Towards intelligent Green and Blue infrastructure

Eirini Malliaraki
16 min readJan 11, 2020

Imagine a city that has cleaned up its rivers and streams, links these to parks, squares, and regenerated brownfield sites, and encourages citizens of all backgrounds to plant trees in public spaces and grow community gardens. Imagine well-managed wetlands, forests and flood plains that absorb water, reduce the risk of floods and increase resilience to climate change. Urban trees and green roofs will be reducing the local temperature, energy demand, and carbon emissions, and offer habitats for urban wildlife. Air and water quality will be improved, our communities will be healthier and will have more opportunities to encounter and interact with nature.

Addressing the interconnectedness of modern urban challenges like public health, flood management, housing delivery, biodiversity decline, and climate change requires a much more dynamic and integrated approach to designing our cities. Green and Blue Infrastructure, as described above, can provide multiple ecosystem services that improve our cities and therefore are key parts of the solution.

However, existing technological infrastructures such as transportation, industrialized energy, and food production and distribution systems give rise to environmental crises. These infrastructures are mediated by information, trade, and networks that change and adapt over time. Given the magnitude, flow, and components of our socio-environmental systems today, the ways we understand and interact with nature in our cities are not fit for purpose anymore. We need a better language to understand the mutually evolving components of urban ecosystems, promote ecosystem regeneration and value nature as an integral component of urban life.

As I will argue in this article, data science, and machine learning can help us make more dynamic digital representations of Green and Blue Infrastructures (GI & BI), proactively design and steward them and connect with nature in new ways.

Many Interlocking Parts

Here’s how these pieces fall together. This is a system’s overview of the links between various ecosystem pressures, conditions, services and policy decisions for cities.

Synthesis of the links between pressures, condition and ecosystem services in urban ecosystems. Adapted from EU’s Mapping and Assessment of Ecosystems and their Services [1] and visualised by me.

Unsustainable land take, air and water pollution, noise, and invasive species are some of the pressures facing our cities. We know this because we have condition indicators and various datasets that capture the physical, chemical and biological quality of these ecosystems. For cities, these include urban temperature, air and water quality, noise levels, and metrics that assess the share and composition of built area in relation to population density. Different relationships between pressures and conditions emerge with land take having an impact on all the indicators and affecting urban temperature, air quality, and species diversity.

Ecosystem conditions are then strongly connected to the delivery of urban ecosystem services. For example, temperature is key for several ecological processes and subsequently affects the provision of ecosystem services. Ecosystem services are inextricably connected to different policy objectives such as reducing the impact of heatwaves. For example, many cities have targets for reducing maximum summer temperature and for regulating the local climate. Trees and shrubs provide shade and have a cooling impact on the surrounding neighborhoods. So, monitoring urban temperature, soil sealing, the composition of green space and the rate of land take at the cost of ecosystems with a cooling capacity are key indicators to inform decision making.

Learning Urban Environments

The combination of new data sources, increased computational power, and recent advances in machine learning offer new opportunities for expanding our knowledge about natural ecosystems in our cities from data. To achieve its full potential, we need a common vision for an open digital ecosystem of urban environmental data, algorithms and policy insights (see The Case for a Digital Ecosystem for the Environment, UN, 2019).

It would be an open, adaptive and distributed system — just like a biological organism. This digital ecosystem can provide the situational intelligence we need to secure fair stewardship of food, water, biodiversity, health, energy, materials, and other ecosystem functions and services. It can also help us evaluate strategies for governing blue and green infrastructures across sectors and scales and achieve long term sustainable transitions.

This digital ecosystem poses lots of unique and interesting questions related to:

Sensing and data

  • How do we optimally place and use sensors and other methods to gather data?
  • How do we best store, curate and openly share this data?
  • How do we create standards and ensure quality?
  • How do we identify what data and knowledge are available, who the providers are, and ensure it is accessible for users?


  • How do we catalogue and integrate data?
  • How do we create safe environments/havens/APIs to share ready for analysis data?
  • How can we democratise access to HPC?
  • How do we better expose existing catalogs of datasets and create collaborative environments that make this data easy to access and visualise?

