General challenges in spatiotemporal prediction

Environmental Mapping and Spatiotemporal Modelling

We focus on large-scale, high-resolution mapping of environmental and spatial processes – such as air pollution, water quality, or infectious disease – using geoscientific data from multiple sources (e.g. Earth observations, aerial imagery, mobile sensors, and citizen science). We also aim to understand the contributing drivers behind these patterns.

Key Methodological Challenges We Address

• Data assimilation and information integration

Integrating geospatial data from diverse sources helps overcome the limitations of single-source datasets. This requires harmonizing data with different spatial/temporal resolutions and measurement supports.

• Spatial and spatiotemporal heterogeneity

Relationships between predictors and responses can vary across space and time. For example, the effect of traffic on air pollution may differ depending on fuel or engine types. We explore: Where and when do relationships shift, and how can we identify these zones of change?

• Optimizing spatial and spatiotemporal prediction methods

Many models exist for spatial prediction—but how do we select the best one for a given purpose? How do modeling choices vary for different goals (e.g. health studies vs. risk assessment vs. planning)?

• Model validation and uncertainty quantification