spatial_predict: Spatial MBM prediction
In mtalluto/mbm: Multifaceted Biodiversity Modelling

Description Usage Arguments Details Value

Spatial MBM prediction

1	spatial_predict(x, prdat, coords, method = c("slow", "fast"), ...)

`x`	A previously-fit MBM object
`prdat`	New dataset to be used for prediction; either a raster stack or data frame. See 'Details'
`coords`	matrix with 2 columns containing X-Y coordinates for `prdat`, required if prdat does not have a `coordinates` method.
`method`	How to compute the spatial predictions; see 'Details'
`...`	Other named parameters to pass to `predict.mbm`.

prdat can either be a raster stack with new variables (and spatial information) for prediction, or a data frame-like object with previous predictions from predict.mbm with 4 columns: 1. site1, 2. site2, 3. mean, and 4. sd.

For rasters, if a layer named 'names' is included (recommended), this layer will be used as sitenames, otherwise they will be assigned unique numbers.

If method is "slow", spatial predictions will be computed by first predicting dissimilarity to all pairs of raster cells, then performing an ordination on the dissimilarity matrix to produce an RGB raster of spatial predictions.

For method == 'fast' (currently not implemented), computation is sped up by first performing hierarchical clustering on the predicted dissimilarity matrix for the calibration data (which will have already been computed when mbm was run) to produce cell categories. Each raster cell will then be assigned the category of the calibration data point that is closest environmentally. Then, we compute the dissimilarity matrix of the categories (based on the mean environmental values). The ordination is performed as with the slow method on this dissimilarity matrix.

An object of class mbmSP, which is a list with three named items: fits is a 3-band gridded SpatialPointsDataFrame giving the first three prinipal components of predicted pairwise dissimilarity, stdev is a SpatialPointsDataFrame giving the mean of pairwise dissimilarities among all other sites in a given site, and pcoa is the principal coordinates analysis for the fits. Both fits and stdev can be made into rasters using raster::stack() and raster::raster().

mtalluto/mbm documentation built on Aug. 13, 2019, 9:43 a.m.