Forecasting of sky images

Hello everyone. I'm designing a system to forecast the availability of 
solar radiation for photovoltaic power stations. The forecasting system 
will use regularly-taken images of the sky, each of which is a full grid 
of binary values (each element indicates cloudy or clear-sky conditions 
for roughly a 40m x 40m area). These (PNG) images are captured every 
minute using a "sky camera". I'm looking to use R to predict the images 
(i.e. the movement of clouds) in an automated fashion for a short time 
period ahead (up to 30 minutes) by utilising trends in the sequence of 
images. Does anyone know of a way to use R to undertake the 
extrapolation of such images please?

There are a number of predictive approaches in the meteorological 
literature (where such methods are included under the term "nowcasting") 
but it's not clear how the algorithms could be used in R without a 
substantial amount of programming. There is also the issue of a lack of 
statistical rigour in the methods. So I'm looking for modelling tools 
(from spatial statistics?) which will allow the statistical forecasting 
of images and which have been implemented in R. I've searched through 
the available spatial statistics packages such as "spacetime" and "geoR" 
but nothing immediately presents as suitable (although I've considered 
using localised spatial logistic regression employing the 
temporally-lagged values from surrounding mosaic elements as predictors).

Many thanks for any suggestions

Ross Bowden
PhD Candidate, Murdoch University, Western Australia.