Modeling Gentrification

Gentrification is the process of middle income people taking over older neighborhoods in central cities by purchasing the homes, renovating them and changing the character  of the neighborhoods completely.  This process has been ongoing for decades and represents part of the back to the city movement of both capital and people in America, (Slater and Smith 1979.  While gentrification is good for urban revitalization, the process is very controversial as it involves one class of people displacing another class. It is actually the opposite of the filtering down process in housing theory.

There have been efforts to model gentrification so that local municipalities and community leaders can have a sense beforehand as to which neighborhoods are likely to suffer gentrification. This got me into thinking about the modeling framework most appropriate for simulating this land use pattern.

Logistic regression is a good choice. Markov Chains is also worth considering, but the process may not be random, so the Markov Chain might fail. Perhaps agent based models might be a good choice.  Below are a few links I am beginning to put together on the topic of gentrification modeling.





Rank Size Rule and Polynucleated Cities

This week I managed to read Michael Batty’s (2001) paper on polynucleated urban landscape. Admittedly, the paper was published a long time ago, but the topic is still relevant as there is still a lot of interest in understanding the structure of system of urban places.

Batty’s paper essentially argues that the appearance of edge cities in the US undoubtedly give rise to added complexity in the American urban landscape.  However, this apparent complexity does not change the underlying rank size distribution of the system of cities in the country and the log normal distribution of cities is still maintained.

The paper uses almost 100 years of data from the British urban system to illustrate the point that although the spatial structures of systems of cities may evolve over time, the underlying rank size pattern remains consistent because urban structures are remarkably persistent over time.

Both the historical rank size graphs and the agent based models that were presented are plausible, however, as usual, the number of runs in the agent based model was not effectively justified. Still, a well argued paper.

I also read Brian Berry’s  et al (2011)  paper entitled  “The city size distribution debate: Resolution for US urban regions”. Another good paper from a stalwart geographer.  However, I was stunned to discover that in their effort to show that the American systems of cities closely follow a rank size distribution, they used some very large construct for metropolitan areas, including combining the Detroit-Cleveland areas a single urban area.

In my opinion, the American urban system follows the rank size rule, but not closely throughout the entire extent of the curve.



Agent Based Modeling

Agent- based modeling involves creating and giving  instructions to different classes of virtual agents and observing the patterns that emerge, in both time and space, as the agents interact with each other and with their environment. These emerging patterns offer insights into how a phenomenon might grow and change over time.

Agent based models have proven useful in simulating the growth of many complex systems in diverse areas such as biology, economics and transportation. In geography, scholars have used agent based models to simulate urban growth, disease spread, agriculture development, etc.  The modeling approach is different from statistical modeling such as regression analysis, but there is much similarity to rule-based, cellular automata modeling.

A good place to start learning how to create agent based models is look at the various sample models available through NetLogo and Repast Symphony.  In NetLogo my favorite sample models include the Sheep-Wolf predation model, the HIV model, the Tijuana Bordertown, and the Ant model.  In Repast, I think the Zombie-Demo model is a good one to start with, particularly because documentation exists for it.

A great set of tutorials for learning NetLogo are:


Both NetLogo and Repast allow for GIS to be integrated into agent-based models.   After learning the basics of non-spatial agent-based models,  it would be good to start working through tutorials that deal with GIS layers within these modeling frameworks.   For NetLogo, you can look at:


Agent-Analyst is an ESRI extension that allows ArcGIS to communicate with Repast for GIS modeling. Downloading and installing Agent Analyst is easy, but it needs a bit of configuration before it works.  I found this stackexchange post to extremely helpful in configuring Agent Analyst.






MapWinGIS.ocx is an open source mapping control that is used for adding GIS/mapping functionality to Windows Forms based applications. Code can be written in Visual Basic 6, VB .NET or C# and can be commercial or open source. If you ever thought about building your own custom GIS application, then MapWinGIS.ocx is a great tool to use.



Many organizations share their spatial layers as map services. ESRI has a product called ArcGIS Server which is used to deliver map services to websites and desktop applications that are connected to the Internet.

GeoServer is a free and open source software (FOSS) that exposes your data as geospatial web services. If you want to share your data as web services, GeoServer could be a cheap, but robust alternative to ArcGIS Server.


