Building The Data Dashboard – From Inception to Release

The following piece was written by OAS Communications Coordinator Ryan Bower for the Snapshot Wisconsin newsletter. To subscribe to the newsletter, visit this link.

This edition of the Snapshot newsletter is focused on a new Snapshot Wisconsin product, the Data Dashboard. The Data Dashboard is an interactive tool that let’s the public explore the data that Snapshot has collected in a new and exciting way. This dashboard marks an important step towards Snapshot’s goal of making its data more accessible to the volunteers who helped collect it, as well as the public more broadly.

You can learn more about the development of the dashboard, as well as what features are currently available on it, in this edition of the Snapshot newsletter, or you can check out the dashboard for yourself at https://widnr-snapshotwisconsin.shinyapps.io/DataDashboard/.

The idea of the Data Dashboard originated about two years ago, stemming from a desire to bring what Snapshot has learned full circle back to the volunteers and public. However, no one knew at the time what would eventually manifest from that desire. No one knew that the Data Dashboard was on the horizon.

The core purpose of Snapshot Wisconsin is to provide data for wildlife decision support and bringing citizens into that process. Citizen science projects like Snapshot Wisconsin rely upon the public’s aid to accomplish tasks that would be unfeasible or impossible otherwise, and making sure volunteers feel fulfilled and satisfied with their time investment is important to the Snapshot team.

As part of her dissertation research, Christine Anhalt-Depies, the project coordinator of Snapshot Wisconsin, investigated why volunteers joined the program, hoping to learn their motivations and reasons for staying. Her surveys taught the team much about what they needed to focus on to keep volunteers feeling fulfilled.

Two of the survey’s findings stood out to the team. First, many volunteers cared deeply about contributing to wildlife monitoring and wanted to see what wildlife was in their area and the state. Volunteers also wanted to see the fruit of their work, how the project was contributing to wildlife management decisions across Wisconsin. The Snapshot team knew from these findings that they needed something that would close the loop, bringing the data full circle.

To help close the loop, the Snapshot team decided to focus on making Snapshot findings more available and accessible to the public. They needed a tool that everyone could use and learn from. With that daunting task in front of them, the team started by investigating what kinds of platforms were even available for this type of visualization tool.

The Snapshot team explored many options but ultimately settled on R Shiny, an extension of the statistical software R, and soon a prototype was being built. The initial prototype of the dashboard started as a map of Wisconsin, but the team wanted to show more. The team needed a specialist with a vision of what the tool could become.

A Fresh Face Joins The Team

In November of this last year (2019), a new member joined the Snapshot Wisconsin team, and he would play a leading role in developing the Data Dashboard.

Ryan Bemowski, Data Scientist and Engineer at the Wisconsin DNR and developer of the Data Dashboard, recalled what it was like joining the project at this time. “There were many discussions still going on about what the tool was going to look like. The team was trying to figure out how they could best display the data in a way that was meaningful to volunteers and the public,” said Bemowski. “My primary task was to get the visualizations made by deciding what data to display and come up with ideas for the visualizations.”

Bemowski spent the next eight months reimagining the look and feel of the Data Dashboard. Bemowski kept one central piece of the prototype, the map of Wisconsin, but simplified it to show the percentage of cameras in each county or management unit that have detected the selected species at least once.

During these months, he also dived into what other data would be shown on the dashboard. Snapshot Wisconsin cameras have taken over 47 million photos to date, but many of these photos don’t contain animals. The majority of the photos that do contain animals will be classified correctly. However, the team knows that not all the photo classifications are accurate, complicating their use. Some species are particularly difficult to classify, often being mistaken for similar species, such as wolves (below left) and coyotes (below right).

Bemowski and the team wanted to show as much data as possible. However, they didn’t want to add data to the dashboard that was inaccurate, so they needed a way to determine how accurate the community was at classifying each species.

Bemowski explained the team’s approach to solving the inaccuracy issue. “We ran a round of accuracy analyses to see how accurate [classifications] are for each species, but we ended up only able to do the analysis for certain subsets of species,” said Bemowski. An accuracy analysis is a test to determine how many of the photos of a species are correctly classified using a sample of photos. The team’s accuracy analysis yielded an accuracy rate for most species. However, it didn’t work for certain species. “When you run an accuracy analysis, you must have a big enough sample for each species. In our case, some species like beaver just didn’t have enough photos to be analyzed properly.”

Additionally, a few species were analyzed but still excluded because of their low accuracy rate. Volunteers are overall very good at identifying most Wisconsin species, but some species are difficult and had a low accuracy. Bemowski mentioned that the team will continue to run more accuracy analyses and work towards getting all species on the dashboard eventually.

