Summary of Week 3
So, there it is, the final step of the third week of our course on Geohealth. We are glad that you are still with us! During this week you have studied two topics: Agent-Based Simulation and Spatial Statistics.
Regarding Agent-based simulations you have learned that we do primarily simulate because we are trying to understand a process. You also learned that agent-based modelling is an individual modelling approach and that each entity is simulated individually, not as a population but as an agent.
The components of an Agent-Based simulation are agents, environments and time. Each model can have multiple agents and many different environments. These environments can be the spatial layers from your GIS. Behavior of agents is often rule based, which can be visualized as an if … then rule. Agents that function is this way are not very smart. We can make them smarter when we implement artificial intelligence.
When you want to design your own agent-based simulation model, you have to start with your understanding of the process. You need to identify your agents and environments, the attributes and the behavior.
Furthermore in this week you have learned about Spatial Statistics. We’ve discussed spatial dependence, the variogram and spatial prediction (also known as interpolation or kriging). We looked at examples in the environment and health.
You learned first about the concept of spatial dependence, as expressed in Wado Tobler’s first law of Geography. “Everything is related to everything else, but near things are more related to each other” is referred to as the “first law of geography”. You looked at examples from remote sensing, air pollution and malaria.
You next learned about how to quantify spatial dependence using the variogram. The variogram is a plot of the semi-variance. Nearby points are expected to have similar attribute values, so the semi-variance is low. More distant points are expected to be dissimilar so the semi-variance is larger. The key parameters of the variogram are the sill, nugget and range. These define the maximum spatial variability, the non-spatial variability and the limit of spatial variability respectively. Specifically the range defines the maximum geographic separation (lag distance) where two points are expected to be correlated.
Finally you learned that the variogram can be used for spatial prediction and for making maps when we only have measurements at a limited number of locations. You were shown examples from air quality and malaria.
In the last week of this course, we offer two new topics that are presented by Ente Rood of the Royal Tropical Institute (KIT). They are both a continuation of two previous in-depth topics, accessibility and health facility planning from week 1 and spatial statistics from last week. Where the topic on accessibility of health facilities focused on network analysis, we will now focus on using roster-based operations. We hope you will enjoy these new topics!
© University of Twente