Covid-related surge in global wild bird feeding: Implications for biodiversity and human-nature interaction

b. Bird species richness We obtained nation-level bird species data from BirdLife International to measure local bird species richness in each country and territory. As noted earlier, this is a proxy measure for the biodiversity of an ecosystem, and it is distinct from species diversity, which is often measured using the Shannon-Wiener Index [40]. BirdLife International [41] provides spatial maps that present the distribution of bird species across the world. We overlaid national boundaries with bird species polygons to calculate the number of bird species in each country and territory. These data categorized bird species into four groupings based on seasonality: Resident, Breed- ing Season, Non-breeding Season, and Passage (See Table A1 in S1 File and Fig 2). c. Lockdown timing To examine how lockdowns affected people’s engagement with birds, we use data from the University of Oxford on the timing of lockdown initiation by country and territory [42]. More than 90 countries and territories implemented shelter-in-place policies at the beginning of the Covid-19 pandemic [43]. d. Internet access Internet accessibility can be correlated with a country’s Google search intensity as the search intensity can be higher in countries with a higher percentage of populations using the internet. We obtained internet access data on the percentage of populations using the internet in each country in 2019 from the World Bank. Internet access is utilized as control while exploring the relationship between the presence or absence of search interest in bird feeding and bird species richness.

III. Methods To assess how Covid-19 lockdowns affected bird feeding interest, we compared Google Trend search intensity before and after the lockdown for countries with sufficient search interest using the panel search intensity data from Google Trends. Our empirical approach was depen- dent upon the fact that lockdown timing varied across nations, as shown in Fig 2 Panel A. To estimate how interest in bird feeding changed after Covid-19 lockdowns, we used an event study empirical framework common in economics. This framework estimated how bird feeding search intensity differs across time, relative to the week before a country starts its Covid-19 lockdown Yct ¼ a þ Xk¼ 2 k¼ 52 skZk ct þ XK k¼0 tkZk ct þ Gct þ �ct ð1Þ The expression is indexed at the country c and week t level. Zk ct are a set of variables that indicate the number of k weeks relative to the week a country c’s first lockdown (k = 0 is the week that the first lockdown was initiated). The σk and τk measure changes in search intensity in each week before and after the first lockdown in a county, respectively. For each week, we plotted the coefficients σ and τ for each week on a graph to indicate the average change in bird feeding before and after the week before the first lockdown, after controlling for the average bird feeding response of that week and for that country, Γct, and the average overall, α. The error term, �ct, is unlikely to be independent across weeks for a country and instead will be cor- related. For this reason, we clustered the standard errors for each country. This specification

allowed us to test whether there were trends in bird feeding that predated lockdown by assess- ing the vector of coefficients σk. This also allowed us to determine how the response evolved after lockdown by assessing the vector of coefficients τk. PLOS ONE Covid-related surge in global wild bird feeding PLOS ONE | https://doi.org/10.1371/journal.pone.0287116 August 2, 2023 4 / 15