Citizen Science 1

The Digital Observatory for Protected Areas (DOPA) is a biodiversity information system developed as a set of interoperable web services at the Joint Research Centre of the European Commission. The DOPA is designed to assess, monitor and forecast the ecological state of Protected Areas (PAs) at the global scale in order to support decision-making and fund allocation processes. Because the collection and processing of information addressing spatio-temporal changes in habitats and anthropogenic pressures on marine and terrestrial protected areas are at the heart of the system, the Copernicus Sentinel missions offer unprecedented opportunities to improve biodiversity conservation.

The purpose of this contribution is to discuss the use in the DOPA of a broad range of remotely sensed products as well as the planned developments. If a number of services based on earth observations can be rolled out without additional efforts in terms of post-processing of the data, most will require some more significant efforts by integrating additional data sources before the information can be delivered automatically. Last but not least, a number of essential products still require significant ground based validation which is likely to come best through a citizen-science based approach considering the global dimension of the required information. The collaboration of the DOPA with such platforms, in this case the Geo-Wiki developed by the IIASA which provides citizens with the means to provide feedback on satellite imagery, will be discussed and the benefits for the community involved in biodiversity conservation further addressed. 

Currently the images of the ISS are the only colour images of the Earth at night. Since acquisition is manually by astronauts not existed any process procedure standardized data. To facilitate the scientific use of these images the project Cities at Night was created. This citizen science project seeks to classify, locate and geo-reference images taken by astronauts. At the same time, it has created a pipeline reduction can generate different maps of impact in different areas. Examples include, Visibility stars, impact on melatonin suppression, energy efficiency and CO2 emissions, aerosol content or impact on air quality.

Large-scale integration of satellite and crowdsourced data was unfeasible until very recently due to the excessive cost of acquiring and analysing large-scale data, the lack of open source data on geographical locations as well as the fundamental cultural distrust for crowdsourced data and its potential quality issues. This has now changed thanks to the public availability of satellite images; the pervasive use of mobile phones; large banks of messages and photos from social media; and the willingness of all stakeholders to collaborate for the common good. Indeed, the scale of this is borne out by the success and the wide-scale benefits of crowdsourcing adoption in initiatives such as Wikipedia, Amazon Mechanical Turk, etc.

The recently (March 2016) finished European Space Agency (ESA)-funded Crowd4Sat project investigated and analysed the potential for ESA to capitalise on crowdsourcing and citizen science in its activities and products with a particular focus on those concerning observations from satellites. This covered how crowdsourcing  could be used for the validation and enhancement of ESA products with the ultimate aim of supporting better data exploitation, improving citizens' lives and educational activities, and encouraging wider science engagement.

We will present the final results from two core activity areas. The first area is an analysis of the current crowdsourcing communities, initiatives, stakeholders, and trends including a review of the challenges and opportunities of incorporating crowdsourcing and citizen science within ESA activities and outreach. The work in this area addressed the historical, societal and economic context for crowdsourcing and citizen science and reviewed in detail approaches for successfully engaging citizens with initiatives and, importantly, sustaining that engagement for the lifetime of the initiative and future initiatives. In addition, a strategic roadmapping exercise was conducted focussing on the short- and mid-term objectives for ESA; the wider EO and space sector; and society as a whole.

The second area targeted key scientific and societal problems through four demonstration projects in which the combination of crowdsourcing / citizen science and satellite-based remote sensing was shown to have a positive impact. The first demonstration project employed participatory crowdsourcing of snow covered area (SCA) from hikers and integrated this with Sentinel-1, ENVISAT, RADARSAT and MODIS data to enhance the precision of snow coverage identification in the Pyrenees. The crowdsourced data was collected via mobile devices using a custom app; this data was subsequently analysed and integrated into an existing, commercial SCA processing chain. The second demonstration project used unprecedented amounts of opportunistically crowdsourced vehicle movement data (telematics) with in situ and satellite remote sensing (CORINE land cover; DEM) for improved pollution modelling and mapping for local authorities of large metropolitan areas. The third demonstration project integrated opportunistically crowdsourced data from social media (e.g., Twitter, Flickr) with OS data from Sentinel-1, Landsat-8 and MODIS to improve the overall quality and timeliness of pure satellite-based flood emergency mapping services. The fourth demonstration project used participatory and opportunistically crowdsourced images combined with CORINE land cover data from IRS P6 LISS III and RapidEye for improving the accuracy and timeliness of land cover information for water management.

