tianshengs.github.io

Follow me on GitHub

QGIS Modeling : a use case of San Juan, Puerto Rico

Introduction

Measuring distance and direction is a common theme in GIS analysis. The concept of distance and direction, moreover, can be especially helpful for us to understand the organization of urban landscapes (learn more). While one can easily use GIS tools in ArcGIS to calculate the direction and distance from a point, with the advance of open source technologies, one can create a self-defined model to calculate the direction and distance from city center. For this project, I have created my own direction and distance calculator model. The model is then used to analyze the percentage of Hispanic people in San Juan metropolitan area of Puerto Rico. This is my introduction into open source GIS and the use of QGIS for GIS analysis.

Study Area

lede puerto-rico-1

San Juan is the largest city of Puerto Rico. Located in the North coast of Peurto Rico, it is a lovely old city with around 400,000 people in its core and 2 million in its metropolitan area. Although the population of Puerto Rico is highly hispanic, we might expect to see some more diversity in this large metropolitan city. Specifically, it is interesting to see how the percentage of Hispanic people changes by distance or direction from central San Juan.

Data

I downloaded the Census Tracts shapefile and County Within Urban Area shapefile of Puerto Rico from the United States Census Bureau and information of the Hispanic or Latino origin by race per census tract in Puerto Rico from American FactFinder. The County Within Urban Area shapefile is used to crop out the San Juan study area from the Census Tracts shapefile.

Here is the geopackage of data for the case study, including the original data that I downloaded and the shapefiles that I created through the process.

Models and Method

a. Creating model using field calculator to calculate distance and direction

Here is my initial model using field calculator to calculate distance and direction. The user can define a city center as a point or selected features, from which a point with the medium coordinates of the centroids of all selected features can be calculated. Finally, the direction and distance from the city center point to the centroid of each feature in the shapefile is calculated with new fields in the attribute table (distance and direction) of the output. Finally, I created a documentation for the model that I created.

initial

This is the specific field calculator query I used to calculate distance:

distance(centroid($geometry), 
make_point(  @Mean_coordinate_s__OUTPUT_maxx,  @Mean_coordinate_s__OUTPUT_maxy   ))

This is the specific field calculator query I used to calculate direction:

degrees( azimuth(  make_point(  @Mean_coordinate_s__OUTPUT_maxx,  @Mean_coordinate_s__OUTPUT_maxy ), centroid($geometry)))

b. Update the old model with SQL quries

Then, I modified and updated the model by doing the following:

  • Added a Help webpage to my model.

    The “Documentation help URL” is especially helpful, which directed the users to my Github page. I also added one for my original model.

  • Used Execute SQL to calculate distance. 
    SELECT*, st_distance(centroid(st_transform(geometry, 4326)), (Select st_transform(geometry, 4326) from input1), TRUE) as  DistSQL
FROM input2
  • Used CASE statement to classify the direction data into N, E, S, W. The CASE statement can be helpful to calculate fields based on a set of one or more conditions. For my model, I used the CASE statement to classify the direction data into N, E, S, W based on the direction of the data: 
    CASE
WHEN attribute(concat(@fieldname, 'Dir')) >= 225 AND attribute(concat(@fieldname, 'Dir')) < 315 THEN 'W'
WHEN attribute(concat(@fieldname, 'Dir')) >=135 AND attribute(concat(@fieldname, 'Dir')) < 225 THEN 'S'
WHEN attribute(concat(@fieldname, 'Dir')) >= 45 AND attribute(concat(@fieldname, 'Dir')) < 135 THEN 'E'
ELSE 'N'
END
  • Applied the transform function of SQL to calculate azimuth while transforming it to the World Mercator projection (EPSG:54004). 
    degrees( azimuth( 

transform(
make_point(@citycenter2_maxx ,  @citycenter2_maxy  ), 
layer_property(@citycenter2 ,'crs'), 
'EPSG:54004'
), 

centroid(
transform(
$geometry, 
layer_property(@inputfeatures, 'crs'),
'EPSG:54004'))

))

The problem that I faced is that the Execute SQL only worked with a single point, but not with selected feature data. Therefore, to get a single point from selected features, the user may now need to just run centroids and mean coordinates separately before they use my Model to calculate.

After creating the model, I opened QGIS 3.8.1 to do the following:

  • Added the Hispanic or Latino origin by race per census tract in Puerto Rico CSV file into QGIS. Layer->Add Layer->Add Delimited Text Layer is the best way to add CSV files into QGIS because it tries to detect data types correctly rather than assuming everything is a text string. I imported the layer with no geometry as a comma-delimited file.

  • Added the Census Tract shapefile and County within Urban Area shapefile of Puerto Rico into QGIS.

  • Selected the San Juan Metropolitan Area from the Census Tract shapefile and Use the QGIS function Select by Location to crop out the census tracts located within San Juan Metropolitan Area for analysis.

    Here is the cropped study area:

    Captured

  • Joined census data to Census Tract shapefile by opening the properties of the shapefile layer and using the Joins menu. To keep the data neat, I only selected the fields that I want to join: latinx, the column with the number of latino population and total_pop in each census tract, the column with the total number of population in the census tract. I joined the census data to Census Tract shapefile by the corresponding GEOID of each census tract, which is a unique feature in both files.

