CONTEMPORARY GHOST TOWNS

An attempt to visualize pre-crises housing speculation and its affects on population flows in Hillsborough County, Florida.

Objective:

The goal is to visualize the financial crisis’ affect on population flows, as well as to draw attention to the tensions created by a surplus of vacant houses coupled with a rise in need.

Focus Area:

Hillsborough County, FL.

Florida experienced high levels of home foreclosures due to rapid home construction in the years leading up to the crisis. - source

FOCUS TIME RANGE:

2000 - 2010

______________________________________________________________________________________

To begin, this chart, created with NAHB data shows that the national number of authorized housing starts fell drastically in 2008. 

For Hillsborough County, Florida, housing starts, per Census Data looked something like this

In Florida, prior to 2005, housing starts were on a dramatic incline.  This means that the county planners and developers anticipated continued growth.  When the crisis occurred, housing developments that were expected to be filled, did not meet expectations, and many of those that were in the process of being constructed were stalled.

Photo Source

In conjunction with unoccupied and stalled new housing developments, foreclosures began to rise.

To visualize the rise in home foreclosures, I began with 2008 national county level foreclosure data provided by the Department of Housing and Urban Development in order to see where Hillsborough County stood. 

Hillsborough County fell in the 350,001-400,000 (number of foreclosed homes) range.  Because it was not at either extreme level of the spectrum as the crisis began, it seemed a perfect county to focus on, in order to really see the affects of the Financial Crisis.

So, with Hillsborough County in mind, I decided to parse my HUD FL foreclosure data set further in order to see the breakdown of 2008 home foreclosures by town within the county.

Outside of downtown Tampa (Tampa City), Brandon (where I grew up) and Town n Country top have the highest levels of foreclosed homes.

[ISSUES] In order to truly show that the crisis severely affected population flows however, I need foreclosure data from at least as recently as 2010.  This has actually presented an issue in my project.  The most reliable foreclosure data sets seem to be provided by HUD, however their foreclosure data does not extend beyond 2008.

In my initial project planning stages, I intended to create a series of visualizations using the Metropolitan Area Quarterly Residential and Business Vacancy Report.  This report documents vacant and abandoned properties as reported by the United States Postal Service.  The data is organized by quarter for the years 2008-2010.  There are two different data download options:

Fields for MSA file:

CBSA: 5-digit CBSA code
METRO_NAME: CBSA Metropolitan Statistical Area Name
QUARTER: Description of quarter, ie., “June 2009”
AMS_RES: Total Count of Residential Addresses
AMS_BUS: Total Count of Business Addresses
RES_VAC: Total Count of Vacant Residential Addresses
BUS_VAC: Total Count of Vacant Business Addresses
RES_VR: Residential Vacancy Rate (RES_VAC/AMS_RES)*100 
BUS_VR: Business Vacancy Rate (BUS_VAC/AMS_BUS)*100
RES_VR_CH_PQ: Residential Vacancy Rate Change from Previous Quarter
BUS_VR_CH_PQ: Business Vacancy Rate Change from Previous Quarter
RES_VR_CH_PY: Residential Vacancy Rate Change from Previous Year
BUS_VR_CH_PY:Business Vacancy Rate Change from Previous Year

Fields for Metro Division file

MDIV: 5-digit CBSA Metropolitan Division code. All MDIV codes end with a “4”.
MDIV_NAME: CBSA Metropolitan Division Name
CBSA: 5-digit CBSA code
METRO_NAME: CBSA Metropolitan Statistical Area Name
QUARTER: Description of quarter, ie., “June 2009”
AMS_RES: Total Count of Residential Addresses
AMS_BUS: Total Count of Business Addresses
RES_VAC: Total Count of Vacant Residential Addresses
BUS_VAC: Total Count of Vacant Business Addresses
RES_VR: Residential Vacancy Rate calculation (RES_VAC/AMS_RES)*100 
BUS_VR: Business Vacancy Rate calculation (BUS_VAC/AMS_BUS)*100 
RES_VR_CH_PQ: Residential Vacancy Rate Change from Previous Quarter
BUS_VR_CH_PQ: Business Vacancy Rate Change from Previous Quarter
RES_VR_CH_PY: Residential Vacancy Rate Change from Previous Year
BUS_VR_CH_PY:Business Vacancy Rate Change from Previous Year

The goal was to create graphs and maps drawing from the total counts of vacant residential addresses for Hillsborough County for the years 2008, 2009, and 2010. 