Data Science & AI

  • How can we fuse information from multiple data sources with uncertainties and biases that vary over space and time;
  • How can we ensure privacy, fairness, accountability, robustness, transparency for environmental analytics and systems?
  • How can we achieve physical consistency via hybrid simulation and data science approaches?
  • How can we integrate probabilistic forecasting methodologies into decision-making?
  • How can we better understand risk and model extreme events?

Applications and Governance

  • How do we best translate abstract mathematical outputs into actionable, contextual and relevant policy insights?
  • How do we develop open-source tools and user-friendly interfaces?
  • How can better support the climate adaptation activities of local communities, including citizen scientists?
  • How can we better incentivise and involve users and local communities in the development of such interventions?
  • How can we develop better tools and resources to cover social and economic implications across timescales?
  • How do we equip individuals with the skills, information, and training to enable them to generate, communicate and use decision-relevant data?
  • How do we create new processes and tools that enable effective ecosystem services governance, not only within organizations but between the different organizations and sectors?

Where can data science intervene

In the following sections, I give examples of how data science is used in the context of green and blue infrastructures in order to: a) help us expand our understanding of space and time; b) proactively steward and maintain UGI; c) collectively sense and act; d) create new planning and design affordances, and e) create post-human encounters.

a) Expand our understanding of space and time

A multiscale approach is needed to assess the total contribution of ecosystems towards human well-being and to understand the incentives that individual decision-makers face in managing ecosystems in different ways. The use of new datasets and distributed sensor networks can expand our spatiotemporal understanding of nature in cities.

We can use GIS data to identify and describe the full range of urban green spaces we could not systematically see before. For example, project Greensurge has developed an inventory that contains 44 green infrastructure elements like facade-bound green walls, bioswales, house gardens, botanical gardens, cemeteries, green allotments, biofuel production sites, as well as blue components like estuaries, canals, and marshes.

Sensor networks for monitoring stormwater, air pollution uptake, and urban heat lands have also been deployed in various cities and enable us to monitor ecosystem pressures at a more granular level.

48-hour predictions of air quality (NO2) Central London and a running route that changes shape to minimise air pollution. Source: Alan Turing Institute

One such example of tracking ecosystem pressures is CityNet. Citynet is a bioacoustic assessment tool made of two neural networks — CityBioNet and CityAnthroNet — which measure biotic and anthropogenic audio activity respectively. When used together, they accurately measure daily patterns of both biotic and human acoustic activity from real-world urban sounds and facilitate the investigation of the impacts of anthropogenic activities on wildlife.

New data sets have also been used to study long, cyclic, natural phenomena, especially in relation to climate and plant and animal life in cities. Ecologists from Ghent University analyzed four decades of archival footage from cycling races to study climate change impacts on trees. They found that the trees had advanced the timing of leafing and flowering in response to temperature changes. In footage after 1990, more and more trees were visibly in bloom at the time of the race.

Archive footage of the Tour of Flanders obtained by Flemish broadcaster VRT — Flanders Classics

Another example comes from Green City Watch, who identified 531 parks in 26 cities in Indonesia using publicly available data from OpenStreetMap and calculated the park’s infiltration capacity, green versus paved ratio, and the number of trees. This could provide a new systemic understanding of the health of these places so that we can take better care of them.

Example visualization built for Indonesian policymakers to display the ‘bare ground’ metric for the pilot study, Source: Green City Watch

b) Proactively steward and maintain

Stewarding and maintaining green infrastructure is essential for maximising its benefits and health. Some infrastructure such as green roofs require minimal maintenance but parks need mowing, weeding, and watering. These costs often fall to local authorities, which have been facing budget cuts in recent years.

To address this challenge, TreeMania mounted sensors in 5,500 trees in the Netherlands. These sensors collected real-time data on soil moisture and sent emails updates to tree managers and workers. This sensor network was installed after a severe drought caused significant tree loss from June to September 2018 and helped local authorities keep newly planted trees alive [3]. Also, in the Dutch village of Geijsteren, instead of tree managers, local residents receive the watering updates and could directly participate in tree maintenance.

