Scent Smart Toolbox for Engaging Citizens into a People-Centric Observation Web Name of event Location and date Name of presenter, partner organisation

Facts & figures H2020 project funded under the topic SC5-17-2015 Demonstrating the concept of ‘Citizen Observatories’ 10 partners from 6 countries Coordinator: Institute of Communications and Computer Systems (ICCS) 3 Research institutes 3 SMEs 2 NGOs 1 Large industry 1 Public Body Duration: 3 years (9/2016 – 8/2019) EU funding: €3,264,675

Partner map

Motivation Flooding is a major environmental issue related to land-cover/use Urban development, critical infrastructure, soil fertility Assets close to flood plains Driver for economic development Societal risk in case of flood loss potential Pluvial flooding in Coibra, [Source: Imperial College] Flooding in Chennai, [Source: Oxford Analytica] Flooding in Vietnam, [Source: Floodlist]

How does land-cover/use influence flooding? Large scale diversions, earthworks Flood embankments Urbanisation, changing natural surfaces to artificial areas [Source: Noah systems] Agricultural practices related to choice of crop, cropping pattern & soil drainage system Forestation: density of forest, canopy, soil cover Land-use malpractices, e.g. excessive logging, conversion of forested areas into other land- use types, conversion of land through forest fires. [Source: Telegraph]

Scent at a glance Better use of available data, repositories and in-situ monitoring systems through innovative tools and algorithms, aiming at complementarity and not yet another independent framework (i) A number of open platforms exist (e.g. Flickr) with an enormous set of images generated by users, which are often raw and of dubious quality; thus remain unexploited for land resource monitoring. The annotation of these images is a cumbersome task, requiring a substantial volunteer set. Scent will develop and provide freely to applications and websites CAPTCHA plugins, asking Internet users to annotate such images. (ii) Current remote-sensing or satellite-based maps, e.g. Copernicus or Google maps are often outdated, lack precision concerning important parameters for land-resource management (e.g. type of vegetation or obstacles in a river bank) and frequently their segments reflect different time periods within a day or year. This renders them unreliable, especially in order to predict crucial environmental phenomena, such as floods. SCENT will design and apply advanced machine learning algorithms to segment, classify and annotate images from such maps as well as from crowd-sourcing. This task will enhance their precision and they may then be used for search, analysis and visualization of specific areas of interest.

Scent objectives People-centric Observation Web Develop OGC-compliant web services and resources for GEOSS portal Scent crowd-sourcing platform Collect image & text from citizens, thus extend the in-situ infrastructure with citizen-sourced observations Scent intelligence engine Implement innovative machine learning to classify & annotate images/text sourced from citizens & existing open platforms. Serious gaming applications Annotate unutilised images from popular platforms, e.g. Flickr Image: Shutterstock

Scent objectives (2) Quantification of impact of land-cover/use changes Improve flood risk maps and spatio-temporal flooding patterns. Simple authoring tool for policy makers Customise Scent toolbox according to target group, regional or social characteristics. Testing & validation of the SCENT smart toolbox Run 2 large scale pilots of great environmental impact; the urban case of Kifisos river in Attica and the rural case of Danube Delta. Increase active engagement of citizens, associations, local authorities Create a sustainable citizen movement

