GIS4930 Module 4 - Final Presentation (Food Deserts PPT)

Weeks of working with open source software have come together finally. Working with QGIS, Tilemill, MapBox, and Leaflet has been quite an experience, and definitely was enough to develop preference for some software over others. The creation of a web map essentially from scratch was a challenge initially but easier to do as time and experience increased. Completing an analysis for food deserts in my hometown (Marshall, Michigan) was fun as well as interesting, since it produced some results which I had not expected. Along the way, much more was learned about food deserts, their impact, and potential solutions. One of my favorite finds was the USDA's Food Access Research Atlas which can be found here with additional explanatory information available here. The USDA's site enables anyone to investigate food deserts by census tract.

The culmination of the project was the assembling of all the information and experiences into a PowerPoint presentation complete with audio which was a new experience. A link to the presentation is included here, along with a screenshot and link to the web map created for this project.

Food Deserts in the City of Hospitality: An Analysis of Marshall, Michigan and Vicinity

(PowerPoint Presentation)

Food Deserts of Marshall, Michigan

(Web Map)

Screenshot of web map created with QGIS, Tilemill, MapBox, and Leaflet

For the time being that is enough about food deserts. Now it's time to move on to food desserts.

GIS4930 Module 4 - Tiling & Basemap Creation

Compilation of the grocery store layer included referencing several online sources plus verifying the information in Google Streetview. Online sources included Yellow Pages, Manta, and Google Maps which turned out not to be very reliable. Two grocery stores were shown in completely wrong locations. This was verified by personal knowledge and double-checking street view. Verifying grocery stores in the outlying rural areas involved more research to verify whether they were full-service or convenience stores (street view, store websites or descriptions, and reviews were used). Grocery stores outside the study area were included, because of the 10-mile marker used to define rural food deserts. (http://www.ers.usda.gov/data-products/food-access-research-atlas/documentation.aspx) Addresses for the grocery stores were tabulated in Excel, saved as csv file, and geocoded before performing the Near analysis to determine which urban census tract centroids exceeded the 1-mile marker and which rural census tract centroids exceeded the 10-mile distance from a grocery store.

Centroids of the census tracts were calculated to provide a comparison point for the one-mile marker food desert determinant (urban) and 10-mile marker (rural). Interestingly, the centroid for one of the rural census tracts landed inside one of the city census tracts. Both city census tracts were considered to be food deserts which surprised me. If one of the grocery stores had been located on the east end of town or farther south, at least one of the city census tracts probably would not have been classified as a food desert. The grocery stores are both located near the western edge of town (and extreme NW part of the larger urban census tract). Because of the town size and number of grocery stores, this type of analysis might not be as informative as anticipated. Perhaps an evaluation by census block would be more appropriate to really define the at-risk areas, or an assessment by residents’ age and mobility/transportation issues could be done. I was surprised that at least one of the two southern rural census tracts was not a food desert. However, the grocery stores in that area are of a smaller nature and do not carry stock nearly as comprehensive as a chain grocery store would. So while not technically food deserts, these rural areas might still be prone to a lower-quality diet unless supplemented with home-grown produce.

The process of creating a web map of food deserts included tiling the shapefiles described above. Tiling a shapefile enables more rapid loading of layers into a web map because the layers are displayed in smaller portions (tiles). In TileMill each of the data layers for this map was centered in a tight-fitting bounding box with a centralized marker. The zoom level was adjusted so that the resulting tiles would be 10MB. Each of these map layers had up to 19 tiles. The default world layer was removed by editing the code, and the symbology (colors, opacity) for the layers was adjusted as needed. Each layer was exported for use in Mapbox and by default was sent off to a folder not of my choosing. I opted to re-organize those layers into project folders prior to adding them to the map in Mapbox. The method of saving layers or projects in both TileMill and Mapbox was not very intuitive for me (save/don't save, save again, export, etc.), so several times some steps had to be repeated in order to ensure that I had the desired product. As so often happens, this was just enough exposure to new software to let me know how much there is to learn! The screenshot below shows the food deserts, census tract boundaries, and grocery stores added to the revised basemap in Mapbox.

