Using CartoBuilder As a TARDIS For Exploring Geologic Relationships

Using CartoBuilder As a TARDIS For Exploring Geologic Relationships

If you're a Dr. Who Fan I apologize for the potential clickbait but geoscientists also frequently have to navigate the space-time continuum, sadly without the luxury of a time machine. However after receiving a few emails from CARTO I finally decided (and had the time) to take CartoBuilder for a spin and I think some of the new functionality could be used in interesting ways to help teach geologic concepts and relationships. After watching the introductory video on widgets I immediately wondered how CartoBuilder could be used to explore the recent earthquake swarms in Oklahoma. A number of studies (including this nice summary) have concluded the increase in seismicity is linked to subsurface fluid injections associated with increased fracking activities. However it's pretty difficult (Google's results aren't compelling) to find a nice looking map summarizing this seismic anomaly. The only visual I really liked was created by Dan Nguyen who used a nice Tufte-inspired, small mutiples approach:

Source: http://blog.danwin.com/oklahoma-earthquakes-r-ggplot2/

If you're interested in teaching or exploring your own data with R, Dan also provides a nice Github repository with all the necessary code and R notebooks to explore how he created both this image and his earthquake animation. For research purposes there's no question R has a more extensive suite of user-supported packages and is without question more customizable, but for out-of-the-box simplicity in teaching environments I think CartoBuilder is really exciting.

---

Unfortunately I didn't discover Dan's repo until after I finished my map or I would have used some of his data. To create my map I used the following data:

• I used the USGS Earthquake Catalog to download all the earthquakes within Oklahoma since 1980 of magnitude 1.5 or greater, which resulted in this CSV file.
• I downloaded and merged shapefiles of the gas/oil plays from the Energy Information Administration.
• I downloaded shapefile from the Oklahoma Geological Survey depicting faults.
• I used a file of gas/oil wells I already had from a previous project

NOTE: Click here or on the map to view in a new window. The layout works better in a full-size window rather than as an embedded element.

In the static map above you can see the earthquakes as orange dots and since the timeline widget below the map is using a date field within the earthquake table, the vertical bars are also orange. There are three widgets on the right side of the interface: total earthquakes, earthquake magnitude, total wells, and play name. You can filter the map results by highlighting data of interest within the timeline, magnitude and/or play widgets. For example you could isolate only those earthquakes that happened between 2009 and 2015 or only those earthquakes with a magnitude greater than 3.0 or combine those two selections to focus in on the recent seismicity of concern. I included the play name widget to demonstrate the ability to include and symbolize categorical data but also overlay drilling operations with known fault systems. The really cool thing about widgets in CartoBuilder is they are included with your final map so that others can explore the data you want to share. This is pretty similar to the shareable Dashboards in Tableau, but in my opinion easier to create and share. 

Another nice feature is the ability to reveal the underlying SQL and CartoCSS syntax in most of the panels - it doesn't appear possible yet to expose the CARTO.js used to create widgets. From a teaching perspective this is a great way for learners to explore and tinker with the code in a stable sandbox. But it is also useful for greater cartographic control over your map features. For example, since the oil and gas plays aren't an integral part of this story, I wanted them to have dashed lines so they were less obtrusive. To accomplish this I switched into the CartoCSS editor (lower left corner in the first image) and added a line-dash-array property to the layer symbology (second and third images). The only disadvantage is once you opt into the editor environment you need to stay there or reset all your changes (fourth image) - so you're no longer using the WYSIWYG interface. I don't think this is a real issue other than potentially slowing down new users unfamiliar with CartoCSS. But if you're a frequent user of halos or patterns, you'll quickly be switching to this Sublime-esque environment.

It's pretty easy to imagine the variety of inquiry-driven discussions we could jump start using the CartoBuilder TARDIS. Ultimately the bottleneck is finding relevant and sanitized geologic data. However it's worth the extra effort because in addition to exploring geology over space and time, geology students would also develop greater fluency with geospatial concepts and be required to navigate common syntax used to query and visualize spatial data. TheSQL statements used to filter or join data in the CARTO environment are transferable to desktop PostGIS environments. Most importantly for introductory courses, all of this is performed relatively easily compared to a full-blown GIS or by using Python or R. The Torque animation below (pretty similar to Dan's version) was created using the same earthquake data and CartoBuilder in under 5 minutes.

I'm developing another interface to help explore geologic relationships using the new analysis tools available in CartoBuilder. In my preliminary testing though, the underlying PostGIS engine supporting these cloud-based analyses is blazing fast compared to other... um, "competitors." I have shared my opinion about using PostGIS before to explore geologic relationships and I think the installation and database management required of a typical desktop setup may be a factor in seeing a slow adoption but CARTO offers painless access (both GUI and command line) to this powerful tool.

