Building up local knowledge on restoration: lessons learnt from organizing a set of crowdsourcing campaigns

Restoration of degraded land is an important national goal to achieve Indonesia’s environmental targets. To map both land cover and land degradation, Indonesia needs timely, high quality data and the necessary tools. We have addressed this issue by running a sequence of crowdsourcing campaigns. Our aim is not only to collect the data but to also potentially present a way for citizens to contribute to larger environmental policies and strategies. 

Focusing on land cover identification and tree cover change, we planned and ran a set of  pilot crowdsourcing campaigns in two provinces in Indonesia. We analysed the data from these pilot campaigns, and then used the insights obtained in the subsequent crowdsourcing campaign on land cover identification, upscaled to national level, which is currently ongoing. The campaigns were run using a mobile application developed as part of the RESTORE+ project. Through this application, we presented volunteers with simple microtasks by showing them satellite images and asking a simple yes/no question as to whether the image shows a particular land cover class. The application implemented a scoring system, which additionally performs a quality control of the data contributed by the crowd, and users competed with each other to classify the satellite images displayed by the application. 692 volunteers have actively engaged in the pilot crowdsourcing campaigns and have contributed more than 2.5 million satellite image interpretations.  

Based on the insights from the pilot campaigns, as well as an expert consultation session in Indonesia, the crowdsourcing application was modified to ensure, first, a uniform number of interpretations across the images, and secondly, higher quality data by allowing users to focus on geographical areas familiar to them, as well as to see the larger area surrounding the target sample.   

We analyzed the data collected and will present issues regarding data quality, comparing the accuracy of the contributions from the volunteers with the accuracy of the data collected by a set of experts. We show that a citizen science approach is promising and can complement scientific analyses and can provide potential inputs to policies on landscape restoration. A crowdsourcing approach to image interpretation can also help to shorten the time needed for data collection, making the process more cost-effective. In addition, the collective ownership of the results ensures their legitimacy and increases the chances of data acceptance.

We also focus on transparency and the importance of open data. We present how we have made data generated by the crowd accessible in order to empower citizens in exploring and process the data further, thereby actively participating in environmental decision making.