In 2013, The Great Elephant Census embarked on a journey to understand the alarming decline of elephant populations across Africa. National parks and wildlife staff in 18 countries with support from seven NGOs coordinated by Vulcan to carry out the most extensive pan continental elephant survey since the 1970’s.

The efforts resulted in an unprecedented data collection. From the flight logs from the survey to the geolocation of elephants and gorgeous footage of African wildlife, the Great Elephant Census was able to collect the most comprehensive elephant conservation database in history.

The footage recorded from the aerial survey was beautiful, and the message conveyed seemed simple enough yet touching. So when Vulcan contacted the Office for Creative Research, I could not help but wonder: why would one need data visualization when the story to tell is so beautiful and relatable?

As it turns out, elephant conservation is a very complex issue. The richness of the collected data quickly revealed that we would not be able to condense this complex system of causalities and correlations into an over-simplified narrative. For this reason we set out very early in the process to go beyond storytelling, and create a platform that guides our audience in learning about elephant conservation through exploration and interaction.

Elephant range in Africa — African Elephant Specialist Group, IUCN

As Ted Schmidt — senior program manager-conservation at Vulcan — put it, OCR was commissioned to create what would become the "main non-scientific artefact of the Great Elephant Census." Together with Vulcan, we shaped this collaboration around the 3 following goals:

- Create a tool for policy makers (e.g CITIES representatives) to make informed decisions based on state-of-the-art elephant data.

- Engage the general public with the story of the Great Elephant Census, by showing the incredible human effort behind the status report, and the complexity of the science involved.

- Provide audiences with a direct access to the data through an API explorer, allowing journalists, teachers, students, artists and developers to take the data beyond our initial vision.

Our collaboration with Vulcan revolved around research and exploration, which of course led us to learn about elephant conservation ourselves. The framework of our analysis was defined around multiple avenues, from the complexity of the conservation ecosystem itself, to the overall situation of elephant populations across the African continent, to an initial look at the data at hand.

An early look at elephant populations gain and loss across Africa

As for each of the OCR's projects, we excitedly took a stab at the data available to us. We went through a number of research papers and status reports describing the decline (and in rare occurences gain) of elephant population across Africa. This gave us a broad understanding of the conservation effort led in the 18 countries involved.

Not forgetting that the surveying effort was a story on its own, we also looked at the process of counting the elephants. The wealth of data included in the Great Elephant Database gave us access to thousands of hours of flights logs, raw sigtings and meta information about each expedition.

3 early explorations of the database, looking at survey designs and flight logs

This initial look at the data collected by the Great Elephant Census enabled us to plan for creating the right backend tools for exploiting that information, and creating technically viable concepts.

An incomplete map of the international conservation eco-system

In parallel, our research led us to analyze the landscape of the elephant conservation ecosystem itself. Understanding the numerous actors involved would help us tailoring our message for each user group, and provide them with the appropriate tools for learning, exploring, and taking action for the elephants.

A studio charette was organized to decide on the broad orientations of the project

Our team concluded this initial phase by synthesizing the extensive research gathered. At that point, our production took the form of a framework for approaching multiple target audiences (from conservation experts to the general public and policty-makers), and a typology of concepts that could resonate with each of them.

Some of the concepts selected to become interactive prototypes

We drafted a number of storyboards defining a dozen of creative concepts, from immersive game experiences to informational dashboards and social media attractors. Each of them would later become its own prototype, in order to narrow down our options and validate our intuitions.

In conjunction to the research phase going on in New York, I was lucky enough to be sent for a field research trip in Tanzania, South Africa, and Bostswana. It was an opportunity to meet a number of researchers and conservation specialists, whose help and guidance had a critical impact on our work.

More importantly, it was a chance for me to discover the reality of elephant life on the field. If visualization sometimes fail at capturing the weight of data, this experience in Africa was a mandate for me to always remember what was at stake: the life of a keystone species in the African landscape.

The research phase of the project provided us with enough insights to start developing concepts out and tools for data exploration. This was the object of the next step: the prototyping phase. We set out to create as many quick prototypes as possible to either validate or reject the concepts drafted during the initial phase of the project.

This phase was brilliantly led by Jane Friedhoff and Sarah Groff Hennigh-Palermo. You guys rule.

The pace of elephant counts at a regional level — Jane Friedhoff

We quickly created a number of map-based visualizations to show the process of elephant counting itself. These maps showed effectively how the distribution of elephant populations varied across the continent. It also told the story of data collection protocols and their shortcomings.

A timeline exploring the impact of socio-economic factors on elephant conservation — Jane Friedhoff

Other prototypes also attempted to analyze the data from a time-based approach. In the graphics above (concept by Jane Friedhoff), the baseline shift over time allows to compare different variables on a similar scales. In this way, we can (for instance) explore how the evolution of GDP in specific countries may or may not have an influence on elephant loss or habitat loss. This prototype made the final cut.

