Introduction

Urban growth and densification in recent decades are putting major pressure on biodiversity and ecosystems in general, through the reduction and fragmentation of sustainable habitats and degradation of natural resources (Horváth et al., 2019). Plants and wildlife face more difficulties in surviving and moving through landscapes dominated by humans, as urbanization reduces the amount and quality of blue (i.e., water) and green (i.e., vegetated) areas that support life. This affects important biological functions such as reproduction, spreading, migration and resource use, and eventually lowers the genetic exchange between populations, which can affect their variation and adaptation (Donati et al., 2022).

Nature-based solutions (NBS) are defined as multi-benefit ecosystems, often engineered but always inspired and supported by nature, that cost-effectively “provide environmental, social and economic benefits and help build resilience” (Cohen-Shacham, 2016). Biodiversity is therefore inherent part of the NBS, which aim for more diverse natural features supporting pre-urbanisation-like levels of ecological services. However, NBS still predominantly target human-centric benefits and comforts, ranging from low to high levels of engineered elements, encompassing blue and green features, while connected to “grey” (man-made) infrastructure. NBS has been gaining popularity recently due to its multifaceted approach to address a number of local issues from climate, water and air pollution to green space promotion for biodiversity and inclusion. While biodiversity is often not considered in traditional urban planning, NBS gives us a chance to reconcile humans and nature to co-exist in urban spaces, which is also increasingly explored in architectural design thinking (e.g., Braidotti, 2020; Haraway, 2016). Solutions like tree pits, rain gardens, wetlands, and other NBS, provide a wealth of habitats and resources for biodiversity conservation in cities and rural areas, while at the same time increasing human health, decreasing urban heat, and managing urban water and pollution (European Commission et al., 2020). However, this exact multifaceted nature of NBS is challenging to urban planners, due to a large number of, sometimes conflicting, priorities (Kuller et al., 2019). Human economy, amenity, health, and well-being are usually at the core of new urban NBS practices. Urban planning is usually focused on serving the general public, but through the vision of a few individuals, without fully understanding community dynamics, the history of the place, or changing priorities (Coyne et al., 2020). More “liveable cities” (Pineda-Pinto et al., 2022) is something everybody is striving for. However, liveability has a varied meaning in different communities. Additionally, economic, and political circumstances often dictate NBS investments and urban planners need to adapt to them (Iftekhar et al., 2022). Hence, the planning and design of NBS have been ad-hoc, and opportunistic, with biodiversity often treated as a nice ‘benefit’ rather than an active goal (Kuller et al., 2021). Consequently, non-human, or ecologically-centric (eco-centric) urban design often poorly incorporates plant and animal benefits. While some urban planning rules dictate immediate actions towards biodiversity promotion, these are usually not sustained over the project life-time and often become neglected when human-centric issues surface, as witnessed in most major infrastructure projects in the cities (Hawken et al., 2021). Additionally, researchers are arguing that we are missing a more profound understanding of how to enhance biodiversity in urban spaces, and, furthermore, how to incorporate this in explicit planning approaches (Ignatieva et al., 2023). NBS could be the key, but how do we reconcile so many evolving priorities between human and non-human users of blue-green spaces?

A further opportunity and complexity is the advancements in computer science, which have made planning of multilayered NBS systems significantly easier over the years (Kuller et al., 2019). Many urban planning models include NBS integration in the urban fabric (e.g., Kuller et al., 2019; Bach et al., 2020; Prodanovic et al., 2022b), but models are still limited by their users’ initial boundary conditions, or model assumptions, and are often run for human-centric design. Even the most advanced models nowadays are not capable of incorporating all the human-centric considerations (e.g., history of the space, shifting community interests, etc., - Coyne et al., 2020; Naserisafavi et al., 2022), and ecological considerations are even less promoted. However, with the rapid and recent advancement in Artificial Intelligence (AI) technology, we are starting to incorporate AI techniques in decision-making where multicriteria decisions are not easily balanced by our own experiences (Koumetio Tekouabou et al., 2023). With the increased amount of data worldwide and the expense of AI capabilities, there is a question of how we can best leverage AI (particularly machine learning) to create integrated NBS solutions for co-existence between humans, animals, and plants.

In this paper, we cover recent efforts in urban NBS design across human-centric and eco-centric solutions, while also looking into AI-focused developments in urban planning, and discuss how do we integrate machine learning in creation of positive NBS biodiversity outcomes. We have intentionally separated human and ecological aspects in this discussion, which may appear non-sensical to some and obvious to others. This, however, reflects a long-standing dichotomy and state-of-mind in environmental engineering and urban planning in industrialized contexts, that is actively being questioned (e.g., Gray, 2002), also in light of climate change, which renders our ‘reactive’ approaches as no longer sustainable and requires us to revisit the natural cycles.