Before deciding this is what I would do, I should have dug further as it turns out these data sets are temporarily unavailable and there is no estimated time frame for their availability. 

So, the next thing I explored was a series of data provided by policymap that addresses levels of home ownership for 2000 and 2010.  I filtered the data to show Hillsborough County in 2000 and 2010. 

The two maps show a significant decline in home ownership levels between 2000 and 2010, especially in certain suburbs of Tampa, like Brandon. 

A decline in home ownership in the suburbs as well as in the downtown metropolitan area correlates to rising foreclosure rates.

Unfortunately, my project is ending without my ideal final map included.  What I had intended was to create a final choropleth map

Layer 1: Pre-crisis housing development plans as drawn from a Hillsborough County Proposed Land Use dataset.

Layer 2: Empty/vacant housing that prove that development plans were not sustainable and thus a retraction in sprawl

What my project does demonstrate though, is a significant fall in housing starts coupled with a rise in foreclosures/decline in home ownership for Hillsborough County, FL.

GIS Final Project Process

This will be a rough post where I work out what exactly I am trying to do, and also have all of my stuff in one place in order to not lose my mind. 

Final Project Proposal:

For my final project, I hope to explore datasets that might illuminate some of the 2008 financial crisis’ affects on population dynamics/shifts as well as what the crisis and said shifts imply about the space(s) in which we currently live.  I hope to find ways to visualize housing vacancies in the sprawling housing developments that blew the bubble that eventually burst.

I will use 2008 as the center of what I mean when I refer to “pre” and “post” crisis. 

I realize that my skill set is not yet strong enough to create maps that can truly give the situation the emphasis it deserves.  So, I have to keep reminding myself that the point is to show some of what I have learned to do.

First Attempt:

After a 3 hour wormhole of exploring census data and getting stuck, I decided I needed to just make a map…any map, in order to see a deliverable come to fruition and hopefully become motivated again.  So, I returned to the Gridded Population of the World data set that I was sure could help me visualize some aspect of the story…any aspect.

The goal was to use the data set to create a split map of projections of population densities for 2005, 2010, and 2015 that were estimated in 2004.  Then, I would find other data sets to show that the projections were not realized; in part because of the financial crisis and the way it has affected population movements.

Aaaannnnddd this is what I got!  I actually like it, aside from the blue that mysteriously appeared after I had already saved and exported it.  I think it says way more than what meets the eye, though. 

I’m losing my mind. 

So, that didn’t produce anything of true value in telling my story or showing off my ESRI skills…but it did make me laugh.

Second Iteration:
Next, I found this data set that illustrates US home foreclosure estimates for 2008.  Back into ESRI, but running low on energy, I created this map which I hope to return to and make more compelling (as well as correct my misspelling of ‘foreclosure’).  But for now, it is a start.

While the data set is at a national level, I used a sequel query to pull out just the Florida data.  I then grouped counties together based on their numbers of foreclosures.  I think it is interesting that Hillsborough County is the only county in the median grouping of 350,001-400,000 foreclosures.  Because this data comes from the very beginning of the crisis, narrowing in on the median county pre and post 2008 might provide results  more directly linked to the crisis than a county on either end of the spectrum.

Third Iteration:

Next, I found Policy Map an online mapping program that has some interesting data on demographics, mostly from the US Census.  

I made two maps on the site that show the percentages of Hillsborough County home owners in 2000 and 2010. There are certainly areas that decreased in home owner percentages post-crisis.  In Brandon, a cookie-cutter suburb of Tampa, the percentage of home owners fell from the 71.11%-84.85% range in 2000 to the 60.27% or less range in 2010.  Of course this drop can be caused by various factors, but as the area is full of the kinds of homes that were built during the boom and then filled with many residents who are likely to have taken sub-prime mortgages, I feel it is safe to say that a portion of the dip may be directly related to foreclosures.

The maps are slippy, so you can click and drag around Hillsborough county to compare rates of homeowners in various neighborhoods in 2000, when the economy was on the up, and then in 2010, a few years and many devastated people later.

