Does scale matter in predicting species distributions? case study with the Marbled Murrelet

Area of Science:

• Wildlife ecology and habitat modeling
• Spatial analysis and species distribution

Background:

• Hierarchical selection orders (microsite, patch, home range, population block, geographic range) are crucial for defining spatial scales in habitat models.
• Traditional conditional models based on these orders are often inadequate for generating accurate predictive habitat maps.
• Existing single-grain models may achieve high accuracy if they capture the most limiting spatial scale.

Purpose of the Study:

• To propose and evaluate a novel two-step approach for accurately mapping animal use probability.
• To assess the utility of this approach in elucidating scale-dependent selection effects for wildlife.
• To compare the performance of the two-step multi-grain model against traditional single-step models.

Main Methods:

• A two-step modeling strategy was developed: Step 1 involved single-grain analysis for each selection order, and Step 2 combined key variables into a multi-grain model.
• The Marbled Murrelet (Brachyramphus marmoratus) dataset was used as a case study, analyzing five hierarchical selection orders.
• Information theory criteria were employed to compare model performance.

Main Results:

• The two-step multi-grain models demonstrated superiority over simpler one-step single-grain models for the Marbled Murrelet, based on information theory.
• The proposed models effectively allow for the interpretation of variable selection at different spatial scales.
• Classification accuracy of the two-step multi-grain model was comparable to traditional one-step multi-grain models that do not account for selection orders.

Conclusions:

• The two-step multi-grain modeling approach is strongly recommended for its ability to reveal scale-specific habitat selection patterns.
• While not necessarily enhancing predictive map accuracy, this method provides valuable insights into the scaling effects of habitat variables.
• The primary advantage of the two-step method lies in its explanatory power regarding scale, rather than solely in improving predictive performance.