Many drought forecasting models use SPI values directly as the predictand ? Instead of that can we make a rainfall foresting model and then calculate the drought indices?
You can predict first the rainfall and then compute the SPI from the predicted rainfall, but I think it won't really change the skill at the end since SPI is just a sort of a calibrated standardization of the rainfall. So I think it won't really change the skill
I think these two methods have different results. As you know the SPI is based on fitting a distribution to monthly precipitation data. In the first method, for calculating the historical SPI, you use the distribution parameters estimated from historical precipitation data series. But in the second method, you calculate the SPI after forecasting precipitation, so you estimate the distribution parameters for the predicted series (or for a series consists of both historical and predicted series). Since, the forecasting has error, the distribution parameters of this new series may have difference with historical ones. Thus, this new SPI series may differ from the historical SPI.
As you know when we want to calculate SPI, we should know about the distribution of rainfall data, In my opinion for your comment, it is better to forecast rainfall for future period, then with analyzing the achieved data you can easily calculate SPI for the specific future period.
Also, I refer you to visit my website:
www.agrimetsoft.com
It has already several tools that they may be work for you, in this regard.