50 likes | 167 Views
Gis-based Landscape Appreciation Model. Towards the development of GLAM version 3 Sjerp de Vries Landscape Centre, Alterra, Wageningen The Netherlands. Background.
E N D
Gis-based Landscape Appreciation Model Towards the development of GLAM version 3 Sjerp de Vries Landscape Centre, Alterra, Wageningen The Netherlands
Background • GLAM is a model that predicts the attractiveness of the countryside to local residents, based solely on physical characteristics of the landscape on which information is available in national GIS-databases • Version 2 uses four indicators, each with five levels: Naturalness, Historical distinctiveness, Urbanization, Skyline disturbance. • Spatial resolution of the model: 250 x 250 meters (6.25 ha) • Predictive validity is reasonable: 47% of variance explained • Problem: usability to evaluate policy measures is still low, because small changes often do not lead to different indicator values
Research questions and method • Level of spatial detail is quite acceptable, but how to improve the sensitivity of the model to more subtle/smaller changes in the physical appearance of the landscape? • Step 1: recalibrate the model based on recently gathered data on landscape appreciation (larger dataset, more areas rated) • Step 2: error analysis using recalibrated model • Where do predictions deviate most from actual attractiveness scores? • Is there some structure to be discerned in the direction or size of errors? • Spatial clustering • Type of landscape
Main results thus far • Recalibrated version of GLAM 2 available • Based on study “Beleving naar gebieden”, part of “Belevingswaardenmonitor” • About 300 demarcated areas rated by people living nearby • Explained variance: 38% (lower than in previous validation phase!) • Naturalness of the areas as main focus • Adding average rating of naturalness: explained variance 76% • Correlation GIS-indicator with rating of naturalness: r = 0.61 • Naturalness indicator performs quite reasonable already • Correlation of rating of naturalness with prediction error: r = 0.57 • But very important to improve it further! • Preliminary results based on www.daarmoetikzijn.nl • Over 700 postcode areas with at least 10 participants • Rating of attractiveness of countryside surrounding place of residence • 51% of variance in average attractiveness rating explained by GLAM 2
Preliminary conclusion and next steps • Feasibility of improved version of GLAM seems high • Already identified clear source of errors: naturalness • Still unused GIS-data on naturalness are likely to be available • Next steps • Looking at spatial pattern of errors • what is wrong with the present naturalness indicator? • Develop and implement new naturalness indicator • Calibrate new version of GLAM