Capabilities compared to other SDM packages
Species distribution modelling (SDM) approaches have been around for
quite some while and as a result a number of ecological modelling
focused packages have been developed. In general most of them are
customized towards specific purposes and modelling paradigm. So isn’t
this just another SDM package? Indeed it is, but
has a number of particular features that set it apart from most other
- It focuses particularly on integration as a guiding principle and different ways of how heterogeneous sources of evidence can be integrated.
- It puts a strong focus on biodiversity types, in particular Poisson-Process models (PPMs) as the default way of analyzing presence-only data.
- It follows an object-based and modular programming philosophy,
taking inspiration from
- It supports a number of Bayesian SDM approaches and algorithms, a field traditionally less represented owing to computational constraints.
- It is customized to create and modify spatio-temporal scenarios, including IIASA integrated land-use assessment model GLOBIOM.
Thus overall, the idea package is in part trying to bring some innovation to the SDM modelling world, while also trying to bring together strengths from different existing tools.
A Non exhaustive list acknowledging other SDM packages in R and how
they compare to
ibis.iSDM is provided here:
- hSDM -> Bayesian framework for hierachical and mixed models. Fast, but little flexibility with regards to weights, offsets and different datatypes.
- multispeciesPP -> Package that allows integrated SDMs, however it has not been further developed since a few years and key gaps remain particular with regards to different modelling approaches.
- inlabru -> Package specifically for Lox-Gaussian-Cox Process (LGCP) models with INLA, now integrated also as engine in ibis.iSDM
- pointedSDMs -> Another wrapper for INLA that allows to integrate different datasets in a SDM. Less focus on priors, offsets and scenarios.
- biomod2 -> Popular package for ensemble modelling, but fixed on specific (non-Bayesian) engines and data types and with no integration options.
- sdmTMB -> Package for fitting spatial-Temporal SDMs for very specific biodiversity data.
- modleR -> similar as biomod2 and a wrapper to construct ensembles of models.
- kuenm -> Another wrapper for Maxent.
- flexSDM Similar as biomod2 a wrapper for SDMs, but coming with several helper functions for data preparation and cross-validation.
Besides SDMs there are also new packages available on spatial and integrated species occupancy models, such as spOccupancy. Occupancy modelling however requires some very specific biodiversity data and information to infer detectability of species occurrences.