* Any views expressed in this article are those of the author and not of Thomson Reuters Foundation.
A new study indicates that protected areas have saved wildlife from decline in many tropical forests. To get their findings, researchers sorted through 2.5 million camera-trap photos taken in 15 protected areas in rainforests around the world. How did they pull off such a feat? Jorge Ahumada, executive director of the Tropical Ecology Assessment and Monitoring (TEAM) Network,which produced the study, explains.
Question: What exactly can camera-trap photos tell you about how tropical species are faring?
Answer: We start with the raw data: a time-stamped image of an animal that has been identified by a scientist. We have around 500,000 raw images a year coming into our database from TEAM camera traps.
There can be a lot of problems with camera-trap data if you just take it at face value. You can’t just count the number of pictures of a particular species and use that to accurately estimate the population of that species in the forest. Because sometimes an animal will walk in front of the camera 100 times, and then you will have 100 pictures of that one animal. The other problem is when you don’t see something in a camera trap, that doesn’t mean that the species isn’t there; it could mean that it just didn’t walk in front of the camera trap. This is a very important distinction that few people make.
So the first thing we do with our raw data is we simplify it. We divide each site’s data in a given year into 15 time periods (roughly four to six days for each period), and then using some algorithms, we check if the species was there during each of these time periods. If it was there — whether only once or 100 times — we write “1.” If it wasn’t, “0.” We create what’s called a binary matrix for each species at each camera-trap site, each year.
From these matrices we calculate what we call occupancies for each species — the proportion of camera-trap points where a species was seen in a given year. Within these calculations, we correct for false negatives, or the probability that the species was not seen given that it is really there.
We then use the occupancies to produce something called the Wildlife Picture Index (WPI). This is an indicator that takes all the occupancies of all the species at a given site, or continent — basically any grouping you want — and averages them out. Then it shows you the average of all these occupancies relative to the first year they were measured. So the first year is always going to be 1. A WPI that goes below 1 is a decrease; a WPI above 1 is an increase.
Q: How can you process all these data so quickly?
A: The WPI is powered by something called the WPI Analytics System, a system we built with HP which allows us to analyze the TEAM data much faster than was possible even a few years ago. For anyone who’s interested in seeing what we’ve learned so far, we have a website that shows the results of the WPI. Behind the scenes, we have HP servers rapidly cranking the numbers. The system can recalculate the whole occupancy of all the species in the whole network overnight. And once you have that, you can recalculate the WPI based on the new information scientists have added.
Our methods are public, but given that they require complicated simulations and statistical modelling, they’re not usually something scientists can just do quickly. With the WPI Analytics System, we’re taking all that away and putting it in a box so that people just enter the data, and it calculates the results for them.
Q: Who is using the WPI, and how is it helping them?
The data we’ve collected has been an especially useful tool for park managers. If a park manager looks at the WPI from his site and detects a problem like species decline, he can then quickly alert someone higher in the parks department. For example, in Uganda’s Bwindi Impenetrable National Park, TEAM site manager Badru Mugerwa analyzed the camera-trap data for the African golden cat using the WPI system and noticed the cats were being seen less frequently in places with heavy tourist traffic. He then went to Ugandan wildlife authorities and said “Hey, we need to change the routes of the tourists if we want to preserve the cat.”
Now that this data analysis is available so quickly, researchers like Badru don’t have to wait until someone publishes a paper 10 years later before they can confidently report scientific trends to the people who need to know about them. Instead, they upload the data, and by the next day it’s already analyzed for them.
It also helps that the WPI summarizes the biodiversity information in a way that is understandable to policymakers. This way, you don’t have to explain the scientific details to a government official. You can just tell them, “If it’s below 1, you have a problem. If it’s above 1, you don’t.”
Q: What are the next steps for this type of work?
A: World domination! We created the system, we’ve demonstrated that it works, and now we’re trying to amplify it.
In November we launched a new website called Wildlife Insights that is essentially like the WPI that we had, but on steroids. Although right now TEAM camera-trap data is the main source of the WPI, we are starting to include data from other groups that has been collected in a similar way to our own. When others submit data, we determine whether their data-collection practices meet our standards. Consistent data-collection methods help ensure the accuracy of our results.
The idea is that more and more people and organizations will eventually contribute to this global WPI. We’re also working with national governments like Brazil and China, convincing them to set up monitoring systems like TEAM’s to monitor their countries’ protected areas and then use the data to inform decisions about how best to management the species and forest. We are setting up pilot systems in both countries and we eventually hope these governments adopt them as their official monitoring system for wildlife.
This article previously appeared on the blog of Conservation International.