Food Deserts in Marshall, Michigan
with Census Tract Boundaries and Grocery Stores
Created with TileMill and Mapbox

GIS4930 Module 4 - Tilemill & Leaflet

The exploration of opensource software for GIS purposes continued this week with the focus on Tilemill and Leaflet. Shapefiles were imported into Tilemill which already had a basemap in place. Arcmap was used to classify the food deserts by population (QGIS could have been used also) before selecting the color scheme in Color Brewer. This color scheme was entered into the Tilemill source code. The Leaflet code editing was terribly confusing, but with a lot of referencing previous instructions, double-checking coding changes, verifying paths, and re-reading of instructions, I made it through. OpenCage Geocoder was utilized to find latitude and longitude for specific addresses. Markers and shapes with popup text boxes were edited in the code. Layer groups were created and the ability to toggle layers on or off was coded. A geocoder was added to the source code for finding locations. Interestingly enough, it went to locations within a few blocks of the input addresses, but not to the exact address location specified. Its reliability indicates that it should be investigated further. Throughout the process of editing the Leaflet code, the text and html files were saved incrementally, keeping previous copies in case something really got messed up during the editing process. The gods of coding must have been with me this week though, because I made it through all the Tilemill and Leaflet coding without any errors. What a lovely surprise that was!

Here's the link to the final web map: Food Deserts of Southeast Escambia County which opens up this:

Food Deserts of Southeast Escambia County

GIS4930 Module 4 - Intro to QGIS (Prepare Week)

Prepare week for this module started off with an introduction to QGIS, an open source, desktop GIS. The similarities between QGIS and ArcGIS are significant, but there are several differences as well. QGIS did reveal a few quirks, one being that an added layer did not show up in its correct location, but what could possibly be bad about UWF being in the Bahamas at this time of year? UWF's location was easily corrected by removing and re-adding the layer to the map. A difficulty I encountered was not being able to find adequate help resources (in a timely manner) for issues I experienced while using QGIS. Adding and adjusting bar scales was much more involved in QGIS than it is in ArcGIS. The use of data frames in the composer was a particular challenge for me. Frequently my data frame would be void of information after changes were made elsewhere in the composer. Hopefully more experience will alleviate those issues.

All that being said, a definite advantage to QGIS is that it's free and available to anyone. This would be beneficial for small businesses and non-profits which do not have the financial resources available to invest in ArcGIS.

University of West Florida

(Introductory QGIS Map)

For this module's project, QGIS was used to perform several processes which have been utilized in ArcGIS. These included re-projection, clipping, selecting by attributes, selection by polygon (this turned out to be a challenge), and centroid generation. From this information, food deserts were calculated based on whether the centroids of the census blocks were within 1 mile of a grocery store.

Statistics were also generated in QGIS after using an edit session to delete unnecessary columns in the attribute table for the food desert shapefile and again for the non-food desert shapefile. The results of these processes are shown in the following map of Food Deserts and Non-Food Desert areas in Pensacola, Florida.

Food Deserts and Non-Food Deserts of Pensacola

Escambia County, Florida

GIS4930 Project 3 - Statistical Analysis with ArcGIS (Analyze)

The examination of various socio-economic factors in facilitating the ability to predict methamphetamine lab locations began with the Ordinary Least Squares (OLS) regression to help assure the creation of a good model. Initially all independent (explanatory) variables under consideration in this study were applied to the dependent variable (meth lab locations) as a group. The results were examined for redundancy, statistical significance, and a correlation between the independent and dependent variables. Following this critique of the results, the variable was evaluated for consideration of inclusion based on considered importance to the study. The goal of the OLS regression was a return of a high Adjusted R-Square value. Independent variables were removed individually in an effort to maximize the Adjusted R-Square value based on a critique of the results. This was followed by another run of the OLS tool. This process was reiterated numerous times until satisfactory results were achieved. A table of the final OLS results follows:

Final OLS Results Table

To inspect for bias, the OLS regression's Jarque-Bera Statistic Score was checked along with the histograms of a scatter plot matrix. Finally, the Standard Residual values for each census tract were added to the map for visual critique. Most of the two-county area was predicted accurately with this model as shown in the map below. Additional work would include resolving the high and low residuals.

