The main files for the data are 5873inv.txt - list station numbers, lat/lon/elev, station name 12 files stnx1.txt-stnx12.txt - each row contains one month's data for one station. -1 is used as missing value code (note this includes, for example, always giving the missing value code to September 31, but there are also a lot of genuinely missing values in these datasets). The way I use these files myself is to run a Fortran program "decode" that expands the station files into 5734 individual station files - that takes up about 3GB but is a convenient format for direct loading of individual data series into Fortran or R (or any other programming language that reads data directly from ascii files). I have included a compliled version of "decode" for Windows (if you copy this, need to copy the "salflibc.dll" program as well - it's based on the free online version of the Salford systems Fortran compiler) but I have also included the program code (decode.for) in case you want to compile it on unix or linux. The way you use this is very simple: in Windows just open a command window and type "decode". The program then creates all the new data files (takes 5-10 minutes on my machine).