Scent architecture Scent low-cost data collection tools: Citizens’ portable devices with embedded GPS and cameras capture ground-level images from designated areas where citizens are directed to during field campaigns, motivated by the gaming ARG app). This real-time data will calibrate land-cover maps (Google Earth, OpenMaps, Copernicus). Also, citizens and NGOs or municipalities will use low-cost portable sensors for sampling soil features, essential for assessing land-cover changes related to flood maps. Scent will also deploy ready-to-fly digital sensors mounted on multicopters for a quantitative mapping of environmental parameters (e.g. vegetation state, wetland coverage etc). Scent crowd-sourcing platform: Based on an existing prototype (CIVICFLOW platform, see section 1.3.5), a cloud-based crowd-sourcing solution will be developed to motivate citizens to provide, qualify and interpret information on land-use. This will be delivered in the form of an open, freely accessible hybrid app. Scent serious gaming applications: These will have a strong narrative and collaborative dimension and measurable goals so that players are given a sense of purpose. The aim will be three-fold: i) the online annotation of relevant images from open platforms that were not semantically annotated, ii) the participation in the ARG through actual field visits for new images collection and iii) the implicit annotation of images from a broader Internet community through the Captcha plugin version. A reward system will further foster user engagement. Scent intelligence engine: A machine-learning driven intelligence engine will be developed for a semi-automated classification of outdoor images based on low-level features and discrete object recognition. Feedback from human users through (c) will provide additional calibration. Moreover, the engine will collate user-supplied textual metadata descriptions provided through (b) based on semantic similarity of the natural language descriptors. Scent authoring tool: An easy-to-use customisation tool will be targeted at policy makers of different countries and social characteristics. It will be available as a web application to render citizens and citizen authorities as co-designers of the core SCENT tools, e.g. the crowd-sourcing and gamification platforms. Scent numerical models: They will quantify the aforementioned and existing data sources on land-cover and land-use changes into flood risk maps and spatio-temporal flooding patterns. The output will support the decision making of relevant policy makers (e.g. land-use planners) when assessing their decisions impact on flood risks. Scent harmonisation platform: This will ensure that the observation pool available through the SCENT citizen observatory (annotated images data/meta-data, sensor information, numerical models output quantifying flood risks) will be harmonised according to established data schemas (OGC, INSPIRE) and will be fed into the GEOSS portal and to relevant national authorities in a transparent way using web services.

Large scale demonstrations Urban demo: Kifisos river catchment in Attica, Greece Rural demo: Danube Delta in Tulcea, Romania

Large scale demonstrations (2) At least 200 participants in each demo, aiming to reach ~ 1000 SCENT NGOs (HRTA, SOR) 5 municipalities around Kifisos mobilised by Attica regional authority External volunteer groups & environmental associations Individual citizens mobilised through SCENT dissemination campaigns Birdwatchers @ Danube Delta Online SCENT toolbox validation Download gaming app in personal devices for image annotation from e.g. Flickr Use CAPTCHA plugin and implicitly annotate pool of images from open repositories

Expected impact Lowered cost and extension of Copernicus/GEOSS in-situ components Better decision making through the empowerment & active participation of citizens and associations Improved co-operative planning with special impact on land resources management. Enhanced implementation of governance and global policy objectives Increased deployment and market uptake of innovative in-situ monitoring techniques Increased role in the business of in-situ monitoring of the environment

Target audiences Gaming and software industries, website developers Municipalities, regional governments, ministries of environment Media and social media influencers Civil protection agencies, NGOs, environmental networks Photo-sharing platforms Photo-sharing organizations (eg Flickr) interested in automated classification of user-uploaded content, Social network providers wishing to ‘sell’ user-generated content to municipalities regional governments, ministries of environment so that they enhance their land-resource management processes Open and commercial map providers (within Europe and also internationally) wanting to improve accuracy of their maps without investing on additional in-situ/remote-sensing equipment/satellite subscriptions Municipalities, regional governments interested in improving their land-resource monitoring processes Regional educational directorates or individual schools motivating children on land-use monitoring in a neighbourhood scale, Web-site providers interested on verifying ‘humans’ through more innovative/attractive ways than plain text Citizens including children and schools Open & commercial map providers Scientific audiences

Current status and next steps Development of gaming apps Scent Explore and Collaborate as well as captcha plug-in Integration of sensor measurements, implementation of sensor management collection app Registration of drones and flight plans in Romania and Greece Development of crowdsourcing platform backend and frontend Development of intelligence engine Calibration of hydrological models Preparation for pilots starting in August (Danube Delta) and September (Kifisos)

Dissemination channels Networking with environmental associations, grassroot orgs & NGOs Training Dissemination material Events and publications Early childhood education programmes Links with other Citizen Observatories Digital and traditional channels

Find us on social media Twitter: @SCENT_EU, Facebook:@ScentEU, Instagram: scent_eu YouTube: scent EU, LinkedIn: Scent H2020

Name of presenter Email Organisation This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement no 688930.