Meth Lab Locations and Standard Residual Values
Charleston, West Virginia, with Kanawha and Putnam Counties

GIS4930 Project 3 - Statistical Analysis with ArcGIS (Prep)

Statistical Analysis is the focus of the third project in Special Topics. This particular study will analyze a variety of socio-economic factors to determine whether there is a connection to the location of illicit methamphetamine production labs in Kanawha and Putnam Counties, West Virginia. With 2013 being a record-setting year for meth lab busts in West Virginia (Eyre), law enforcement officials will appreciate having additional information that may help locate existing meth labs or prevent additional ones from starting business.

Preparation for the statistical analysis included creating new attribute fields for the census data and then using the field calculator to determine percentages of some data. Utilizing percentages rather than raw numbers for the census tracts will provide more equalized comparison of data. Python scripting was also used to add additional attribute fields and perform related calculations in a more efficient manner. Drug Enforcement Administration (DEA) meth lab bust information with geocoded data was joined to the Census data as well. Tidying up the attribute table was done by turning off fields which wouldn't be used for the statistical analysis. The following basemap of the study area was created. It includes cities and towns which have had meth lab busts in the past.

Socio-Economic Factors and Meth Lab Locations

Study Area: Kanawha and Putnam Counties, WV

Resource:
Eyre, E. (2014, February 19). W.Va. Senate OKs anti-meth prescription bill. Charleston Gazette. Retrieved from http://wvpress.org/news/w-va-senate-oks-anti-meth-prescription-bill/

GIS4930 Module 2 - Mountain Top Removal, Appalachia Coal Region (Report)

The  Mountain Top Removal (MTR) method of coal mining in the Appalachian Mountains has been shown to impact the surrounding hydrology (Petrequin). As part of the study of applying GIS to MTR impacts, this project included creating stream and basin features. This was done by mosaicking four DEMs into one layer and then applying several Spatial Analyst Hydrology tools.

Analyzing the impact of the MTR method of coal mining on an area involved comparing imagery from two different time periods. ArcMap and ERDAS Imagine were both used with 2005 imagery to develop a signature file which was then applied to 2010 imagery to develop a map of MTR areas.

Only a portion of the 2010 data was analyzed by Group 3 for SkyTruth. The 7 bands of the 2010 Landsat imagery were consolidated using the Composite Bands tool, and the imagery was clipped to the group's study area in ArcMap. A layer with 50 classes was created in ERDAS Imagine using the Unsupervised Classification tool. Areas which appeared to be part of MTR were classed and symbolized by color accordingly with the remaining areas being classed as NonMTR. Many objects such as stream or river banks, buildings, and roads have identical spectral signatures as MTR and were included initially. They were removed from the MTR features later. In ArcMap, the classified image was reclassified with MTR being assigned a value of 1 and all other Class Names assigned blank values.

From this information, the MTR raster was converted to polygons. MTR features within 400 meters of major rivers or highways were removed from the MTR polygon layer, as were those within 50 meters of streets and other rivers. Features smaller than 40 acres were removed as well. An accuracy assessment was done with a result of 96.7%, and a comparison against the 2005 dataset was made. There was an overall decrease in acreage attributed to MTR from 2005 to 2010, but the data needs to be critiqued further. The 2005 dataset included features of less than 40 acres, while the 2010 data was restricted to features containing more than 40 acres. The 2005 dataset had more than 8,000 features, while the 2010 dataset had fewer than 500 due to the acreage restriction.