The Soft Launch, Feedback And Improvements

After much work, Bemowski and the team were ready mid-July to test out their tool with a soft launch to a handful of volunteers. They chose the 1,700 volunteers who host Snapshot trail cameras and asked for feedback on the dashboard. Around 90 responses came back, and three main themes emerged that the team began to think about.

First, the volunteers wanted clearer explanations about why species were included or excluded. Second, there was a desire that certain species be added to the dashboard, namely wolf and elk. Lastly, volunteers wanted to see data at a more granular scale, such as being able to see the data for specific cameras instead of countywide.

After reading the survey responses, Bemowski and the team set about incorporating this feedback and getting the dashboard ready for the full launch to the public. The first theme was easy to address. Bemowski added a clickable pop-up (below the species list) to the dashboard that shows the accuracy rate for each species. Simultaneously, Bemowski and the team were able to finish the accuracy analyses of three more species and added them to the dashboard: Snowshoe hair, pheasant and fisher.

The second theme that emerged was a desire to see certain, high-interest species on the dashboard. Bemowski explained, “The reason we excluded some species [for the soft launch] is because, below a certain accuracy, we know some of the data is incorrect. If a species has a 50% accuracy rate, for example, we know that 50% of the data being shown is correct and 50% is incorrect.” The data for species with an accuracy rate below 95% is still stored and used in other ways, but the team wanted all the data on the dashboard to meet a minimum quality standard.

The team thought of a clever way to balance the public’s desire to see high-interest species like wolf and elk with their data accuracy standards. The photos of species with low accuracy and species of special interest go through an extra round of classification by species experts to confirm which photos are correct, so the accuracy of the expert-classified subset of photos is 100% and meets the data quality standards. The Snapshot team collectively decided that the best way to include the data for elk and wolves was to only use the expert-confirmed photos on the dashboard. Bemowski said, “We want to show correct data, as accurate as possible, without holding anything back.” This decision was the best compromise the team came up with.

The last theme focuses on granularity, or the scale to which you can “zoom in.” Some volunteers wanted to see beyond the county scale and look at data from individual cameras. Bemowski was conflicted about the topic of granularity though. “We don’t want to give the exact latitude and longitude of the cameras because of privacy concerns, but we still wanted to give as much granularity as possible.” Showing data at the county level is as granular (or “zoomed in”) as Snapshot can currently go, but the team hopes to eventually have individual dashboards for each camera’s host so that they can see the data for their cameras. However, that is still in the planning stage.

What’s Next For The Dashboard?

Now that the dashboard is released, Bemowski reflected upon what the dashboard means to the project. “I’ve realized recently how big of a step this dashboard is for Snapshot Wisconsin: To go from collecting millions of photos to showing people what the data is saying and really analyzing that in depth. That’s a big step. Looking back, the intention has always been to transform Snapshot data into actionable outcomes,” but the Data Dashboard marks a major step towards Snapshot sharing its findings with the public.

The Data Dashboard’s current version is focused on making Snapshot’s findings available to the public, but more updates and additions to the dashboard are already on the planning board. “The Data Dashboard is an evolving product,” said Bemowski. “We are going to go through many iterations of it to improve the quality and abundance of data and species available. The next step is to build upon the dashboard so it can be more meaningful for decision makers, such as the county deer adivory councils (CDACs) and other wildlife management organizations around Wisconsin.”

Bemowski offered some advice for anyone going outside to interact with nature. “Improve your chances of seeing your favorite species by using the Dashboard to observe wildlife during their times of peak activity or use the dashboard to see if your favorite species have been detected in your county. Even if you don’t see your favorite species, you may be surprised at other wildlife you do see.”

Bemowski and the Snapshot team are momentarily celebrating reaching this important milestone before they continue to improve the dashboard. Bemowski said, “I wish I had something like this as a kid. I grew up in Wisconsin. I grew up a hunter and fisher; someone who enjoys the outdoors. I always heard that there isn’t enough information about wildlife, that people just don’t know what is there, and that all the ‘guesses’ from wildlife organizations [about what is around] aren’t true.”

“But the Data Dashboard is as close as you can get to really knowing what is out there,” continued Bemowski. “Drawing from a growing collection of over 2.3 million animal sightings from over 2,200 cameras locations across Wisconsin, the dashboard is a really awesome way to discover and experience these animal sightings.”

If you want to explore the Data Dashboard yourself, you can access it at https://widnr-snapshotwisconsin.shinyapps.io/DataDashboard/.