Hillsborough County Home Owners 2000

Hillsborough County Home Owners 2010

Fourth Iteration:
I decided to attempt to visualize that housing authorizations rose leading up to the crisis, and then fell drastically in 2008 to juxtapose with visualizations of foreclosures that began to rise in 2008.  I honestly don’t know what this shows in regards to my overall thesis, but the wormhole took me here.  I imported individual 2000, 2002, 2004, 2006, 2008, and 2010 housing start authorization data from the National Association of Home Builders, parsed the huge data sets, squeal queried the hell out of what remained, then joined each individual yearly set together in order to create one graph showing yearly housing start authorizations.   The end result is nothing amazing as esri does not really allow for building beautiful graphs/tables…but it took hours, so voila!

Fifth Iteration:

Next, in Geocommons, I imported Hillsborough County spatial data from the  Department of Housing and Urban Development.  Once I had a map of Hillsborough County created, I explored the Planning Commission’s datasets and downloaded a 2011 existing land use shape file.  The same site also has a future land use shape file, so my hope is to juxtapose the 2011 with the future plans, and then somehow relate that to my overall topic.  

In the works…

I can possibly further explore Hillsborough County’s foreclosure data. Geocommons has some data sets that could potentially work.

In order to tell my story more successfully, it would be beneficial to find data on housing speculation pre-crisis.  If I can find some sort of data set that shows developers were planning to build in 2008/2009… I can then argue that their plans were halted because of the crisis and support my argument with the foreclosure data I am already working with. 

I realize that as I go, my initial question is shifting in that I am looking less at flows of people and more at foreclosure rates…

Hillsborough County Home Owners 2000

Hillsborough County Home Owners 2010

GIS: Some New Maps

Here are a few new maps I have made in class, with brief explanations.

Hey mom!  I will teach you how to make maps next time I visit. A computer whiz like you can pick it up in no time!!

Lab 5 Map 1: This map shows all land parcels that are in a vernal-pool extension zone.  Hypothetically it could be used by a local government to alert land owners who are within vernal pool zones.

Lab 5 Map 2: This map shows potential campsite locations that are within a certain requested distance from botha roadway and access to water. 

Lab 5 Map 3: This map shows regions that might be worth solar-power investment.  The hypothetical company wishing to invest in solar power generation established a set of criteria for investment regions, including (a) they are close to a heavily populated city, (b) they are in an area of low elevation (c) they experience high rates of cloudless days(d)they are close to the equator.  This map is separated into three maps highlighting regions of cloudless percentages of 60%, 70%, and 75%.

Lab 6 Map 1:This map further expands onLab 5 Map 3by targeting cities that were found to have a 60% cloudless rate that are also within a 50 mile radius from a city with a population of 1 million or greater.

I swear that I will eventually start using my newly acquired mapping skills in relation to Beirut’s potential transportation system…as well as other interesting urban focused questions.

GIS Problem Set 1.2: Report

For the second part of my project (explained in the last post), I decided to focus on Nouakchott, Mauritania.  As mentioned previously, the city has a low level of base OSM data, a high level of risk, and quality satellite imagery.

Some brief research about Nouakchott’s urban plan illustrated that a major issue for the city is its “rapid and unplanned sprawl”.  According to Sameh Wahba’s 2001 report, the density of the city is challenged by the citizens’ desire for large parcels of land and single-family homes.

My Open Street Map UserName is DesireeLaVecchia (original, I know) so if you desire to visualize any of the summary of edits listed below, feel free look at Nouakchott in OSM.

A very brief list of my edits:

- Removed one of two identical “highway”=”primary” in order to clean up the map a bit.

- Added the name of the “Friday Mosque” to an already existing mapped plot of land corresponding to a place of worship.  Per the Wikipedia page that described its location, I was able to identify its name.  Via Bing! Satellite imagery, I was able to narrow in on the physical building within the already mapped plot of land, which I further mapped out.

- Did the same thing as above with a market place.

- Added the opposite one-way direction of Ave. Abdel Nasser.

- Mapped 2 personal swimming pools! (useful in the case of a heatwave or if it is for any reason necessary to know where the richest Mauritanians live, of course)

- Chose an area with many buildings and mapped them out.  I felt interested in the particular plot that I worked in because it seems illustrative of the sprawl described at the beginning of this post.  I used “building”=”yes” for those structures that I could not distinguish as commercial or residential and “building”=”residential” for those that clearly resemble informal housing structures.  I feel confident in going with “residential” in instances where the structure is not uniform and appears to have been added to over time - as is typical with many informal/makeshift housing structures.

- Added an agricultural field, and a footpath going through it.  I wonder if this is part of the Green Belt project.

