Post provided by Sarah Goslee
For an ecologist interested in long-term dynamics, one of the most thrilling experiences is discovering a legacy dataset stashed away somewhere.
For an ecologist interested in long-term dynamics, one of the most daunting experiences is figuring how to turn that box full of paper into usable data.
The new tool HistMapR, described in ’HistMapR: Rapid digitization of historical land-use maps in R’ by Alistair Auffret and colleagues, makes one part of that task much easier.
Historical maps with coloured areas denoting different land cover or use are a valuable record, but difficult to analyse. This R package automates much of the time-consuming and tedious process of turning paper maps into classified categorical raster maps.
A map is scanned, imported into R, and the software is trained by clicking in different areas of each category. It then automatically classifies pixels based on which colour they are most similar to. The resulting classification is assessed manually. The process can be repeated with slightly different parameters until a good fit is achieved.
The authors found 80-90% agreement between HistMapR classification and manual digitisation (sources of error included clarity of original maps and scan quality). Using HistMapR reduced the time needed for digitising a series of historical land cover maps from two months to two days. Ecologists interested in long-term dynamics should be cheering!
The HistMapR package is available on GitHub and you can find example scripts on Figshare, so you can get right to work.
‘HistMapR: Rapid digitization of historical land-use maps in R‘ by Auffret et al. is a freely available Applications article (no subscription required).