Investigating the Spatial and Temporal Variability in Whooping Crane Habitat Selection Preference in Kansas State

Keywords: Whooping crane, Habitat conservation, Habitat Assessment, Species distribution modeling, SDM, Machine Learning, Random Forest, Poisson point process distribution, Kansas
Abstract Type: Poster Abstract

Abstract

The whooping crane (Grus americana) is an endangered species facing critical migration challenges. This makes the identification of suitable stopover habitats necessary for its survival and recovery. This research focuses on modeling habitat suitability within the Kansas portion of the Aransas-Wood Buffalo population’s migratory corridor, an area impacted by wind energy developments. We build upon recent research using machine learning algorithms to create species distribution models (SDM) for this migratory avian species.

We developed an SDM for the whooping crane using a random-forest technique to inform management decisions within Kansas. The model uses six predictor variables for our observation data to estimate the relationship between environmental characteristics and stopover site suitability and telemetry data from 58 tagged birds between 2010–2016 (n = 1253). In similar studies, presence-only data is commonly used by randomly generating the background data (“pseudo-absence” points), but this approach has several weaknesses. This study utilized a Poisson point process for generating 2000 spatially weighted pseudo-absence points for the model.

The result was a prediction map showing areas where whooping cranes would most likely be found. Relative probability function ranged from 0.007 to 0.82, with a prediction accuracy of 92%. The prediction map showed that 75.07% of the study area was unsuitable and posed low conflict for energy development projects. In comparison, 24.93% posed mid-high-level conflict and should be designated critical habitats for conservation.

This research supports whooping crane recovery while contributing to the body of knowledge on species distribution modeling techniques to further wildlife conservation efforts.