Student Research Reports

View Research Report

Although the accuracy of global land cover data products has greatly increased in recent years due to technological advancements, the visual understanding of Earth’s surface that scientists obtain from existing land cover maps often diverges from ground observations obtained during field investigations. Today, many high resolution land cover maps such as the European Space Agency’s (ESA) 2020 WorldCover (WC) map stand at an accuracy below 75%. Faced with these accuracy limitations, scientists have turned to in-situ citizen science observations such as those from GLOBE Observer to supplement existing land cover data and to increase its accuracy. Our research focused on increasing the impact of citizen science by identifying the key environmental and geographical factors associated with discrepancies between an official land cover map, WorldCover, and citizen scientist land cover classification of satellite imagery through Collect Earth Online (CEO). According to our statistical analysis conducted using computer-generated confusion matrices, the agreement between citizen scientist and WorldCover land cover classifications was highest in areas with mostly homogenous land cover. There is a relatively strong negative association between land cover diversity and classification agreement. Additionally, we observed that classification agreement is negatively correlated with the highest amount of shrubland classified between the citizen scientists and the WC map. This is because shrubland is often confused for other types of vegetation and vice versa. Using the numerous associations we found, we were able to identify the types of areas in which citizen science observations will be most useful in providing new insights into land cover. This information can potentially help provide a more effective and streamlined method for scientists to document and collect impactful crowd-sourced data. By helping to improve global land cover maps through citizen science, our research may assist professionals in diverse fields fight some of the world’s most pressing issues, including those involving natural resource management and mosquito source reduction. Key words: land cover, data accuracy, citizen science, remote sensing