Public parks serve an important societal function as recreational spaces for diverse communities of people, with well documented physical and mental health benefits. As such, parks may be crucial for how people have handled effects of the COVID-19 pandemic, particularly the increasingly limited recreational opportunities, widespread financial uncertainty, and consequent heightened anxiety. Despite the documented benefits of parks, however, many states have instituted park shutdown orders due to fears that public parks could facilitate SARS-CoV-2 transmission. Here we use geotagged social media data from state, county, and local parks throughout New Jersey to examine whether park visitation increased when the COVID-19 pandemic began and whether park shutdown orders were effective at deterring park usage. We compare park usage during four discrete stages of spring 2020: (1) before the pandemic began, (2) during the beginning of the pandemic, (3) during the New Jersey governor’s state-wide park shutdown order, and (4) following the lifting of the shutdown. We find that park visitation increased by 63.4% with the onset of the pandemic. The subsequent park shutdown order caused visitation in closed parks to decline by 76.1% while parks that remained open continued to experience elevated visitation levels. Visitation then returned to elevated pre-shutdown levels when closed parks were allowed to reopen. Altogether, our results indicate that parks continue to provide crucial services to society, particularly in stressful times when opportunities for recreation are limited. Furthermore, our results suggest that policies targeting human behavior can be effective and are largely reversible. As such, we should continue to invest in public parks and to explore the role of parks in managing public health and psychological well-being.

We assessed park visitation at our selection of candidate parks using publicly available, geotagged social media data from Instagram. To gather data on park usage, we wrote a Python script to collect data from publicly accessible Instagram photos. We first matched candidate parks to their corresponding Instagram location tags, of which there were often several for a park, and visually verified that all location tags corresponded to the correct parks using Google maps. After we had created this list of accurate location tags for the parks that we identified in the NJ OSPRI dataset, we ran our Python script to collect the metadata of all photos associated with each tag; we collected the user handle, date, location tag, photo caption, and hashtags for each photo.