As of today, all remaining lectures have been cancelled as classes move online for the foreseeable future due to the Covid-19 pandemic.

Today we began our presentations on health geography. The topic of my paper was calculating and modelling access to healthcare in terms of car travel times and bus accessibility in East Anglia. This was an interesting read as it is a rather old paper,using techniques and tools that would be considered obsolete by today’s standards.

The study involved using a vector road network of the region divided into nodes, with locations of general practitioner services placed at the nearest node. Driving-time distance was then calculated from individual postal codes, to find car travel time from each residence.

The other part of the study involved mapping bus networks and classifying them based on service degree. GP locations were buffered 800m and all bus routes passing through were identified. These bus routes were each again buffered 800m and all postal codes lying within the buffer were then counted as being bus accessible.

Naturally, there were a lot of flaws with this study, namely that there are a large number of simplifications and ignored variables, along with some rather arbitrary decision. The techniques used also seem very unwieldy by today’s standards, as the study could likely have been done by using an Origin Destination Cost Matrix or Generate Service Area tool in a fraction of the time and with greater precision. However, it was nonetheless interesting to see how GIS was carried out using vector data in the days before ArcMap.

Due to this paper, I found another presentation extremely interesting, that also covered accessibility to GPs in a different region of the UK. However, this presentation was on a much newer paper, where the authors actually did use Create Drive-Time Area tools in ArcOnline and carried out GWR analyses to great detail, making it a far more advanced and recent version of the paper I presented on.

Today we began thinking about our final project ideas. I am considering doing a species distribution modelling project revolving around the use of the Maxent modelling software.

Spatial data is of great concern to population health studies. People and entities do not exist in a vacuum, and as such are influenced by and influence those around them. This can lead to geographically correlated issues where space and place are of huge importance, hence the need for GIS in health studies.

This is especially true as populations are often distributed unequally. Wealthier people tend to be concentrated in one area, and people of certain races in others, with varying densities.

There are many applications for health GIS, such as disease mapping, cluster detection, environmental hazards, or modelling health services.

Disease mapping is the epidemiological study of how a given disease spreads, where it goes to and at what rate. This is important as it can be used to predict which areas would be impacted next or the most, and can play a key role in developing resilience or response policies.

Cluster detection is the use of GIS to identify areas with high instances of a given health condition. This often requires the use of a Geographically Weighted Regression analysis in order to determine correlation of a health outcome with any of a variety of possible factors.

GIS can also be used to monitor environmental hazards. This involves analysis of risk exposure as well as mitigation techniques.

Modelling health services is another important role of GIS. For example, GIS can be used to find the ideal location for emergency services through the use of a service area tool, which creates drive time maps and can be used to estimate time-distance from response areas.

However, the use of GIS in health geography also comes with some limitations. One key issue is that of data. Often, health GIS requires public census data, which is always out of date and frequently incomplete or flawed. Flawed data leads to flawed analyses. Furthermore, people are extremely dynamic, and it is difficult to take all the possible variables into account, which often results in regression analyses that indicate one of a myriad number of possible underlying variables.

The topic of today’s lecture was an introduction to health geography. Considered to be an improvement by some over “medical geography”, health geography is concerned with not simply epidemiology, but also the spatial analysis of access to and quality of healthcare, and has a variety of applications, such as optimal ambulance routing.

Health geography can be divided into three main areas: disease ecology, healthcare delivery, and environment and health.

Disease ecology studies the geographical spread of infectious diseases.

Healthcare delivery analyzes the efficacy of modern healthcare systems and how they deal with and are impacted by geography. For example,this can include average patient distance from hospitals, or travel times from high injury rate areas to treatment centres.

Environment and health is a branch of geography that studies how place can impact one’s health. An example would be how proximity to a coal plant is correlated with issues such as lung disease.

Wednesday was the first day  of in-class presentations on landscape ecology. However, I ended up presenting on Friday.

The topic of my presentation was a paper written in 2019 analyzing the impact of night time light pollution on bat presence probability and landscape habitat connectivity. The study took place in the city of Lille, France and data was gathered through audio recordings of bats echolocating, as well as from a NOAA light raster.

Analysis in QGIS was used to construct bat presence probability based on light pollution and recorded data, with an algorithm to create a probability raster. Linkagemapping Toolkit in ArcGIS Pro was used to map least cost pathways that represent movement corridors for bat movements.

