Undergraduate Geospatial Python Projects

earthballThis week my GIS Programming students presented their programming projects to ESRI. First, I cannot say enough to thank ESRI for taking time out of their schedule to meet with our students – the staff was helpful, encouraging, and provided great feedback to the students – what an honor it was to get their feedback.  I am so thankful to be a part of a GIS community that is so supportive of one another.

Now, this was a really special class of undergraduates – and some of them were part of that special group of students that presented their research at an undergraduate conference.  It was small, so we could do some really cool things.  In fact, in the middle of the semester, the students wrote a paper comparing the geocoding accuracies of Google Maps and the United States Census Bureau.

Things were going so well that I decided in lieu of a final exam, we expanded their final projects a little more, and arranged for the staff at ESRI Charlotte and ESRI Redlands to join us on a WebEx that included demonstrations and a code walk-through.  Below are each students’ presentation, and some of the Q&A from ESRI:

noahNoah Krach.  Noah is an amazing undergraduate.  Recently, we lost one of our graduate research associates, and Noah stepped in to provide technical support on a National Science Foundation project in Lake Victoria.  Without missing a beat, Noah was all over the project, and he used his time in my class to create an Arcpy tool to extract, translate, and load (ETL) gigabytes of Landsat imagery.  This tool does a lot, and I can’t even begin to describe all he did, you’ll simply have to watch and learn.

Check out his video, and you’ll see why we are so excited that Noah will be around for another semester.

cc

Caitlin Curry.  If you follow my blog, you’ve already met Caitlin.  She finished her summer internship I told you about, and during the middle of it, her boss wrote us to say what an excellent worker she was (he prefaced his email by saying he never does that, but was so impressed with Caitlin, he had to let us know).  We are impressed with Caitlin, too.  And, as I have now grown to expect, Caitlin did an amazing job with another ETL type tool using Arcpy, where she downloaded, unzipped, and processed earthquake data and critical infrastructure.

I did a lot of emergency response work with earthquakes in a previous life, and what Caitlin did here would have been so useful.  I think you will enjoy seeing how she integrated many different Python packages with Arcpy to provide an early warning application for emergency responders.  And just as a heads-up, Caitlin uses Python to download everything while the script is running – so you just give the script to a user and it works without any operator knowledge of the underlying data = really cool, and efficient.

mb

Matthew Bucklew.  After my first lecture this semester, Matt told me he built his own computer this summer – just for fun.  So, I knew he wasn’t your ordinary  geographer – he likes to try new things, and if something is done in a conventional way, Matt is going to try and be more innovative.  Matt created a great Arcpy application to locate renewable engery stations needed by automobiles.  His Python scripts use ArcGIS for analysis, but at the same time, seamlessly brings in the Google APIs to provide directions to the nearest locations.  For good measure, he also brings in other packages like heapq.

At the moment, Matt’s program works on a desktop, but his hope it to turn this application into a cloud based solution for use with mobile phones.  Keep an eye out for what Matt comes up with, and if you watch this, you’ll see it is an excellent tutorial on how to mash up bunches of Python packages with Arcpy.

jmJessica Molnar.  Like Caitlin, Jessica is another student you’ve seen before.  She’s got such a big heart, and is always looking for ways to apply GIS to humanitarian and ecological solutions.  In this project, Jessica created an Arcpy application to identify locations for community gardens in Baltimore City with special consideration for locations within food deserts, near churches and schools, and on suitable soils for growing food.  Jessica’s program also found those locations that were already owned by the City, but were vacant.  Let’s hope the City makes use of this to build a more beautiful Baltimore (BTW, Jessica wrote her program to work in any location in the State of Maryland, so any community can use this tool!).  I think Jessica may eventually roll this into a cloud based solution – hey Jessica, I think we found a project for graduate school!

 

jtJohn Tilghman.  John’s family owns an orthodontist practice, and John decided to use PostGRES/PostGIS along with a number of other different Python packages to perform market area analysis.  John integrated PostGRES, Google, and the Pygal libraries to create the first stages of a geodashboard to assess the effectiveness of marketing strategies, and other metrics.  In the video, you’ll also see how he created a distance decay algorithm in SQL to determine at what point customers drop off from visiting the practice.  With just a little bit of information (addresses and marketing strategies), John was able to extract a ton of business information – in fact, our guests from ESRI were surprised the John wasn’t already a business major!

This is an excellent presentation to watch for those of you who are interested in using Python with Open Source GIS – you’ll learn how to integrate FOSS4g and Python for a business analytics tool.

 

jyJosh Young.  Josh created an Arcpy script to assemble tons of location based data that might be useful for someone thinking about moving to a particular location.  Now, in Josh’s case, he chose location based data he deemed important for the neighborhood (download speeds, elementary school, crime statistics, distance to the downtown, etc.).  But ultimately, what Josh has shown us is how to create a template that integrates multiple Python packages and online data to provide very useful information.

