Choice of predictors and complexity for ecosystem distribution models: effects on performance and transferability

• DOI: 10.1111/ecog.07269
• Publication Date: 2024-06-10T12:05:47Z
• Authors: Adam Eindride Naas; Lasse Torben Keetz; Rune Halvorsen; Peter Horvath; Ida Marielle Mienna; Trond Simensen; Anders Bryn
• Keywords: Transferability

There is an increasing need for ecosystem‐level distribution models (EDMs) and a better understanding of which factors affect their quality. We investigated how the performance transferability EDMs are influenced by 1) choice predictors 2) model complexity. modelled 15 pre‐classified ecosystem types in Norway using 252 gridded to 100 × m resolution. The major ‘Nature Norway' system mainly defined rule‐based criteria such as whether soil or specific functional groups (e.g. trees) present. were categorised into four groups, three represented proxies natural, anthropogenic, terrain processes (‘ecological predictors') one spectral structural characteristics surface observable from above (‘surface predictors'). Models generated five levels Model evaluated with data collected independently training data. found that trained only performed considerably more transferable than ecological predictors, increased complexity, levelling off approximately 10 parameters reaching peak at 20 parameters, while decreased Our findings suggest enhance EDM transferability, most likely because they represent discernible types. A poor match between define processes, plausible explanation why predict results indicate that, cases, same not well suited contrasting purposes, predicting where ecosystems explaining there.