Greencity Watch also used Synthetic Aperture Radar (SAR), an active microwave remote sensing technique to map standing water in Amsterdam’s Noorderpark. These initial, promising results can help local authorities prevent infrastructure degradation and tree rot caused by excessive flooding.

Synthetic Aperture Radar analysis of Amsterdam’s Noorderpark during a flooding event. Source: Green City Watch

c) Collectively sense and act

New data-driven platforms enable a new form of intelligence to emerge: distributed, real-time and connected. Groups of citizens can now channel their insights, views, and data towards tackling urban challenges and coordinate their actions at an unprecedented scale.

Several mapping and crowdsourcing initiatives now provide large scale sensing capabilities for our cities. For example, with Forest Watcher app citizens can monitor areas of interest, view deforestation and fires alerts and collect their observations regardless of connectivity.

Another big tree mapping project is Treezila. Participants are invited to contribute data to scientific investigations and they can design, execute and publish their own, autonomous investigations based on the Treezilla dataset. To this day they have mapped 1,087,389 trees across the world and they have calculated how much local air quality has improved, how much stormwater was filtered and how much carbon dioxide was removed and stored.

SPIPOLL, a participatory science project, aims to study pollination networks i.e the complex interactions between plants and insects, but also between the visitors of the flowers themselves. Another project comes from London, where Primary school pupils carried backpacks fitted with air quality sensors on their journey to and from school to help monitor the levels of particulate matter and nitrogen dioxide in the air. The data allowed scientists to analyse at which point of their journey school children are exposed to the most pollution and then make recommendations for how they can reduce their exposure in the future.

Source: Kings College London

Crowds can also discover new information. In 2016, 2,300 volunteers in New York City mapped every single street tree recorded the presence of trunk and bark damage. This was further analysed by researchers who found the invasion of the Callery pear, a species responsible for repulsive smell, but also being susceptible to bark inclusions and limb breakage during strong storms [4].

Last but not least, data from social media can provide new insights about the location and quality of green infrastructure. For example, A team at University of Illinois at Urbana Champaign used social media data from Flickr and Instagram, satellite aerial images and computer vision to identify urban green stormwater infrastructure. They were able to detect potential green infrastructure sites, which can then be investigated in more detail using on-site surveys. This approach is tackling the challenge of GI being installed by various different parties and cities who typically do not know where GI is located. They aim to make the study of its impacts and developing new GI easy.

d) Create new planning and design affordances

Cities need to make lots of decisions about the planning of green and blue infrastructures. They need to strategically connect these networks, plan for climate change, locate the best places for habitat enhancement [5], guide infrastructure developments away from sensitive areas, and identify multi-functional zones where compatible land uses can support healthy ecosystems.

Planners can now also integrate human preferences and behaviours into their planning. They can explore a larger parameter space that includes for example how appropriately designed green spaces affect human wellbeing and health. For example, researchers at the University of Illinois at Urbana Champaign developed [6] a design framework that integrates stormwater management requirements with criteria for human wellbeing. The team used machine learning to identify specific patterns that promote human wellbeing in urban green spaces. This model was then linked to hydrological models to evaluate GI designs in terms of both water resources and human health benefits.

Decisions don’t only affect humans. For example with “Building for Birds” online tool, decision-makers and planners can evaluate different development scenarios and how they affect habitats for different species of forest birds in fragmented areas. Planners can manipulate forest fragments and tree canopy and determine the best designs for conserving bird habitat, which is useful in already fragmented landscapes where planners are trying to decide which tree canopies to conserve.