Methodology

This study employs a multi-faceted approach to comprehensively explore the planning and implementation of NBS in urban environments. Hence, we integrated narrative review, integrative review, and systematic review methodologies.

Central to our methodology is the narrative review, chosen for its capacity to construct a cohesive and comprehensive story elucidating the multifaceted dimensions of urban NBS planning. Drawing inspiration from the work of Greenhalgh et al. (2005) on narrative synthesis, our narrative approach allows us to weave together historical context, theoretical frameworks, and contemporary practices. Embracing a human-centric, eco-centric, and AI-centric lens, our narrative review illuminates the complex interplay between socio-cultural, ecological, and technological factors shaping NBS strategies in urban settings.

Complementing the narrative synthesis, an integrative review methodology is employed to synthesise a diverse range of studies across disciplinary boundaries. As advocated by Whittemore et al. (2005), our integrative review process involves systematically analysing and synthesising the findings from different sources to construct a comprehensive understanding of urban NBS. By bridging disciplinary differences and identifying connections between seemingly varied research strands, the integrative review facilitates a nuanced examination of NBS planning, fostering a holistic perspective.

Furthermore, a systematic review approach is utilized to delve into recent advancements in NBS research, focusing primarily on literature from the past five years. Our systematic review process ensures a thorough and unbiased examination of the latest scholarly contributions in humanitarian engineering, biological and ecological sciences, and computer sciences. By synthesising contemporary insights with foundational works, our systematic review enhances the robustness and relevance of our findings, providing valuable insights for researchers and practitioners alike.

To prevent geographical bias and promote inclusivity, our study adopts a global perspective, drawing upon literature and practices from diverse regions worldwide. This approach aligns with the call for global collaboration in addressing urban sustainability challenges (Colding et al., 2019). By synthesizing knowledge from varied contexts, we aspire to foster cross-cultural learning and facilitate the exchange of innovative NBS solutions on a global scale.

In summary, our methodological framework integrates narrative, integrative, and systematic review methodologies, underpinned by a commitment to inclusivity and methodological rigor. Through this multi-faceted approach, we aim to provide a comprehensive and nuanced understanding of urban NBS planning, contributing to both scholarly discourse and practical decision-making processes.

Human-centric urban NBS design

NBS practices, arguably, have the highest importance in the urban context, where rapid urbanization is chipping away at natural spaces and blue-green environments are at a premium. Urban NBS, at its core, is designed for the benefit of humans by providing re-naturalisation, hence it is no surprise that human-centric approaches towards NBS design are dominant. While nature alone can increase the resilience of urban spaces to diverse climate phenomena (e.g., temperature fluctuations, severe rainfall events, earthquakes, etc.), this type of resilience is not always optimal for human well-being (Nguyen et al., 2022). Through significant engineering efforts over the past three decades, urban NBS practices have been adapted to provide the most benefits to humans for city cooling (e.g., trees, parks, wetlands, etc.), flood protection (e.g., vegetation barriers, retention basins, rain gardens and wetlands, etc.), and pollution reduction (arguably all vegetated systems are involved in this) (Fletcher et al., 2015; Deletic et al., 2019; Probst et al., 2022). These are some of the most important engineering achievements for NBS design and implementation. However, the real selling point of urban NBS technologies are less-tangible benefits (harder, if not impossible to objectively evaluate and quantify) which are health benefits, amenity increase, and economic benefits. Green, natural environments are known to increase human happiness and comfort levels (Wai et al., 2022), while there are studies that show a significant increase in mental and physical well-being (e.g., Bowen et al., 2017). Due to this, NBS is desirable in urban environments and people are willing to spend more money on properties that have NBS features close-by. For example, a study in Sydney, Australia, showed a 6% percent property value increase within 50 m of rain gardens, equating to a total $1.5 million Australian dollars in value of a single rain garden (Polyakov et al., 2015). While researchers have tried to design frameworks that evaluate these non-tangible benefits (e.g., Hamann et al., 2020), these are always location-specific and have not been widely accepted in the research community. While non-tangible benefits are less represented in the typical NBS cost-benefit analyses frameworks, literature shows that these benefits (such as amenity, well-being, historical context, etc.) are often the most important for the communities. Researchers have seen only marginal interest of the communities for practical, tangible benefits such as pollution or flood control (Naserisafavi et al., 2022), and even less interest in biodiversity outcomes of NBS (Ignatieva et al., 2023). Hence, there is still a need for well-developed, inclusive framework to monetise all benefits provided by NBS in urban settings.

