# Analysis of spatial data over time and space

I have a data set having year-wise monthly average of minimum and maximum temperatures of 32 stations around the country since 1948. The latitude and longitude of the stations are given as well. I have been asked to dig deep into the data set and analyze the much I can. The main objective is to understand the trend and possibly predict the future values.

I have little experience to work with such data sets. I usually like to use R. I have three questions in particular.

1) I can see that the latitude and longitude of an station are given to be 25.72 and 89.26 respectively. The corresponding Y and X coordinates are calculated as 844822.3 and 426172.9. I don't understand how they have calculated this. Can anyone explain?

2) What model should I fit to the data to get a good prediction of future values? Since I have the data since 1948, I might use a time series model, but here I have the latitude and longitude information too. So is there any model that can use these information too? Is there any particular package in R that will be useful for me?

3) Can I use R to find interesting plots mapping my findings or different summary statistics in map of the country? Some 2D or 3D plots will be excellent for me.

• I am also quite concerned about the over-fitting problem that might occur by using the latitude and longitude as regressors. – Blain Waan Jul 4 '14 at 9:29
• For 1 the coordinates are likely in some type of projection. The answer to 3 is generically yes. 2 is difficult to answer in this format, IMO the best anyone could do is give general reference to the lit. – Andy W Jul 4 '14 at 12:07
• There is far too much here to address. To get a handle on these issues you will need to learn about representing geographic data (projections and coordinate systems); and then about spatio-temporal data analysis, which is a kind of marriage of time series analysis (which you need to know very well) and spatial data analysis (geostatistics); and have a deep understanding of climate models. Those things will take anybody years to learn and more years to learn well, but if you absolutely must do it yourself, take a look at Cressie & Wikle, Statistics for Spatio-Temporal Data (J. Wiley, 2011). – whuber Jul 4 '14 at 15:48
• Thank you so much for the reference. I'll try to learn it and yes I understand that it will take time. But thanks for you hints. – Blain Waan Jul 4 '14 at 19:02
• Also, look at the spacetime package in R! – Subir Mitra Mar 26 '16 at 13:36

That´s a big task for one person. I can suggest you only some books related to 1) and 3). If you want to analyse your data in R a good reference is Applied Spatial Data Analysis with R and thesp package. On the book´s webpage you can find sample data and the code used thorughout the book. Another books are:
Spatial Data Analysis in Ecology and Agriculture using R, a free book A Practical Guide to Geostatistical Mapping, or Displaying time series, spatial and space-time data with R. If you will work with rasters in R, than there is a raster package . Or you can do it in GIS software, for example ArcGIS or SAGA GIS, and check the GIS Stackexchange