
ARMA result period depend on your input period, if you develop a monthly model, your model results are in monthly scale, thus you have no problem with time scale. also WEAP have flexibility in time steps (data view,general,years and time steps). on the other hand ARMA,ARIMA,SSA,... models are temporal models, not spatially distributed models.

Thank you so much Majid Vali !