A layer package was created with this data and submitted to the group leader for compilation into one dataset for the group's study area, and a map service to present the group findings online was created. This map service was used to create an online map. A soils runoff classification layer was also added to the map. This additional data was selected because runoff from MTR sites impacts the surrounding areas. The online map can be viewed here, although the soils layer is not available to all viewers.

Resource:

Petrequin, M. (2012). Hydrological Impacts of Mountaintop Removal in Appalachia: History and Solutions (Colorado School of Mines, Department of Environmental Science and Engineering).  Retrieved from uwf.edu.

GIS4930 Module 2 - Mountain Top Removal, Appalachia Coal Region (Analyze)

Assessing the impact of Mountain Top Removal (MTR) mining on an area requires comparing imagery of that area taken after a period of time has passed. The Analyze portion of this module helped develop the skills to do just that. Utilizing data from SkyTruth, 2005 Landsat imagery was used in ArcMap to select training samples in known MTR polygons as well as in nonMTR areas. After changing class names, values, and colors for the samples they were compared for overlaps, and a signature file was created to be used in Supervised Classification. The same data was used in Erdas with the "Grow" tool to select several samples to add to the Signature Editor table.

The class was split into several different work groups of 3-4 students to analyze the 2010 data for SkyTruth. Each group was responsible for 2 to 4 images. After the preliminary exercise, the 2010 imagery was prepared by first using the Composite Bands tool to consolidate the 7 bands of the Landsat file and then clipping the imagery to the group's study area in ArcMap. In Erdas the Unsupervised Classification tool was used to create a layer with 50 classes. Areas which seemed to be part of MTR were classed and symbolized by color accordingly. Distinguishing the MTR areas was difficult for me. The remaining areas were classed as NonMTR and symbolized by a different color. Because identical spectral signatures can be assigned to very different types of features, many objects which are not MTR areas are included in the MTR class. This will be accounted for at a later time.

Back in ArcMap, the classified image was reclassified with MTR being assigned a value of 1 and all other Class Names assigned blank values. The resulting image is shown here:

2010 Landsat Imagery Classified for MTR
(NOTE: This image includes some nonMTR features with the same spectral signature as MTR.)

From this information, the MTR raster will need to be converted to polygons and smaller MTR areas removed. Buffers will be created around roads and rivers, and MTR areas within the buffers will be removed. An accuracy assessment will be done as well as a comparison with the 2005 dataset. Each group member will submit packaged data to the group leader for compilation into one dataset for the study area. Finally a map service to present the group findings online will be created. Look for this next week!

GIS4930 Module 2 - Mountain Top Removal, Appalachia Coal Region (Prepare)

This was a two-part lab preparing for a project concerning the coal-mining process of mountain top removal in the Appalachian Mountains of West Virginia. The first part of the lab involved learning about LiDAR as well as creating and viewing LAS datasets in ArcMap. Different point symbology and different filters impact how useful the information can be.

The second part of the lab involved creating stream and basin features from USGS Digital Elevation Models (DEMs). A mosaic of four DEMs which Group 3 will be using this module was created and clipped to the extent of the class's study area. Several tools of the Hydrology toolset of Spatial Analyst were used then:

1. Fill tool to fill in low spots/pooling areas

2. Flow Direction tool to assign values to each pixel indicating direction of flow across the individual pixels 

3. Flow Accumulation tool to indicate the number of other cells flowing into each cell (to show possible streams)

4. Con tool to eliminate features which likely are not streams

6. Stream to Feature tool to create streams layer

7. Basin tool to create watershed layer

The created streams and basin layers are depicted in this map:

Group 3 - Streams and Basins

Appalachian Mountains

GIS4930 Network Analyst - Report Week

Producing maps with the routes created last week was the focus of this week's lab. The patient evacuation maps were inserted into a pamphlet template which included information for the patients and their families. Each of the emergency supply routes to the evacuation shelters was shown on its own sheet which included turn-by-turn driving instructions for the National Guard crew members and other emergency workers. These maps were specifically created in grayscale for ease of reproduction as requested; following that same thought process, the maps were created in an 8 1/2 x 11 format for easier handling, storage, and reproduction. 