Data Portals

Classical Models in Geography

Human geography, like other social science disciplines, has it fair share of models. Models are abstractions or simplifications of the inner working of complex phenomena that exist in the real world.  We build models to reflect out understanding of how things work.  The models will provide us with a way to describe phenomena as well as to explain and even predict them.

In geography, the globe is one of the best known models. It is a model of the earth and we can use it to visualize how the earth works in relation to lots of things including night and day, seasons, climate, air flow, etc.

Maps are also brilliant models of the earth. Depending on our needs, we can simplistically represent the entire earth on a map, or just a single street block to show homes that were built in a particular style.

It is often said that that in the 1800s, the French were the first to start representing intangible things on thematic maps. However, this is not true. Way back in antiquity, the Greeks figured out the shape of the earth and started representing climate zones on maps.   Today, most maps are used to represent intangible things like temperature, disease rates, homicide patterns, income distribution, etc. Maps depicting these things are also models.

In Geography, when we speak about classical models, we are talking about a set of models beginning with Weber in 1909 that sought to explain industrial location, agriculture land use, location of cities in a region,  internal land use of cities,  lmigration patterns, transportation patterns, economic growth, territorial behavior of states, etc.

Two decades ago, these models were the foundation of geography programs in the US and elsewhere. Today, many geography majors have never even heard about these models. Yet, these foundation models continue to provide us with a way of thinking about spatial patterns and their logic and conclusions are prevalent in current geography textbooks. Also, some disciplines such as urban and regional planning still rely heavily on some of the theoretical work of this era to guide settlement planning.

In modern geography, the concentration known as spatial analysis is firmly connected to the geospatial sciences and continues to directly extend the work of classical model builders in geography. Spatial analysis takes, as it starting point, the need to understand spatial patterns. Space rather than place is the central focus and the goal is to build mathematical and computer models to represent, explain and simulate patterns created by natural and human processes on the landscape.

Models in spatial analysis have practical applications and relevance across a broad range of disciplines. However,  because they are often taught without reference to their historical connection to geography, people often fail to see the connection between current spatial modeling and classical models in geography.

In my own dealings with spatial analysis, I try not to lose sight of the classical problems in geography because they represent the historical basis of spatial analysis as well as the type of problems that geographers are interested in solving.  It is true that the way we approach problems in geography changes with time and technological developments, but the core nature of the problems remain constant.

Below is a list of some classical models in geography. Scrutinizing the list, you will notice that the common interest is solving problems related to spatial patterns.


Location of Economic Activities

  1. Agricultural location theory – Johann von Thunen (1826)
  2. Least Cost Industrial location theory  – Alfred Weber (1909)
  3. Locational Interdependence – Harold Hotelling (1929)
  4. Profit Maximization – Losh (1952)


System of Urban Places


Internal Structure of Cities

  1. Concentric Zone theory of urban land use – Burgess
  2. Sector theory of urban land use  – Hoyte
  3. Multiple Nuclei theory of landuse – Ullman/Harris
  4. Bid-Rent Theory – William Alonso (1964)
  5. Vance urban realm theory


Economic Development

  1. Growth Pole theory – Francois Perroux (1949) 
  2. Polarization effects -Albert Hirschman, i.e., the negative impact of a growth pole on surrounding regions. Trickling down refers to the positive impact of a growth pole or growth center on adjacent regions.
  3. Backwash and spread effects – Gunnar Myrdal, Swedish economist. These are the same concepts as polarization and trickling down.
  4. Rostow’s stages theory of economic development
  5. Core periphery model – John R. Friedman
  6. World Systems theory – Immanuel Wallerstein


Population Growth

  1. Demographic transition model
  2. Malthus Population Growth Model



  1. Gravity model
  2. Ravenstein laws of migration



Spatial Diffusion – Torsten Hägerstrand


Territorial Expansion

  1. Heartland/ Rimland theory

Web Mapping

Mapping has experienced a huge resurgence since the advent of the Internet and the proliferation of GIS technology. Digital maps have now become a vital part of our everyday communication and transportation infrastructure.  Knowing how to distribute information using static and interactive maps via the web is an important and highly marketable geography skill.

I use the websites below to introduce students to basic concepts in web mapping. The best way to learn web mapping is obviously by building your own web mapping sites. Start by building simple web pages that include static maps and animated maps. Afterwards, move on to interactive maps using platforms such as ArcGIS Online, ESRI JavaScript API, Google Map, Open Layers, ArcGIS Online.


Web Mapping Links


Sample Websites


Research Topics