Building sets of continuous and ordinal neutral landscape models matching real-world value distributions and correlations
Abstract
A neutral landscape model (NLM) creates a virtual landscape that mimics a real-world landscape sufficiently for landscape ecologists to explore the effects of spatial patterns on ecological processes without the restrictions of real-world experimentation. While a range of NLMs have been developed, little has been presented in terms of creating NLMs with continuous or ordinal values that more closely mimic real-world landscapes. This is important for more applied questions there is also a need to produce NLMs that mimic real landscape values when there are questions about how changes in real landscape patterns may affect real landscape processes. Through the use of three key underlying methods (i) rank rescaling, (ii) blending, and (iii) Kendall’s rank correlation matrix, a workflow is presented that is capable of building a set of continuous and ordinal NLMs that match real-world value distributions and rank correlations. The accuracy and precision of the workflow is demonstrated using an example for a real-world landscape in New Zealand. Accuracy is shown to be high, but precision varies as a function of landscape heterogeneity, with less precise results with more homogenous NLMs. The developed workflow is implemented in the open source NLMpy Python software package, but as the individual methods and hence combined workflow rely solely on commonly used numeric and statistical procedures, these methods and workflow could be easily reimplemented in other programming languages and software.