  • Calculated the percentage of Latino population in each census tract and stored the value in a new column using field calculator.

  • Used the model to calculate the direction and distance from the center. I manually selected the three census tracts as city center for my analysis.

  • Created Data Plotly plots as HTML to show my results. Data Plotly plots are dynamic code written for the internet and can be created using a Plugin in QGIS. For my analysis of the percentage of Hispanic population in San Juan Metropolitan Area, I created a scatterplot of distance and polar plot of direction.

Results

This is the choropleth map showing the percentage of Hispanic within each census tract: sanjuan

This is a graph that I created using Data Plotly that illustrates the relationship between percentage of Hispanic people and distance from central San Juan. As we can see from the graph, while a majority of the census tracts have over 95% of Hispanic population, the few census tracts relatively closer to San Juan has slightly more non-Hispanic population. One census tract of the city center, moreover, has only 68% of Hispanic population.

This is a graph that I created using Data Plotly that illustrates the relationship between percentage of Hispanic people and direction from central San Juan. Because central San Juan is located on a small island in the Northern coast of Puerto Rico, almost all of the other census tracts are located to the south and east of central San Juan. Most of the census tracts to the South and Southeast has over 95% of Hispanic population, while a few coastal census tracts to the east of San Juan have a slightly lower percentage.

Overall, the analysis has shown that the although the majority of San Juan Metropolitan area is pretty homogeneous, the city center and a few census tracts along the coast stand out as regions with slightly more non-Hispanic population.

Discussion: GIS: A tool or science?

One big debate in the world of geography is whether it is a tool or a science. This dichotomy can be seen in different institutions. In China and many countries in Europe, a major in GIS or Remote Sensing is very common. ETH Zurich, for example, despite not having a geography major, has a Geomatics major offering courses in GIS, Romote Sensing and GPS. American institutions, on the other hand, seldom have a major in GIS or Remote Sensing. Instead, it is offered in the geography department, or used as a tool in other disciplines such as Economics.

This question make me think whether Computer Science, my other major, is a tool or a science. As computer science skills getting increasingly important, more and more students from other majors also start to learn how to code and use coding and programming tools to get a better job in the market. Although the competitive job market nowadays emphasize on the “tool” section of coding, computer science indeed has its “science” side: understand background theories, designing algorithms, and writing proofs. In Middlebury College, I like the balanced approach to both theory and practice of computer science teaching, yet I have to admit that many students and parents, with the emphasis of getting high-paying jobs, emphasize on the “tool” side of computer science. However, to have a good computer science background, one should not totally seperate the two sides of computer science. Without writing codes, we will not be able to achieve any specific goal. Similarly, without runtime analysis, proofs and rigid algorithms, we will not design stable and convincing programs and the programming tool itself is simply useless.

Therefore, it is unfortunate and unnecessary to seperate GIS as a “tool” or a “science”. Much like computer science, it can be both and indeed either side of GIS is indispensable to our success in the discipline. Knowing only how to do “GIS” is not enough, one also need rigid geography and geology background because in the end we use GIS to solve a problem. With a study question in mind, GIS analysis can help produce result and make progress alongside other methods. This project, for example, shows the two sides of the coin: while building the GIS model is solely quantitative, the result can be used to understand urban questions that is built into researches and theories talked about in this article.

St. Martin and Wing (2007) has provided a approach to not view GIS simply as a tool or science, but as a discourse, an interdiscipline set of practices and understandings. While I resonate with their idea, I find that the term “discourse” is a bit vague. We do not need to create another term to define what GIS really is as long as we equally value the importance of GIS with other approaches such as a humanitarian approach or a statistical approach to our understanding of geographic and spatial phenomenon. Indeed, I have taken various different courses in the discipline of geography in Middlebury College, ranging from GIS to Urban Geography, which uses mostly statistical methods, and Environmental Change of Latin America, which is a more human geography approach to our understanding of geographic phenomenon. I appreciate that the geography department in Middlebury College arranges courses and professors each with different focus, all of which provide a different yet important perspective into the research of geography.

This debate over GIS as a single entity also has the danger of producing a subject that is static, invariable and a singular technology (St. Martin and Wing , 2007). Indeed, ESRI has dominated the market since the 1980s and the use of the software is extremely expensive. This use of exclusive software to quantified, classified and optimized data to produce accurate and standard results can create technocracy and has posed a threat to the inclusiveness (Sieber, 2004). However, the development of open source GIS can provide a good way for a more engaged approach that Sieber (2004) suggested and heterdox GIS approach which St. Martin and Wing (2007) suggested are “open with alignments and experimentation”. The development of open source software such as QGIS is free and people from all backgrounds can contribute to the construction of GIS, from example, to create a model, and people can also help each other to debug on online platform and forum. This collaborative approach of GIS challenges the hierarchical and institutional nature of GIS in the past and can create an environment that is more open, well-represented and innovative.

References:

Martin, K. St., & Wing, J. (2007). The Discourse and Discipline of GIS. Cartographica: The International Journal for Geographic Information and Geovisualization, 42(3), 235–248. https://doi.org/10.3138/carto.42.3.235

Sieber, R. (2004). Rewiring for a GIS/2. Cartographica: The International Journal for Geographic Information and Geovisualization, 39(1), 25–39. https://doi.org/10.3138/T6U8-171M-452W-516R

Back to Main Page