It would be so easy to take Josh’s work and roll it into a site specific location-based analysis engine.  In fact, one of the people watching Josh’s presentation mentioned that he was moving, and saw how useful this could be for a community.  The best part of it is that Josh did it with all freely available online data for the State of Maryland, so any community can spin this up into a cloud-based solution.

 

image2

Robbie Stancil.  Robbie is our only non-geography major.  You’ve met him before when he worked with me on a National Science Foundation project to use Spatial Hadoop.  Like John, Robbie’s project used Postgres/PostGIS and the Google API to do something quite interesting: he created a mesh of points over community to determine how far the Google API will search in order to find a property address, and compared the concave hull of each series of points for an address to the actual property parcel.  This project got us thinking about some very creative uses – you’ll have to watch it until the end to see the interesting things we came up with.

 

Again, I have to give a huge shout out to the ESRI staff – they were wonderful guests, and really excellent mentors during the Q&A. As these students get ready to graduate in May, I know they will make excellent employees or graduate students – the future is really bright for them. If you are in academia, I hope that you are inspired to expect the very best of your students as I do, and you’ll be so pleased to see what they are capable of doing.

want to learn how to program geospatial solutions like these students? Check out the geospatial courses at gisadvisor.com.   

Advanced Editing Capabilities with PostGIS

Today I want to introduce you to another one of my really great undergraduate students: Jeff Scarmazzi.  Jeff participated in the SUSRC conference that I posted about previously, but unfortunately, his photo and poster got deleted from my camera, so I never posted information about his project – I may eventually get around to posting Jeff’s poster, but for now it will have to wait.

 

Jeff is an exceptional student, and at this point can run circles around me with PostGIS, Leaflet, and Postgres – at the moment, I am still reserving the right to say that my spatial SQL is better than his, but that may not be true.  And, now that Jeff secured a prestigious internship with NASA over the summer, once the Fall rolls around it definitely won’t be true!

What I wanted to show you was a short presentation that Jeff gave on his work with Postgres and PostGIS in my Advanced GIS class.  For this project I had students create a geodatabase with ArcGIS that includes domains, subtypes, topology rules, etc.  Jeff decided to extend his work on this assignment and recreate the same kind of features using Postgres and PostGIS.  This presentation is just a small piece of what Jeff and his fellow students (Austin Barefoot, Carl Flint, and Jordan Roose) created, and focuses specifically on writing triggers and creating roles to automate and quality check digitizing procedures using Postgres.

The presentation is only about 15 minutes long, and you’ll be amazed to see what you can do with PostGIS to creating robust editing rules.

Warning Shameless plug: In case you wanted to learn how to implement PostGIS in an enterprise setting, check out my online course here.  Also, I have another course on how to write spatial SQL with PostGIS here.  

Open Source GIS Course at UMD – the Syllabus

UMD provided me with enrollment numbers, and we are getting a pretty full class.  However, a few people on the “outside” have asked for a little more depth into what will be covered.  So, I put together a quick outline on each of our class meetings.

Keep in mind, in addition to the class meetings, there will be weekly discussions and short assignments.  Each class session is 2.5 hours long.  Those sessions colored orange are classes that I will be on the campus live (please note that this is tentative, and the dates may slide around a bit), giving the lecture and assisting students in the lab.  The other class sessions are a combination of online video lectures or in laboratory exercises.  For those with full time jobs, most of the material can be performed at home on you own devices, although being in the lab with the TA will be helpful.

Again, please feel free to email  geog-gis@umd.edu to get more information about signing up for the class. Continue reading

k-Nearest Neighbor with PostGRES

Well, even though you didn’t ask for it, I decided to create a k-nearest neighbor analysis in PostGRES/PostGIS.  This has lots of potential: find the 5 nearest ATMs to every fast food restuarant; find the 3 closest medical clinics to each school; or find the 10 closest bars to each college.

Now, the LIMIT clause does allow us to find the k closest features to a particular feature.  But in this case, I’m not interested in a particular feature, I want the results for all features.

To do this in Postgres, I’m going to show you two things that I haven’t used until today: rank() and partition. Continue reading

Trade Area Analysis in Postgres / PostGIS

How I used SQL to generate an accumulated sum of spatial data

My friend Tomas does work with business analytics, and wanted to find a way to perform trade area analysis.  He had a bunch of stores, and a map of census areas with populations.  What he wanted to figure out was:

how far do I have to radiate out from each store before I hit 5,000 customers

So for each store, he wanted to basically generate concentric buffers and sum up the population of the census areas before he hit 5,000.  Once he found the census areas, he also wanted to generate a convex hull to show the area.  ESRI has a really nice tool that performs this as part of their business analyst package, called threshold trade areas.

Check it out, it seems pretty cool.

Well, to help my friend I was thinking that I would determine the distance from every store to every census area, and then write a script with a giant for loop to iterate through each record, accumulating the results (of course, I would have to integrate some kind of do..while loop and an if…then to make sure I accumulated different counts for each store until I hit the threshold I wanted.  At that point I began asking myself so, how good of friend is Tomas?