Highline aerial view in NYC. Source: David Shankbone (creative commons license)

Researchers have also used data from Twitter and Flickr to understand park access and visitation in New York City and Minnesota, study interactions with urban green spaces in Birmingham, UK and understand how people engage with different social activities in green spaces in New York City.

f) Create post-human encounters

Data has been used in other interesting ways to highlight or create new connections with otherness. For example, Natalie Jeremijenko in her project Mussel Choir mounted hall effect sensors onto mussels. One mussel can filter as much as 6–9 litres of water/ hour, and by converting the data into sound, the artwork uses the behavior of the mussels as a biologically meaningful measure of pollutant exposure. Stormwater run-off, local weather, and seasons affect the Choir’s performances in new ways that are now visible and audible to the public.

Source: Natalie Jeremijenko
Source: Natalie Jeremijenko

Marshmellow Laserfeast created an artistic interpretation of the sensory perspectives of three species -a dragonfly, a frog, and an owl- that live in British forests. The team used lidar scans, drones, bespoke 360° cameras, and binaural recordings from Grizedale Forest (UK) to create an animated, real-time forest landscape. The audience can now ‘see’ through the eyes of these species, on their journey through the food chain.

Source: Marshmallow Laser Feast

The Intelligent Guerilla Beehive by AnneMarie Maes is a sensing device that mirrors the pollution of the environment. Through colonies of ‘color-changing’ bacteria –living on the skin of the hive- it sends out warnings. At the same time, it is a device that monitors the bees’ wellbeing using computer vision. It is a radically new beehive designed for urban environments.

Source: AnneMarie Maes

Last but not least, Semiconductor films developed a sculpture made from one year’s worth of measurements of the take-up and loss of carbon dioxide from the forest trees, collected from the top of a 28m high flux tower Alice Holt Research Forest. By humanising and contextualising the data, the artists offer a new perspective of the natural world and the ways we are documenting it.

Source: Semiconductor films


Modern urbanisation depends on technological networks of urban infrastructure which structure material metabolisms, i.e. the inflows of material and energy, the outflows of waste and emissions and the retention of materials in the built environment. Given that these technological networks mediate nature and society we should pay more attention to the power mechanisms and unintended consequences they might have.

In a city where citizens may be turned into sensing devices, we need to ensure that their agency is protected. There are questions about how AI tools interfere with our ability and need to connect strongly to nature as well as our ability to transfer this experience to others. We also need to ensure that social and structural inequalities are not exacerbated between the cities of the Global North/South and also within wealthier urbanized countries where disadvantaged communities do not have access to the same investments for smart cities initiatives. Smart cities may create further challenges and unintended consequences on human and ecosystem health. Studies have shown that radiofrequency exposure can damage trees [7], and alter insect behaviour and physiology [8].

Cities also need to be very careful regarding data governance and ensure that public goods and data, as well as citizen’s data privacy and security, are protected from the profit-making objectives of the private sector. For this, alternative visions of governance of green infrastructure have been proposed by design studios like Dark Matter Labs. Their project Trees as Infrastructure provides a new model for nurturing urban forests as live public infrastructure. They propose using smart micro-procurement contracts for distributed tree planting and management. The team aims to highlight the true contribution of natural assets in our cities and reshape how humans engage with them. A similar model of decentralised sustainability is put forward by David Dao.

Source: Dark Matter Labs

Intelligent Urban Metabolic Systems for Symbiotic Cities (and violets)

In this post I demonstrated some of the ways that Green and Blue infrastructure circulates in cities as well as the value of data science and machine learning in understanding, stewarding, sensing, planning, acting and interacting with them.

Cities can be conceptualised as living organisms made up of many different and co-evolving man-made and natural elements. Using the notions of flow and circulation of matter, energy, and information, urban metabolism can link material flows with ecological and social processes [9].

The first wave of urban metabolism studies (see Odum) described these flows in terms of energy equivalents (Emergy), while the second wave takes a broader approach that expresses the city’s flows of water, materials, and nutrients as mass fluxes (using material flow analysis) [10]. These waves have, however, paid little attention to political changes and have conceptualised nature as a static agent.

The latest research paradigm is driven by large and diverse datasets and learning algorithms that have become accessible to researchers and citizens in an unprecedented way. These tools can enable new conceptions of urban metabolism that go beyond just material cycles to include various dynamic, interconnected, and mutually transformative physical and social processes [11], including Green and Blue infrastructure flows, pressures, and services.

