Monthly Archives: October 2010

Species response curves in JUICE and R

Today I spent some time in attempt to fix the function of Species response curves in JUICE. The script for calculation of species response curves is written in R and it has problems from the whole beginning, but as my interest in using species response curves decreased, the same decreased my will for fixing the bugs. To make it short – I haven’t manage to fix it, and there are several reasons for this.

One is package gravy, which is used for calculation of HOF curves. Jari Oksanen has written this package almost 6 years ago (at least the newest version is from the year 2004) and didn’t updated it since that – and in the email two years ago he said he is not going to mantain it any more. This wouldn’t be so big trouble, but since R version 2.11 this packages no longer runs, which makes it imposible to run the script for HOF curves in newer versions of R. I tried to recompile gravy for newer versions of R, but didn’t succeed (mainly because the source code for gravy.dll dynamic library is not publically available, and I didn’t want to bother Oksanen about that).

Anyway, gravy didn’t really perform too good. The algorithm often results into ‘sharp shapes’ (which I documented in details here). I don’t really have ability to fix this, as this goes beyond my mathematical skills. However, seems that Florian Jansen was working this topic through and found some remedy – he reported it two years ago in IAVS in Stellenbosch (presentation could be found here) and he has also written R package, which contains updated version of HOF functions. The packages is rather new and under development (I have seen the latest update from August 2010), so it still needs perhaps some time until it will be reliable. But this would be option to go. The packages could be downloaded from Florian’s website (basic package is vegdata, and the package contains additional function, which are under development – such as those HOF curves).

So, the recent situation is like this: the function about species response curves in JUICE works fine only under R version 2.8 or lower, and it’s perhaps not going to change in recent time. In the future, I hope to remake it from gravy to vegdata package. For those seriously thinking about playing with species response curves, I would recommend use R directly, not the wrapper from JUICE; if you are not R experienced, try to go for CanodDraw for Windows, which offeres GAM and GLM models for species response curves (however, it doesn’t offer HOF). That’s it!

UPDATE (September 2012): There is an alternative, which works right now – I rewrote the original R package for JUICE into JUICE-R function, which can be run with data from JUICE, but it’s not so problematic like the previous solution. It can be found on JUICE-R website in the section Available R scripts as Species response curves. It’s based, however, on the same algorithm as the previous function in JUICE, so the problem with HOF models remains and I would not really recommend to use them…