Assuring food security is becoming a complex problem because of the ever-increasing global population, environmental degradation and the sharp rise in the frequency of environmental extremes (e.g. drought, pest outbreaks) caused by global change. The food security problem is particularly critical in Sub-Saharan Africa where agricultural production is mostly carried out by small-holder farmers that operate in an information-poor environment. The STARS project (http://stars-project.org) is a coordinated set of activities that looks for ways to use remote sensing technology to improve agricultural practices of smallholder famers in Sub-Saharan Africa and South East Asia. In particular, STARS team members use time series of very high spatial resolution Satellite and UAV data (pixel sizes from 5 cm to 5 m) to generate a suite of products that support various stakeholders involved in food production. However, the analysis of very high spatial resolution time series is not trivial and so the pace of data analysis is typically much slower than that of data acquisition. Outsourcing (parts of) the data analysis to volunteers might help to overcome some of the challenging data analysis tasks and unlock the information content of this data.

In this contribution we will report the first results of a crowdsourcing experiment in which we recruited a group of 40 geo-information MSc and PhD students to visually interpret several color composites and panchromatic bands of Worldview-2 images. More specifically, the volunteers were asked to digitize field boundaries and, when possible, to label the crop being grown in each field. For completeness, non-agricultural classes present in the images were also digitized by the volunteers. The crowdsourcing session started with a short training course of about 1 h. The training reiterated the main fundamentals of visual image interpretation methods and demonstrated the use of QGIS to record field boundaries and labels. The training also discussed the spectral and textural characteristics of the main land cover classes in our study area (Koutiala, Sikasso district; Southeast Mali). After the training the volunteers were divided into 9 groups and each group focused on a particular class, namely: maize, sorghum, agricultural field, bare soil, single tree, group of trees, bare rock, irrigated agriculture and plantations of trees.

The quality of the crowdsourced data is assessed in various ways. First, we look at the consistency of the data. This is possible because several volunteers were assigned the same task (e.g. digitize Sorghum fields). Then, we evaluate the geometric quality of the field boundaries using ground data collected by a professional topographer and the accuracy of the class labels using data collected by the ICRISAT STARS team in Mali. After this, we critically review the possibilities of outsourcing image analysis by assessing the added-value of this crowdsourced dataset to improve classical land cover classification efforts. Preliminary results show that crowdsourcing is a cost and time efficient method to map smallholder fields.

Land coverage maps represent invaluable resources for many different environmental studies and applications. Among the different steps involving the generation of these data the validation process is certainly an important research topic, since the knowledge of land cover classification accuracy is an essential information for its proper use. In recent years the traditional techniques of validation have been integrated by new approaches based on the Web and crowdsourcing initiatives; among these, GeoWiki (Fritz et al., 2009) represents one of the most popular examples where users are involved in land cover data improvement by taking advantage of Google Earth images.

The present work proposes a new Web platform for validation of land cover data based on the GWAP (Game With a Purpose) approach (Von Ahn, 2006). According to this technique, the application is presented as a game which uses scores and badges to involve as many citizens as possible in the validation process. While a player detects the land coverage classes having fun and challenging other players, important data for the validation and the improvement of the land cover data are collected and stored.

The gaming application has been introduced during the Mapping Party for the participant of FOSS4G-Europe conference that was held in Como in July 2015. The participant contributions have been taken into account during the event for the validation of 30 meters resolution open global land cover map (GlobeLand30) by using high resolution aerial photos. The participant's contributions from the game have been used to obtain the confidence level of the GlobeLand30 data. The participants reputation has been evaluated in order to validate the data products more accurately.