Within the EU FP7 framework, VOLANTE aims at developing a roadmap for future land resource management in Europe. It will design new methodologies and integrated models to analyse human-environment interactions, feedbacks in land use systems, hotspots of land use transitions and identify critical thresholds in land system dynamics.
In my research I am looking at changes in land uses on a pan-European level for the last decade. I focus on the dominant land use types - forests and agricultural areas - and my particular interests are changes in land management intensities. I aim at identifying the most influential drivers and determinants of change for the above mentioned land use types as well as their spatial characteristics such as spatial non-stationarity or spatial autocorrelation. Furthermore, I will try to isolate the effects of structural, political, or economic events such as the breakdown of the Soviet Union, the application of the Common Agricultural Policy (CAP), etc. In the end, I will use the syndrome approach to define and detect archetypical patterns of land use (intensity) changes. This will enable decision makers to spatially explicitly address disadvantageous developments.
To accomplish these goals I will utilise the following methods:
- Best subset regression + hierarchical partitioning
- Boosted Regression Trees (BRT)
- Geographically Weighted Ridge Regression (GWRR)
- Generalised Additive Models (GAM)
- Matching statistics such as Propensity Score Matching (PSM)
- Self Organising Maps (SOM)
Spatial cluster analysis and geographically weighted regression - exploring patterns and underlying processes of social and health characteristics in urban areas
Processes and characteristics of urban areas at the human-environment interface, such as social segregation or health disparities, depend on a diverse set of socio-demographic, economic, and environmental factors. Due to the heterogeneity of urban areas, one can assume that these factors vary over space in their strength and direction of influence. Geostatistical techniques allow detecting spatial clustering on a global, city-wide and on a local level to explore the spatial distribution of social and health phenomena. To study the influence of the explanatory variables on a target variable, traditionally, global statistical regression approaches, e.g., Ordinary Least Squares (OLS) are applied. Since these approaches emphasize similarities across space by building means, information about the spatial variance is lost. To take this spatial variation into account, a Geographically Weighted Regression (GWR) is utilized, extending the global regression by adding spatial information. This leads to a more detailed picture of the studied target variable.
We apply spatial cluster analysis and GWR to explore the spatial distribution and influencing factors of the social status in Berlin, Germany. A comprehensive set of socio-demographic, economic and environmental data is used.
Joint collaboration with:
- Prof. Dr. Tobia Lakes (Humboldt-Universität zu Berlin)
- Senate Department for Urban Development and the Environment, Berlin, Germany