- Became obsessed (literally hours-long obsessed) with a particular paved roadway that had not yet been mapped.  This roadway is clearly shown in the Bing! satellite imagery and wound throughout a huge area of Nouakchott’s sprawl.  It also connected with several unpaved paths through residential areas, which I also mapped.

So, there you go, Nouakchott OSM basemap!

GIS Problem Set 1.1 (Report)

In my last lab, I looked at hazardous areas and their levels of base data.

Next, the goal is to look at the list of cities I created that might benefit from more base data and contribute to the data for one of them.  I will seek to contribute to the level of base data by accessing public data and metadata and by working in OpenStreetMap.

So, the questions are:

Which place should we map? 

Which place can we map?

First, in an attempt to answer Which place should we map?  I sought to find out which at-risk city is in most need of base map data out of the three on my list: Plymouth, Montserrat, Porto Novo, Benin, or Bissau, Guinea Bissau?

In exploring initial base map data for each of the three, I can quickly rule out Bissau from the list.  It seems that this city has significantly increased its base data levels since the time of the OSM v. Google project. Many forms of infrastructure including main and arterial roadways and even housing and possibly retail infrastructure is now illuminated by OSM.  Plymouth and Porto Novo, however, both seem lacking in this initial stage, with only some limited street data available.  These limited data seem to show some major roadways, but no subsidiary roads or building infrastructure.

Next, I explored both Plymouth and Porto Novo through PotLatch 2 satellite imagery.  For the purposes of adding valuable base data, I think that Porto Novo might now be eliminated as its satellite imagery does not give me much to work with if I want to add infrastructural data to the OSM base data for the city.

Porto Novo Satellite Imagery Screen Shot:

Plymouth, however does have some satellite imagery that can prove beneficial to the quality of my OSM base data additions.

Plymouth, Montserrat Satellite Imagery Screen Shot:

So, at this point I might answer my initial question: Which place should we map?

While both Porto Novo, Benin and Plymouth, Montserrat are places that are in desperate need of base data because they both have very little base data based upon today’s levels of OSM as illuminated in Problem Set 1.1, and are also in high natural disaster risk areas, based upon results of Lab 4, it is also important to consider which place we can map.

Which place can we map?  i.e. which city out of this reduced list of two can be added to in terms of base data in the most effective way?

Which is where I hit a HUGE BLIP in analysis!!  This blip, however, further justifies my initial belief that Plymouth, Montserrat [is] in desperate need of base data, it is just that this need of base data is now past-tense.  It was in need.  It needed base data, majorly, back in the 1990’s when a series of volcanic eruptions ravished Plymouth and left it utterly reduced to nothing.  While some of the infrastructural services have been relocated to other areas of the island, over two-thirds of the surviving population evacuated the island and have not returned.  Its Wikipedia page refers to it as a “ghost town” with population 0.

So, the previously shown screenshot of OSM data layered over PotLatch satellite imagery is showing old infrastructure in the now desolate city.

In regards to Porto Novo, as mentioned above, its satellite imagery leaves us little to work with by way of adding base data.

Alas!  I am basically back at square one in searching for a high-risk city, with low base data, that we can potentially add valuable base data to.

So, working with my Lab 4 data, I isolated several more at-risk cities with low base data by building a query to select attribute.  My query reads: “google” <=2 AND “osm” <=2, meaning I want to isolate all cities that have received a ranking of less than 2 in both google base data and OSM base data.  This gives me new data with which to work, and all of this data was considered (at the time of the OSMvGoogle project) to have low base data. 

With this new, more isolated data, I decided to look at the following cities.  Comments in italics describe my decision making in what city to focus on for the remainder of Problem Set 1.1.

Antananarivo, Madagascar- OSM investigation shows that this city now has significant levels of base data and is no longer in major need.

Port Louis, Mauritius- OSM investigation shows that this city now has relatively higher levels of base data.  I think I can find one with less data and higher need.

Amman, Jordan-OSM investigation shows that this city now has relatively higher levels of base data.  I think I can find one with less data and higher need.

Nouakchott, Mauritania- While there is some OSM base data, PotLatch satellite imagery shows a significant amount of infrastructure that has no related OSM data.  Therefore, this is the city I think both can and should be mapped!

So, it is now my belief that Nouakchott, Mauritania is the city we should focus on mapping.  It is in a region of high risk, has low levels of OSM base data, and has some significant PotLatch data with which we can work.