Future urban metabolism studies will take into account the complexity of these relationships, account all flows entering and exiting cities, measure their environmental effect and generate new co-evolutionary sustainable pathways [12]. This will lead to a re-conceptualisation of the relationship between the urban and nature. From this perspective, urbanisation will not distance humans from nature but will be “a process by which new and more complex relationships of society and nature are created” [13] and cities will become hybrids.

Violets give us a cautionary analogy for this symbiosis. Violets contain beta-ionone which stimulates our scent receptors and temporarily shuts them off completely. Their scent is ephemeral- we cannot perceive their scent more than a few moments at a time and then, after a few breaths, the scent comes back again and registers as a new stimulus. With increasing technological mediation, we need to be careful not to lose our playfulness, attentiveness, and connection to the natural world around us. This is core to our humanity and we shouldn’t become anosmic to it.


  1. Maes J, Teller A, Erhard M, Grizzetti B, Barredo JI, Paracchini ML, Condé S, Somma F, Orgiazzi A, Jones A, Zulian A, Vallecilo S, Petersen JE, Marquardt D, Kovacevic V, Abdul Malak D, Marin AI, Czúcz B, Mauri A, Loffler P, BastrupBirk A, Biala K, Christiansen T, Werner B (2018) Mapping and Assessment of Ecosystems and their Services: An analytical framework for ecosystem condition. Publications office of the European Union, Luxembourg.
  2. Van Velzen, J., Modderkolk, P., Bouma, J., 2018. Nederland zucht onder de droogte [The Netherlands sighs under drought].
  3. Nitoslawski, S.A., Galle, N.J., van den Bosch, C.K. and Steenberg, J.W., 2019. Smarter ecosystems for smarter cities? A review of trends, technologies, and turning points for smart urban forestry. Sustainable Cities and Society, p.101770.
  4. Nitoslawski, S.A., Galle, N.J., van den Bosch, C.K. and Steenberg, J.W., 2019. Smarter ecosystems for smarter cities? A review of trends, technologies, and turning points for smart urban forestry. Sustainable Cities and Society, p.101770.
  5. Barrile, V., Bonfa, S. and Bilotta, G., 2017, August. Big Data Analytics for a Smart Green Infrastructure Strategy. In IOP Conference Series: Materials Science and Engineering (Vol. 225, №1, p. 012195). IOP Publishing.
  6. Rai, A., Minsker, B., Sullivan, W. and Band, L., 2019. A novel computational green infrastructure design framework for hydrologic and human benefits. Environmental Modelling & Software, 118, pp.252–261.
  7. Waldmann-Selsam, C., Balmori-de la Puente, A., Breunig, H. and Balmori, A., 2016. Radiofrequency radiation injures trees around mobile phone base stations. Science of the Total Environment, 572, pp.554–569.
  8. Thielens, A., Bell, D., Mortimore, D.B., Greco, M.K., Martens, L. and Joseph, W., 2018. Exposure of insects to radio-frequency electromagnetic fields from 2 to 120 GHz. Scientific reports, 8(1), p.3924.
  9. Dijst, M., Worrell, E., Böcker, L., Brunner, P., Davoudi, S., Geertman, S., Harmsen, R., Helbich, M., Holtslag, A.A., Kwan, M.P. and Lenz, B., 2018. Exploring urban metabolism — towards an interdisciplinary perspective.
  10. Rapoport, E., 2011. Interdisciplinary perspectives on urban metabolism. Development.
  11. Heynen, N., Kaika, M. and Swyngedouw, E. eds., 2006. In the nature of cities: urban political ecology and the politics of urban metabolism. Routledge.
  12. Donges, J.F., Lucht, W., Müller-Hansen, F. and Steffen, W., 2017. The technosphere in Earth System analysis: a coevolutionary perspective. The Anthropocene Review, 4(1), pp.23–33.
  13. Keil, R., 2003. Urban political ecology: a progress report. Urban Geography, 24(8), pp.723–738.