Emergency Supply Delivery Route Map
with Turn-by-Turn Directions

The map shown here is representative of the supply route maps. It was not a favorite to complete, but is included as an honor for having the distinction of providing plenty of opportunities for me to learn from my mistakes. I feel so much smarter already. There were two particular challenges. One challenge encountered with this part of the project was trying to ensure that there was suitable contrast between the different map elements. The other challenge was fitting all the required information on a single sheet and still retaining legibility.

The final map was created for media distribution to the public. Because the map was requested for use in print media as well as on television, simplicity was expected. The map shows the shelter locations with primary streets along with the addresses of the evacuation shelters and some miscellaneous shelter-related information obtained from the City of Tampa and Hillsborough County. The analysis aspect of this map was completed in ArcMap using Network Analyst and the map was finished using Adobe Illustrator.

The entire project from start to finish was very comprehensive, very practical, and very educational.

GIS4930 Network Analyst - Analyze Week

The analysis portion of the Tampa Hurricane Planning project built on the information prepared in the previous module. The map shows routes from the National Guard Armory to three schools serving as hurricane shelters for emergency supply distribution runs and the routes from Tampa General Hospital to Memorial Hospital and St. Joseph’s Hospital for the evacuation of patients prior to the hurricane’s arrival. The map also outlines the areas to be served by the three emergency shelters (identified on the map).

Tampa Hurricane Planning - Routes

To prepare this map, a new network dataset was created from the streets layer prepared earlier with attributes selected to create the fastest, simplest routes rather than shortest-distance routes with flooded roads closed to the public but possibly traversable by emergency vehicles.

The areas to be served by each of the storm shelters were created in Network Analyst with values selected to avoid accidentally excluding any of the study areas from the analysis and to clearly identify only one storm shelter for any area.

As done in the preparation lab, a map package was created for the map and zipped together into one folder with the metadata files, routing map image, and the process summary. Again, a sample email detailing this information and process was created as if a co-worker would be completing the project. The email included the zipped folder, details of what the City of Tampa had requested, what had been accomplished during this part of the project, and what needed to be completed yet. Items to completed next are different information brochures along with a map and supply list for emergency supply distribution. These will be done during the Report portion of the project.

GIS4930 Network Analyst - Prepare Week

The focus of this module was preparation of data to be used for the production of a map and public information brochure by the City of Tampa in preparation for a hurricane. The City is especially interested in routing for evacuation, emergency access, and post-hurricane supplies, taking into consideration road closures. Elevation zones for the basemap were produced from a Digital Elevation Model which was classified into distinct elevation classes and symbolized. The flood zone for this hurricane (expected surge of 5.5 feet) is symbolized in red to stand out. Higher elevations are shown in other colors with 1-foot increments. Shelters, hospitals, fire and police stations, the National Guard armory, and roads are shown on the basemap as well. Symbols for facilities were shaped distinctly different from each other and generally included an associated letter designation to ensure that there was no confusion caused by color blindness or lack of familiarity. This would be important to anyone already experiencing the duress of an evacuation before an impending disaster.

Tampa Hurricane Planning Basemap

The streets layer was prepared for use as a network dataset with the addition of attribute fields for length in miles, travel time in seconds, and whether the road is in the flood zone. Interstates received the "not flooded" attribute because they are elevated. This information will be utilized in the next module.

A map package was created for the map and zipped together into one folder with the metadata files, basemap image, and the process summary. A sample email detailing this information was created as if a co-worker would be completing the project. The email included the zipped folder, details of what the City of Tampa had requested, what had been accomplished to date, and what needed to be completed yet.