Crossing borders: associating information across different levels of habitat selection, from the individual to the species

• Autores: David Ferrer Ferrando
• Directores de la Tesis: Pelayo Acevedo Lavandera (dir. tes.), Javier Fernández López (dir. tes.)
• Lectura: En la Universidad de Castilla-La Mancha ( España ) en 2025
• Idioma: inglés
• Número de páginas: 300
• Títulos paralelos:
  • Traspasando fronteras: relacionando la información entre los distintos niveles de selección del hábitat, desde el individuo hasta la especie
• Enlaces
  • Tesis en acceso abierto en: RUIdeRA
• Resumen

  • Understanding how organisms select their habitat is a key question in ecology, with direct implications for biodiversity conservation and environmental management. Habitat selection determines how organisms interact with their environment and influences ecological patterns. However, this process is complex, as it is shaped by a diversity of environmental factors acting at different spatial and temporal scales, and mediated by organism-specific traits. To address this complexity, habitat selection is often conceptualized hierarchically, with each level representing a different habitat selection process. Thus, a level encompasses a range of spatial and temporal scales, and a unit of ecological organization (species, population, individual). While analyzing each level independently can yield relevant insights, a more complete understanding requires relate information across levels. This thesis adopts that perspective, focusing on the first three levels of habitat selection, represented by different organization units (species, population, and individual), to explore how information can be compared and transferred across them. By combining theoretical and methodological approaches, the work aims to advance a more integrated framework for studying organism-environment interactions in ecological and conservation contexts.

    Chapter 1 explored the relationship between species- and population-level habitat selection, assessing whether environmental suitability estimates derived from species distribution models (SDMs) can predict patterns of population abundance. Using occurrence data from the six wild felid species in Mexico, two modeling approaches, Maxent and distance to niche centroid (DNC), were employed to generate species-level suitability maps across the continental range. These predictions were then compared to relative abundance indices obtained through an extensive camera-trapping survey in a well-sampled region of central-western Mexico. The results showed that while both algorithms produced broadly similar suitability patterns, their ability to explain local variation in relative abundance differed among species and methods. Thus, performance varied according to species-specific characteristics and data availability. The observed inconsistencies suggest that both methodological factors (such as algorithm selection and input data quality) and ecological factors (such as species behavior or detectability) may influence the robustness of the relationship between levels. This chapter illustrates the potential value of using species-level models to infer population-level patterns in regions with limited monitoring capacity, while also emphasizing the need for caution due to the multiple sources of uncertainty inherent in this approach.

    Chapter 2 examined how both methodological and ecological factors affect the ability to relate species-level environmental suitability to population-level abundance. Specifically, it evaluated the role of the spatial scale of environmental predictors, used as a methodological factor (finer or coarser), and the spatial distribution pattern of the organism, used as an ecological factor (more aggregated or uniform). To address this, simulated and empirical data were used, with increased data quality. The simulation framework allows testing the different scenarios by varying both factors in a controlled manner. And the results are subsequently validated with real-world data from two passerine species with contrasting spatial patterns, in the Iberian Peninsula and North Africa. The results showed that adding finer-resolution variables improved the explanatory power of the suitability models on abundance patterns. An improvement in this aspect is also observed when working with aggregated organisms.

    Even an apparent interaction between organism distribution and the scales of the variables considered can be observed. Where aggregated organisms show more apparent improvements with the first scale improvements of the environmental variables. The chapter concludes that species-level suitability may partially reflect population-level abundance, but the degree of overlap between levels depends on both ecological traits and the methodological framework employed.

    Chapter 3 evaluated the explanatory and predictive inferences derived from habitat selection models based on different data sources and analytical approaches, each corresponding to a different hierarchical level (individual and population). The study focused on the three wild ungulate species inhabiting Doñana National Park, using simultaneous datasets: camera trap detections to represent population-level, and GPS telemetry to capture individual-level. Imperfect detection models (IDMs) were applied to the camera-trap data, while resource selection functions (RSFs) were used for the telemetry data. The comparison revealed a partial overlap in the environmental predictors identified as relevant by each model type, though notable discrepancies were also observed. On the predictive side, both approaches showed a certain degree of similarity in spatial patterns, but their agreement varied depending on the species and the spatial extent of the prediction. These findings suggest that both approaches are not equivalent but rather complementary, as they reflect different processes of habitat selection operating at distinct levels. Moreover, part of the observed differences may stem from the inherent characteristics of the data and analytical methods used, each emphasizing different aspects of organism-environment interaction. The chapter underscores the importance of aligning data type and modeling strategy with the ecological question of interest, especially when working across levels and organization units.

    Chapter 4 compared two strategies for combining information across individual and population levels of habitat selection: an integrated approach using a joint-likelihood model for both levels, and an independent approach based on separate models fitted independently at each level. The integrated approach allows both data sources to contribute simultaneously to shared parameters, facilitating a bidirectional exchange of information across levels. In contrast, the independent approach maintains analytical and data source separation per level, allowing only a retrospective comparison of results. Using African elephant (Loxodonta africana) data from Madikwe Game Reserve (South Africa), GPS telemetry data represented the individual-level, while camera trap data informed the population- level. The study compared ecological inferences within each level and examined how the relationship between levels varied under each approach framework. Results showed that individual-level inferences were similar between approaches, whereas population-level estimates varied substantially. When discrepancies existed between data sources, the integrated model tended to shift population-level estimates toward patterns observed at the individual- level. This adjustment can introduce bias if the individual-level data are unbalanced, as occurred in this study, where only females were tracked. The chapter concludes that integrated modeling enhances inferential power, but increases sensitivity to data quality and sampling bias. Accordingly, the strategy choice to combine levels information should be guided by the characteristics and conditions of the data, with the aim of maximizing robustness of the inferences.

    This PhD. Thesis contributes to understanding how information can be related across different levels of habitat selection, particularly those involving distinct ecological units of aggregation (species, populations, and individuals). Throughout the work, a series of key factors that influence this relationship have been identified and explored.

    Among them, methodological aspects such as the type, quality and resolution of data, model structure, and technique adopted to combine levels, play a fundamental role. In parallel, ecological characteristics such as spatial distribution patterns, behavioral variability, or niche specificity, also shape the degree of correspondence between levels. By systematically analyzing these components, the PhD. Thesis offers a framework to evaluate the conditions under which information can be transferred across levels, and when it is preferable to treat them separately. It also clarifies the type of insights that can be reliably extracted at each level, the degree of overlap and divergence between them, and the limitations of extrapolating information beyond the scale and unit of inference.

    These findings underscore the importance of aligning methodological design with the ecological scale of the question being addressed. While the PhD. Thesis has advanced this understanding through both empirical and theoretical contributions, it also highlights that further research is needed to refine multi-level inference strategies and to better integrate ecological complexity into habitat selection studies. Ultimately, this work provides conceptual and practical tools to enhance the robustness of ecological inference across levels, with clear applications for biodiversity monitoring, spatial ecology, and conservation planning.