Results showed each species of bat impacted in different ways, with one species being adverse to light pollution, one species being more prevalent with increasing light, and one species preferring small amounts of light. Overall it was deemed that light pollution was a good indicator of bat presence, but still less so than distance to wetlands.

One other presentation I found interesting was the study concerning modelling lynx populations. I was able to see how the Linkage Mapper Toolkit was application to modelling species habitat connectivity in various other landscape types.

Today’s topic was ways to quantify landscapes in spatial analysis. Landscapes are impacted by a wide variety of forces. There is a concept in landscape ecology of how “form influences process and process influences form”. For example, this could mean how a forest fire (process) can change the landscape by removing stabilizing tree roots that lead to landslides and rotational slumps (form). This creates a constant feedback loop where the state of a landscape is always in flux. As a result, any given map is merely a snapshot of one of the possible patterns that might have emerged from a process.

One type of landscape metric is spatial autocorrelation, which is divided into first and second order processes. A first order process would be where patterns develop in response to some underlying environmental factor, while a second order process is when patterns emerged as interactions between objects.

Stationarity is another important metric, and is a measurement of to what degree processes may shift over changes in space. This implies a lack of directional bias, meaning that these processes are isotropic.

Patterns themselves have 5 classes of metrics into which they are categorized:

number of classes or cover types

texture measures

degree to which patches are compact or dissected

whether patches are linear or planar

whether patch perimeters are simple or complicated in shape

Today we went over how the spatial element is crucial in many analyses, from social to physical.

One major issue faced in geography is that of scale. Selecting the correct scale is crucial in any project and may completely change the nature of one’s results. This is commonly seen in the Modifiable Areal Unit Problem (MAUP), whereby selecting in inappropriately small or large scale results in skewed analysis results.

Selecting spatial unit boundaries is another issue that often occurs alongside scale-based problems and often serves to confound many projects. Using different boundary schemes may yield completely different, yet still correct, results. This in turn often leads to disagreements, as the data may be used for different purposes.

An example of this is gerrymandering, a practice created by 19th Century American politicians, where constituency lines are drawn in such a way as to ensure victory in a first past the post electoral system.

The extent of the study area must also be clearly defined, as selecting an overly large or small extent upsets the scale.

Spatial analysis must also take into account the effects of spatial autocorrelation. This is when a presence of some quantity expressed in one spatial unit is also seen to an extent in the immediately surrounding neighbour units. Spatial autocorrelation is non-random, meaning that there is often an underlying cause to it, and the degree of autocorrelation can be measured using Moran’s I, which provides a numerical score between -1 and +1.

In ecology, autocorrelation can often be explained by three neighbourhood models.

The first is a grouping model, where similar individuals choose to be located in a certain area, or are constrained in some way to be there.

The second is a group dependent model, in which the individuals in a given area are subject to similar external influences, such as a species of flower prospering in an area due to ideal physical conditions.

The final model is the feedback model. This is when individuals interact with and influence one another to become similar. For example, people who live together will inevitably develop shared interests or behaviours.

The main issue for spatial analysis is determining which cause is at work, especially since they may all be simultaneously true at differing scales.

UBC declared today as a snow day due to extreme weather conditions. No class!

Today we went over the basic concept of GIS, starting with the 5P’s: People, Perspectives, Processes, Patterns, and Places. These 5 are deeply interconnected and interact with one another in GIS analysis. The introduction of these 5 components highlights the importance of GIS in a modern world.

The class also introduced the main focal points of the course: landscape ecology, health geography, and crime geography, as well as the tools that we will be using in each. From the tools, it seems that this course will be highly statistical in nature.

Landscape Ecology

Landscape ecology looks at how living organisms interact with their environment, shaping their surrounds while simultaneously being shaped by them. This unit will focus on how landscapes change over time and how this affects the organisms that live in them.

Health Geography

Health geography is the study of health care and diseases in populations. On one hand, health geography looks at how to deal with distribution of health care, such as where to place hospitals and emergency services. ON the other hand, health geography also examines the spread of disease and health conditions, similar to epidemiology. In this field, health geographers map disease outbreaks, track vectors, and use geographical data to better understand how diseases can be prevented or treated.

Crime Geography

Crime geography maps out instances of crime and uses statistics with the aim of predicting and preventing future crimes. Data used in crime analysis is often compared with surrounding data to identify correlations between certain types of crime and certain elements. This was done in lab 3 of 370 where we looked at data for various crimes in Vancouver.