Header Image Source: http://j-vh.me/2ghv2Dr

An Atlas of Remote Tweets and The Popularity of Null Island

                    Source: http://j-vh.me/1HMmSGi

When I see the word 'atlas' it often conjures up memories of a tattered monstrosity my father saved from the local transfer station. Although its glossy dust cover had long been lost, the embossed majuscule on the front book board still held stubborn scales of faux-gold gilding. It covered the entire coffee table when opened and was a frequent co-pilot in helping make better sense of books like Treasure Island, Around the World in 80 Days and Tom Sawyer. That atlas represented - and in many ways, still does - an opportunity to explore unfamiliar territory, to refine spatial relationships and more often than not, provide an outlet for escape and imagination to run wild. 

This was no less true when I purchased Judith Schalansky's 'Atlas of Remote Islands' published in 2010. I enjoyed both the simplicity in her cartographic representation and the accompanying text for each island, which provided an historical and social context. The emphasis on remote islands ensured that readers would be lured to far away lands while learning about how those places are still related to modern geopolitical realities. Shalansky clearly recognizes her bias in selecting locations by asking "Whether an island such as Easter Island can be considered remote is simply a matter of perspective." 

---

“Whether an island such as Easter Island (pg.100) can be considered remote is simply a matter of perspective.”

One Thousand Remote Tweets 

I went back to Schalanksy's atlas after Eric Fisher at Mapbox released an interactive map depicting 6 billion tweets. Eric noticed that tweets were missing along the Prime Meridian, which resulted in anomalous banding. This made we wonder if the opposite were true - could we see individual tweets where we wouldn't expect them? So with the atlas as my guide I explored Fisher's map to see just how remote Shalanksy's selections were. 

First I created a table containing coordinates for all 50 islands - and for the geohipster crowd I added in Null Island, just because. I took the link to Eric's map and concatenated his link with each set of coordinates in CartoDB and then visited each island. My initial thought was to query the databases by a geographic bounding box using the q=&geocode=0,0,10km parameter but there isn't a publicly available database to query. Fortunately after I started exploring I realized there were so few I could literally just count them on-screen. I tallied the approximate number of tweets (approximate because yes I probably missed some) and added them to the table. The resulting CartoDB map depicts the location of Sholanksy's islands with an info window that reports out the total number of tweets, provides a link back to Fisher's map and a Wikipedia link for further exploration.

---

I tallied a total of 1,005 tweets from the 50 islands that Shalansky included in her atlas but only 16 of the islands reported any Twitter activity. The majority of the tweets were - not surprisingly - from Easter Island (49%) followed by Robinson Crusoe (15%) and Diego and Christmas Islands (5%). Although a third of the islands have internet access, after re-exploring these islands through the lens of Eric's map it is probably unlikely we will see that number dramatically increase. There are enough localities like Southern Thule or Rudolf & Franklin Islands that will remain offline until we start seeing Twitter-supported satellite phones (or maybe that is already a thing).

                     Source: http://j-vh.me/1xVAyPn

Null Island, Where Art Thou?

It was also interesting to see how many people had 'visited' Null Island. While it is likely impossible to decipher whether this was intentional geolocation anarchy, geohipsters at play or geocoders gone wrong, given that Null Island doesn't exist it tallied an impressive 10% of the total Twitter activity. And this is where the realities of the traditional atlas intersect the realities of spatial data collection. 

I wouldn't have found Null Island in my childhood atlas (or using Google Earth today) but it IS a spatial entity - albeit two simple coordinates that could have just as easily been named 'Origin Island.' So while the framing of Fisher's map as the most detailed ever resulted in the folks at Floating Sheep to state it 'rubs us the wrong way' it does provide some interesting details about the intersection between access to and adoption of the Twitter platform and how that data is aggregated and interpreted by various geolocation services (something they also elaborate on).

Null Island is a spatial reality that became a visual reality - and geohipster totem - after accepting that the multiple reports of a position of [0,0] represented 'something' (although initially just a geogoder failsafe).  Visualizations of global flight patterns or wind dynamics also represent these spatial realities we can't 'visit' or see without the visualization process. And without Eric's visualization - most detailed or not - it wouldn't be as easy to decipher something like the distribution of the most remote tweets. And though the medium of the atlas by which we explore the world may be changing, Shalanky's caution about what we consider remote is still warranted.

Although Shalanky's subtitle is "Fifty Islands I have Never Set Foot On And Never Will" I feel fortunate to have visited at least one of them - Deception Island. I was a participant on the Geological Society of America's 125th Anniversary field trip to explore the geology and ecology of South Georgia Island, South Scotia Arc and Antarctic Pennisula through Cheesmans' Ecology Safaris. For a more in-depth geologic exploration of the island you should read Anne Jefferson's and Chris Rowan's recent blog entry on Highly Allochthonous. However I can safely add Null Island to the list of islands that I will never set foot on...

Van Hoesen and Edwards on Deception Island

Photo credit: Heather McFarland