how to impute missing monthly climate data using AMELIA package? I have been trying to impute future monthly climate data based on sets of downscaled data using the AMELIA package. I have monthly precipitation data from 1960-2099 for different geographic regions and I want to impute precipitation data from 2099-2150 using the past trend of precipitation. I want to use bound feature (i.e., max and min values) of different months of a year specific to the geographic region to impute monthly data from 2099-2150 but I can't find ways to do it. I have figured out how to set one bound (Max and min values) for a particular geographic region but can't figure out to set multiple bounds so that data are imputed using that bound for a given month. Is there any way to perform such task? 
I know there are several limitations behind using AMELIA package to impute climate data, but I was wondering if people have attempted to impute climate data using AMELIA package? I find none.
I will appreciate any suggestions/comments on this post. Thanks!
 A: Amelia is designed for filling sparse missing data, not forecasting. Forecasting 50 years based on 90 years of data is certainly not what was intended with this package.
In particular, the multiple imputation in Amelia uses data that is present for a given timestep to predict missing data for that timestep. In your example, this would be like using available humidity, pressure, and temperature data to predict missing precipitation data, but in your case, none of those variables are available for the period in question (2100-2150), so there is nothing on which you can base your imputation of precipitation.
What you probably want to do is forecast using something like an ARIMA model, which will give you a projection that takes account of trend and seasonal changes. However, this is a purely statistical analysis, and won't take into account any longer-term forcing changes over the following 50 years, as the variability may not be fully exhibited in the 90 years for which you do have data. That is, the variability of factors such as to greenhouse gas levels, or orbital forcing over the last 90 years is smaller relative to the long-term changes in those variables. It also won't take account of changes in internal states, such as increases or reductions in carbon sink uptake capacity, or lags in the effects of variable changes (such as the changes in greenhouse gasses).
Depending on how much data you need, you may be better off actually running a climate model a few times, and make an estimate from that. Also, there are studies that have been done much longer than the usual CMIP until-2100 scenario, for example Climate Change to the End of the Millennium by Timothy M. Lenton, and you may be able to use data from such studies.
