Methods in Ecology and Evolution (Aug 2024)

Geographic principles applied to population dynamics: A spatially interpolated integrated population model

  • Brian G. Prochazka,
  • Peter S. Coates,
  • Shawn T. O'Neil,
  • Shawn P. Espinosa,
  • Cameron L. Aldridge

DOI
https://doi.org/10.1111/2041-210X.14334
Journal volume & issue
Vol. 15, no. 8
pp. 1394 – 1407

Abstract

Read online

Abstract A major impediment to wildlife conservation and management, from a quantitative perspective, is dealing with high degrees of uncertainty associated with population estimates. Integrated population models (IPMs) can help alleviate that challenge, but they are often limited to narrow spatial or temporal windows owing to the financial and logistical burdens of acquiring requisite datasets. To expand the spatiotemporal scope of practical IPM implementation, we developed a novel method that expresses demographic relatedness among sampled and unsampled locations using geographic principles of spatial autocorrelation. We interpolated demographic parameters at unsampled locations using parameter estimates from data‐informed locations. Errors attributable to the interpolative process were corrected using a joint likelihood and locally recorded count data (‘cheaper’ and broadly distributed). We evaluated the spatially interpolated IPM (SIIPM) for precision and accuracy under variable levels of spatial autocorrelation using simulated data and a Leave‐One‐Out Cross‐Validation (LOOCV) technique. Conventional IPMs and state‐space models (SSM) were fit to the same simulated datasets to provide a comparative assessment of the novel method. In a final, empirical demonstration we fit the SIIPM to data collected from Greater Sage‐Grouse (Centrocercus urophasianus; sage‐grouse) populations located in Nevada, U.S.A. during 2013–2021. SIIPMs outperformed conventional IPMs when fit to data possessing moderate‐to‐high levels of spatial autocorrelation. Under moderate levels of autocorrelation, the average improvement in parameter estimation was 13.6% for survival, 65.3% for recruitment and 23.7% for rate of population change (λ). When spatial autocorrelation was low, the SIIPM still outperformed contemporary approaches in areas that were geographically close (<67 km) to sampling locations. Under low autocorrelation‐near distance scenarios, we observed SIIPM parameters that were 30.8% (recruitment), 32.5% (λ; IPM comparison) and 54.0% (λ; SSM comparison) more precise than contemporary models. Spatial autocorrelation is often assumed but rarely tested when comparing population dynamics across regions of large geographic extent. We demonstrated that SIIPMs can improve precision of species' vital rate estimation when extrapolating model inference beyond populations for which long‐term monitoring data exists. Specific to sage‐grouse, these results support previous conclusions of broad‐scale spatial autocorrelation in population dynamics and a reproductive‐survival trade‐off previously documented at smaller scales.

Keywords