�ֲ���ý

In research published today, recent PhD graduate Asia Kaiser details how synthetic control methods estimated significant declines in bee observations when traditional analyses didn’t

Since it launched in 2008 as a UC Berkeley student’s master's project, the platform has been a source of both fascination and frustration for researchers.��

The hundreds of millions of observations about the natural world logged by both professional and citizen scientists around the globe are a treasure trove of information about biodiversity. But is that data usable in research? The prevailing sentiment has veered toward doubt, skepticism or an outright “no.”

“I think the feeling has been, ‘Oh, because this data is just being collected opportunistically by nature enthusiasts and not in a standardized, rigorous way, it can’t be used in scientific research,’” says Asia Kaiser, who earlier this month earned her PhD in the �ֲ���ý Department of Ecology and Evolutionary Biology. “If you haven’t planned out data collection in advance, a lot of researchers hesitate to use it.”

Recent PhD graduate Asia Kaiser studied how synthetic control methods estimated significant declines in bee observations when traditional analyses didn’t.

There had to be a way, Kaiser thought, to tap into the vast cache of information logged into iNaturalist without sacrificing scientific rigor, especially data collected in urban environments. The answer, it turned out, lay in economics.

In , Kaiser and co-authors Julian Resasco and Laura Dee, both associate professors of ecology and evolutionary biology, detail how combining iNaturalist records with synthetic control methods, originally used in economics, estimated a significant decline in bee observations in Philadelphia during the two years following Hurricane Ida in 2021, while conventional ecological analyses didn’t detect the decline.

“Basically, the inspiration for this project was thinking about causal inference in ecology,” Kaiser explains. “When we have observational data, can we actually use that to ask questions about drivers of biodiversity?”

‘You can’t just go into people’s backyards’

These questions dovetailed neatly with Kaiser’s research focus, which is bees—specifically, how human land use affects different insect groups and, consequently, the ecosystem services they provide in coupled human-natural systems. Among her research aims is understanding biodiversity in urban environments, improving the resilience of urban agroecosystems, increasing equitable access to fresh produce and promoting environmental justice in cities.��

However, monitoring biodiversity and evaluating drivers of change in urban environments is confounded by several issues: “Cities are mosaics of land-use types, including parks, private properties, buildings, roads and industrial zones,” Kaiser writes in the paper. “As a result, sampling efforts can be complicated by permission and safety issues, and leaving unattended sampling equipment in the field brings a higher risk of theft, tampering and vandalism in cities.

“Given these challenges, measuring biodiversity in cities requires different tools and data streams than those used in natural ecosystems. Participatory science data is a promising solution for monitoring biodiversity in cities; cities are the land use type with some of the highest upload volumes of data to participatory science platforms, largely because upload frequency is strongly influenced by population density.”

Despite the abundance of participatory science data in platforms like iNaturalist, researchers have hesitated to draw from it, relying instead on randomized, controlled and replicable experiments to identify and estimate causal relationships. That kind of science, Kaiser says, becomes more difficult in urban environments due to sampling challenges and historical legacies that shape different neighborhoods, among other reasons.

“If you’re studying a natural area, you could get a permit and go sample all over, but you can’t do that in a city,” Kaiser says. “Even if you get a permit, you can’t just go into people’s backyards.”

The idea of how to bridge the gap between the abundance of iNaturalist data logged in urban areas and the rigor expected in scientific research came to Kaiser when she was assigned to watch a lecture given by a Nobel laureate in economics. The lecture topic was synthetic control methods, which originated in economics as a way to create a nonexistent control group that allows for comparisons between real-world groups before and after an event or intervention.

One of the most famous uses of synthetic control methods in economics was in estimating the impact of Germany’s reunification after the fall of the Berlin Wall on the gross domestic product (GDP) of western Germany. Economists created a “synthetic” Germany from economic data to study GDP with and without reunification.

Though synthetic control methods hadn’t been widely used in ecology research, “I thought it could be adopted with iNaturalist data,” Kaiser explains. She was further interested in studying the effects of Hurricane Ida on her home city of Philadelphia, which included significant flooding.��

“Even though it didn’t have a huge impact on people per se, the effects of the hurricane were really dramatic. Looking at the water levels, the stream gauges had their highest values ever in the 100 years that they’ve been measuring. My feeling was that would have a pretty big impact on bees, because if you look at bee biodiversity, bees are pretty sensitive to precipitation and water. The ones that nest in the ground are really affected by huge flooding events.”

Declines following a hurricane

To apply synthetic control methods to ecological research, Kaiser and her colleagues drew data from the , which collects research-grade iNaturalist data—that which includes, among other points, latitude and longitude, collection date and time and correct identification—as a proxy for bee abundance in Philadelphia.

They analyzed for bee population declines and, in addition to synthetic control methods, also performed the more traditional methods of interrupted time series regression, before-after control impact regression and before-after regression.

Kaiser and her colleagues found that synthetic control estimated a 15.5%—20.9% decline in bee observations in the two years following Hurricane Ida. In contrast, the three more common ecological analyses didn’t detect this decline.��

“That was an amazing moment, seeing this decline in the data and better understanding how iNaturalist data may be able to help us look at the impact of unusual climate events—things that are happening more and more these days, like huge fires, huge floods, abnormally warm winters,” Kaiser says. “Unless you were already collecting data in a region before, you can’t really see the impact before the event, but synthetic control methods might be able to help us in those situations.”

Kaiser adds that this method also might be useful for looking at the effect of policy interventions. For example, the city of �ֲ���ý is establishing pollinator corridors, and Kaiser sees potential in using this method to draw from iNaturalist data in studying the outcomes of these corridors.

Scientists who reviewed the paper expressed excitement and skepticism about using synthetic control methods in ecological research, Kaiser says: “They asked questions about whether or not the decline I’m seeing is a true thing that’s happening or an artifact of the way data has been collected. iNaturalist is very sensitive to observers—wealthy neighborhoods have higher uploads, areas around research universities have higher uploads—but this statistical method can help control for those things.”��

Thanks to the professional and citizen scientists gathering data and sharing it on iNaturalist, Kaiser says she sees potential to apply synthetic control methods to a range of ecological research. For example, “using the bee biodiversity that’s collected on iNaturalist, does that correlate with how well flowers are being pollinated? I think that’s something we’ll be able to study.”

Did you enjoy this article?����Passionate about ecology and evolutionary biology?��Show your support.

��