Integrating expert-based objectivist and nonexpert-based subjectivist paradigms in landscape assessment

This thesis explores the integration of objective and subjective measures of landscape aesthetics, particularly focusing on crowdsourced geo-information. It addresses the increasing importance of considering public perceptions in national landscape governance, in line with the European Landscape Convention's emphasis on public involvement. Despite this, national landscape assessments often remain expert-centric and top-down, facing challenges in resource constraints and limited public engagement. The thesis leverages Web 2.0 technologies and crowdsourced geographic information, examining correlations between expert-based metrics of landscape quality and public perceptions. The Scenic-Or-Not initiative for Great Britain, GIS-based Wildness spatial layers, and LANDMAP dataset for Wales serve as key datasets for analysis.
The research investigates the relationships between objective measures of landscape wildness quality and subjective measures of aesthetics. Multiscale geographically weighted regression (MGWR) reveals significant correlations, with different wildness components exhibiting varying degrees of association. The study suggests the feasibility of incorporating wildness and scenicness measures into formal landscape aesthetic assessments. Comparing expert and public perceptions, the research identifies preferences for water-related landforms and variations in upland and lowland typologies. The study emphasizes the agreement between experts and non-experts on extreme scenic perceptions but notes discrepancies in mid-spectrum landscapes. To overcome limitations in systematic landscape evaluations, an integrative approach is proposed. Utilizing XGBoost models, the research predicts spatial patterns of landscape aesthetics across Great Britain, based on the Scenic-Or-Not initiatives, Wildness spatial layers, and LANDMAP data. The models achieve comparable accuracy to traditional statistical models, offering insights for Landscape Character Assessment practices and policy decisions. While acknowledging data limitations and biases in crowdsourcing, the thesis discusses the necessity of an aggregation strategy to manage computational challenges. Methodological considerations include addressing the modifiable areal unit problem (MAUP) associated with aggregating point-based observations. The thesis comprises three studies published or submitted for publication, each contributing to the understanding of the relationship between objective and subjective measures of landscape aesthetics. The concluding chapter discusses the limitations of data and methods, providing a comprehensive overview of the research.