Short-Term Adventure Plan

Next week, I am getting out of this city.  Thank. god.  It’s going to be a mini adventure to the South, but the real South, not the North in the South (Florida), where I come from.  I am beyond excited.

It is going to be a road-trip in a borrowed car full of blankets to eventually share with so many friends.  The drive will be nearly as long as a flight to Lebanon, and with less run-way fashion.

First stop: Asheville, North Carolina!!  We are staying for two nights in a little cabin/chalet that comes with a HOT TUB ON THE DECK!

After exploring Asheville for a couple nights, we will continue on to north Georgia and meet up with a group of my closest friends from Florida, and also a friend from Indiana.  With our forces combined, we will have two fabulous nights of camping at Cloudland Canyon State Park. 

I am so excited to see things I have not yet seen with people I love so dearly.

GIS Lab 4

I am taking a GIS course this semester, and thought that working through this particular assignment here on my blog might make for a comprehensive report (that includes my maps, tables, and thoughts).  It’ll also allow my friends and family who  read this to understand what it means when I say “I am taking a GIS course this semester” or “Sorry I haven’t called for days (slight exaggeration).  I have been trapped in the mapping vortex”. 

-DISCLAIMER- what follows is an actual class assignment, so expect some maps that are not breathtaking as well as analytical errors.

Natural Disaster Risk Analysis

The goal of this Lab is to identify cities which are at HIGH RISK for natural disasters* and that are also in vital need of base map data.  At the end of this analysis, I will present a list of several cities which my data shows as being high risk and under mapped.

*Natural disaster here means one or more of the following:

drought (drought)

seismic (earthquakes and volcanoes)

hydro (cyclones, floods, and landslides)

My first goal is to illustrate cities that have base map data coming from a comparison of Open Street Maps and Google that also fall within regions considered to have substantial levels of multi-hazard risk by a project conducted by the World Bank & Columbia University.  The following map illustrates this by showing substantial risk areas as yellow striped polygons and base map data as blue points within the polygons.  

Map 1:

It is important now to spend some time understanding which cities are considered to have high and low levels of base map data.

The following graphs illustrate capital cities and their level of base map data from Google (in Graph 1) and OSM (in Graph 2). The rating scale here ranges from 0-5 with 0 being the cities with the least base data and 5 being the cities with the most base data. The blue bars again represent places covered to some extent by the OSM/Google data and designated as hazardous.

Graph 1

Graph 2

So, in using my original map and these graphs as an initial framework for illustrating what cities might benefit from more base data, I aim to give priority to cities that fall within these categories:

1) at risk of multiple hazards (as demonstrated by yellow-striped regional polygons)

2) have low levels of base-data that falls within the OSM/Google comparison (as demonstrated by blue points)

though I feel it is also important to analyze the data within the frame of mind that base data may be limited in areas of highest risk but that were simply not included in the OSM-Google data set.  Thus, one or multiple cities on my list may fall within this final category:

3) are at substantial risk of multiple hazards but do not have data regarding their mapping level (these cities will require further review in OSM to determine if we as a class think they would benefit from more base data).

3 Cities (among many I think Need Attention: 

3. Plymouth, Montserrat

Located in the Caribbean Sea beneath St. Kitts and Nevis.

- at risk of multiple hazards

- is one of the three lowest-ranked cities in terms of combined base data, receiving a score of -4.

2. Port Nono, Benin

Located on the West Coast of Africa

- at risk of multiple hazards

- has low levels of base data ranking Google Level: 0 and OSM Level: 1

1. Bissau, Guinea Bissau

Located on the West Coast of Africa

- at risk of multiple hazards

- has low levels of base data ranking Google Level: 1 and OSM Level: 0

Map 2

This map is on its way (the computer lab closed) the goal of this map is to show a global map zoomed into the portion that shows all three of my chosen locations (one being in the Caribbean, and two on the western coast of Africa).  It will show that they are in areas of high risk, and also that two of them have been included in the OSM/Google base map analysis.

This project is very introductory-level, and can benefit substantially from further time spent analyzing my data.  I know that the list of cities that could benefit from improved levels of data is far larger than the three I presented here. Through exporting my attribute tables to excel and removing all cities from the OSM v Google data set that are not at risk of natural disasters per the Columbia/World Bank data set, I could work with a data set more specific to our class objective and thus present a more comprehensive table of at-risk cities with low base data. 

Thus, this report is to be continued.

For those of you hoping to follow the situation in Syria