What I did instead was write an SQL script to do it. I’ve color coded it below to explain what I was doing….

SELECT ST_ConvexHull(ST_Collect(g)) as geometry, max(sumpop) as sumpop, name
INTO tradearea
FROM
 (
  SELECT a.name, SUM(a.totpop) AS sumpop, ST_Collect(a.geometry) as g
  FROM 
     (SELECT stores.name, censusunit.totpop, censusunit.geometry,
      ST_Distance(censusunit.geometry,stores.geometry) as dist
      FROM stores, censusunit
      ) AS a,
     (SELECT stores.name, censusunit.totpop, censusunit.geometry,
      ST_Distance(censusunit.geometry,stores.geometry) as dist
      FROM stores, censusunit
      ) AS b
  WHERE a.name = b.name
  AND a.dist <= b.dist
  GROUP BY a.name, b.dist
  ) AS T1
WHERE sumpop < 5000
GROUP BY name

The middle portion in orange collects the names of the stores, the population of the census areas, the distance between each store and each census area, and the geometry of the census areas for each combination of stores and census areas.  So, if you have 5 stores and 1,000 census areas, you would have a table with 5,000 rows: Continue reading

SQL and ogr to create 4D lines

A friend recently asked me to help him generate polyline shapefiles with Z and M values that he could deliver to a customer.  The problem was, the software he was using supported the import of Z and M values, but did not support the export of those files.  The other problem was, he has zillions of data tables that he needed to export!

Fortunately, PostGIS allows us to create Polyline ZM data rather easily.  The next part was to figure out how to get it exported to a shapefile.  So, here goes:

Assume you have a table with 4 columns: lineID, Z, M, and geometry.  The geometry field is a point feature that represent vertices on a line.  The lineID separates all those points by which line they are part of, and the Z and M values are, well, Z and M.

Make the Polyline ZM

The SQL command to create a 4D line is:

SELECT  lineid, 
ST_MakeLine(ST_MakePoint(ST_X(geometry), 
                         ST_Y(geometry), z, m)
            ) AS g
FROM zpts
GROUP BY lineid;

In this case, I am making a line (ST_MakeLine) as a series of X (ST_X), Y (ST_Y), Z, and M values.  Grouping the results of the query by the lineid allows me to create a line for each set of points. Continue reading

When More is Less…. lessons from processing large data files

My good friend Stuart Hamilton gave me a fun conundrum to try out. He has a file of province boundaries (400 areas) and lidar derived mangrove locations (37 million points – 2.2GB in size). He wants to find the number of mangroves that are contained in each area.  He also want to know which country a mangrove location is in.  An overview of the area is here:

allstu

but, as you zoom in, you can see that there are a tremendous number of points:

stuzoom

The problem

You would think that overlaying 37 million points with 400 polygons wouldn’t be too much trouble – but, it was.  Big time trouble.  In fact, after running for days in ArcGIS, Manifold GIS, PostGRES/PostGIS, and spatial Hadoop, it simply would not complete. Continue reading

New Books – How do I do that in PostGIS, How do I do that in Manifold SQL

I have two new books out – How do I do that in PostGISand How do I do that in Manifold SQL.  

From the back cover of How do I do that in PostGIS:

For those who are unsure if SQL is a sufficient language for performing GIS tasks, this book is for you. This guide follows the topic headings from the book How do I do that in ArcGIS/Manifold, as a way to illustrate the capabilities of the PostGIS SQL engine for accomplishing classic GIS tasks. With this book as a resource, users will be able to perform many classic GIS functions using nothing but SQL.

Continue reading

A Poor Man’s Parallel Processor for GIS

In addition to SQL, I also am interested in processing large volumes of spatial data.  One of the newest rages in “big data” is Hadoop.  According to Wikipedia:

Apache Hadoop is an open-source software framework written in Java for distributed storage and distributed processing of very large data sets on computer clusters built from commodity hardware.

One way this is implemented is a programming model called MapReduce.  Don’t get too excited, it doesn’t have anything to do with maps or GIS – but, it is very clever and powerful for certain types of problems.  The concept is if you have a really large dataset, you divide and conquer that dataset in a number of steps.  For example, say we wanted to know all the people with the name “John” in the phonebook, and say we had 26 computers in a cluster – we might solve this by:

1.  Use each computer (1-26) to find all the “Johns” for the first letter in the last name (A-Z).  That way, you have effectively broken the problem into 26 smaller units.

2.  Once each computer has counted up the number of Johns, you have reduced the dataset (hence, MapReduce) to 26 variables.

3.  Now, count up the total of the 26 variables.

That is an oversimplified version of course, but it helps to illustrate what we want to do.  I understand that the University of Minnesota has created a set of functions called SpatialHadoop.  I want to test this over the summer, but for now I decided to create my own poor man’s version using PostGRES. Continue reading