Improving efficiency of a statistical analysis of complex ecological models, when using the statistical software R by parallelising tasks with Rmpi

As a consequence, it is therefore indispensable to parallelise parts of the code and run them on more than one processing unit. Therefore, the aim of this study is to show the ease of parallelisation within the statistical software R using the package Rmpi. We present the parallelisation of three different applications (optimisation, Bayesian calibration, sampling from distributions), using our complex ecosystem model LandscapeDNDC. We were able to run the Bayesian Calibration at a computing cluster using 24 CPU's in 11.8 days opposed to 236.7 days when using only one CPU. This is an acceleration of the evaluation process by a factor of approximately 20.