A historical comparison of elephant population estimates by country — Jane Friedhoff

We also looked at a number of interactive mechanisms for people to compare elephant conservation efforts from one country to the other, as it appeared that situations could vary tremendously between neighboring nations. This proved to be challenging, considering the inconsistent nature of data collection over time.

Early prototypes are turned into interactive pieces and placed in a broader user flow

From a month of intense data exploration remained a handful of prototypes that all seemed viable from conceptual, scientific, and technical perspectives. OCR team worked together at creating a platform that would stitch these interactive pieces together in a coherent web experience.

The main challenge I faced when designing the interface for the African Elephant Atlas, was to create visual language that would tie all the moving parts of the experience in a coherent fashion. From immersive map visualizations to austere informational dashboards, the entire experience needed to be appealing, entertaining, and engaging.

Elephant Atlas moodboard

Typography and color guidelines

As opposed to past project from the OCR, such as Into the Okavango, the Elephant Atlas is a story for people to experience, instead of a tool to explore. This is why we went away from the neutral color schemes found in most map visualizations, and instead chose a colorful palette that would contrast nicely with the visually rich African landscapes depicted through images and maps.

For similar reasons, we narrowed down the number of interactions available during the experience. Focused on scrolling and basic interactions, the Elephant Atlas finds inspiration in some of the long reads published by the New York Times and other media outlets. Although its narrative is linear, it always gives access to side tracks, to learn more and dive deeper.

The Census page teaches how elephant counting was done by the GEC

The Context page works as a hypothesis generator and helps grasping the complexity of elephant conservation

The Countries page compares situations across Africa; some countries like Tanzania have lost 70% or their elephant population in a decade

To this day, mobile design is still a strange beast for data visualization. Rich media doesn't always play well on mobile devices, either because of limited performances or complex, counter-intuitive touch interactions. We however designed the Elephant Atlas with those constraints in mind, and I can proudly say we made some major breakthrough in that regard.

The responsive layout I designed allowed to display all core pieces of information with equal accuracy and interactivity on all platforms, with minimal development overhead. Tablet devices in particular ended up being the most pleasurable to interact with, while the OCR> historically had trouble designing beyond desktop experiences.

The layout of the interface was thoroughly reshaped to fit a mobile experience

While we could go further in implementing custom user flows for each platform, this project shows some promising leads on how rich data interactions can be translated to mobile format.

Access and release of complete datasets is always challenging, but this project took these complications to a level we had not experienced before. Vulcan had to make legal agreements with numerous actors in order to collect an release elephant data, from governments to conservation agencies. Each of these entities had specific requirements, as their data disclosure policies ranged from full access to full restriction. As we were building a platform for creating public access to the information, we had to navigate this quasi-hostile legal landscape.

Red countries indicate nations for which no geolocated elephant counts could be used.

The reasons behind conservation data restrictions can vary. Besides evil excuses made up by corrupt governments, more legitimate concerns often include animal safety. Indeed, disclosing elephant herds geolocation could be seen as giving away sensitive information to poachers. Specialists involved with the project however stated this concern was addressed by the obsolescence of the data. We also went at length to minimize any risk.

On every map of the Elephant Atlas, elephant geolocation is sampled and randomized to prevent criminal activity. For the most sensitive countries, the regional elephant counts were removed altogether.

How to show the decline of elephant populations in a relatable way? As a medium, data visualization shines at making sense of complex systems and revealing patterns. While it can be impactful, it sometimes feels abstract and dehumanized. It was very clear with this project that any visual treatment we were giving to the data should not obfuscate the lives behind the numbers.

Where most visualization projects would make use of standard density heatmaps, we decided instead to trade a bit of scientific accuracy in order to gain emotional engagement: for each elephant we would light one particle. It would not only show density changes over time: it would show where elephants live and die in a very visceral way.

Comparison of 2 design approaches: hexbins vs scatter plot

Of course, displaying 400,000 dynamic particles on an interative map proved to be a technical challenge. To make this amount of data workable on both client and server side, I pre-rendered data sprite sheets at various resolutions for each region involved in the census. Using elephant density data as well as some Perlin noise to simulate natural distribution, I was able to generate realistic particle systems.

Region geometries are initially captured in vector format, then laid out in a sprite sheet, and finally populated as a data store

These sprite sheets were generated for each year of data records, and then tied together. Besides creating pretty GIFs, they were used to generate interactive time series that could be navigated by scrolling through the interface.

Exploring the platform reveals some dramatic stories (e.g Tanzania lost 80% of it's elephant population over the past decade) and some more optimistic ones. Our approach hopefully will tell these with honesty and humanity.

30 years of elephant population estimates in Zimbabwe

The GIF sprite sheet is broken down per year and animated via scroll interaction in the final interface