While engineers and scientists have been working hard on objectively optimising the performance of nature, application of such technologies has not been as widespread as we hoped. The economic sustainability of current NBS practices and technologies is still questionable (Langeveld et al., 2022). It heavily depends on complexity of NBS applied (e.g., whether it is a very technical solution requiring specific operation and care, or common natural feature), longevity of the solution (short-term or long-term management and care), and method used for economic evaluation (evaluation based only on direct costs and profits, or it also includes benefits of ecosystem services). While community members reap some of the economic benefits through increased property prices, this only marginally comes back to local governments (as a property tax), which also have a cost of maintaining these blue-green systems (i.e., mowing, pruning, fixing, cleaning, etc.). Being one of the most important requirements for the proper functioning of NBS, these maintenance requirements are especially challenging for local governments due to expenses and skills shortage (Oral et al., 2020). Concepts such as zero-additional maintenance NBS have been proposed to try to alleviate these costs (Prodanovic et al., 2022a), however, these are only isolated cases, and not designed for all types of NBS assets. Nevertheless, while economically not sustainable, local governments are still implementing urban NBS to appease local communities, which further deepens human-centric NBS design. This, however, can create an issue of social inequality in the city or, in worse cases, gentrification (Pineda-Pinto et al., 2022), where only higher socio-economic areas (local municipalities) can afford NBS assets, creating a larger divide between social groups within the city (Kuller et al., 2019). While this is not always the case (e.g., Kuller et al., 2021), social inequality can be a major contributor to a negative perception of urban NBS, preventing its widespread application. Additionally, current convenience-based approaches for NBS design, where NBS is constructed alongside other major grey infrastructure projects (e.g., road upgrades, Kuller et al., 2021), do not systematically address urban challenges, but rather create patches of disconnected blue-green spaces reducing community and nature benefits (e.g., prohibiting large green or blue open spaces for recreation or migration of some biological species, etc.).

To truly design urban NBS for human benefit, several considerations need to be made, which are mostly overlooked in current designs (Coyne et al., 2020; Naserisafavi et al., 2022). There is a significant benefit to considering public co-design of urban NBS. It is shown that communities value their urban environment more if they are involved in decision-making and design (Dushkova et al., 2020), and such increased ownership could lead to lower maintenance requirements for local government (Naserisafavi et al., 2022). This process would increase the communication between professionals, landscape planners, local government and the public, which would benefit all parties and at the same time provide the necessary level of education about the potential benefits of NBS. Studies showing such communication and planning have proven positive results for NBS sustainability (Ignatieva et al., 2023). Additionally, human-centric design should consider the history and culture of the place when deciding on appropriate NBS design (Coyne et al., 2020). This is promoted by the concept of Culturally Inclusive Water Urban Design (CIWUD) (Coyne et al., 2020), and further supported in many regions of the world (e.g., China (Hawken et al., 2021), South Africa (Gxokwe et al., 2020), Australia (Frost et al., 2023), India (van der Meulen et al., 2023), etc.). Finally, in an attempt to re-naturalise the cities, sometimes we forget that increased biodiversity in urban areas can negatively affect humans (so-called ecosystem disservices – e.g. von Döhren et al., 2015). Recently, this was seen in the Chinese city of Chengdu, where green facades and vertical forests attracted mosquitoes, resulting in lower human habitation in adjacent buildings and the decline of facilities (Agence France-Presse, 2020).

Eco-centric urban NBS design

Various approaches for ecological management have devoted efforts to the protection of existing, valuable habitats (‘reservation ecology’), the repair and rehabilitation of degraded habitats (‘restoration ecology’) and, in recent decades, the design of habitats in co-existence with humans and the urban environment (‘reconciliation ecology’) (Rosenzweig, 2003). Reconciliation has, in particular, garnered significant attention in the architectural discourse, where post-human ecologies (Braidotti, 2020) and ‘cohabitation’ (Haraway, 2016) are emerging themes in design thinking. They actively place the worldview of human superiority over nature in question and seek to redefine human’s role as a steward of nature and the environment as one designed for all living organisms (Gray, 2002). Irrespective of the form of ecology adopted, urban NBS practices provide the enabling means to protect and restore natural capital.

Habitat degradation and landscape fragmentation, as a result of urbanisation, reduce not only species diversity, but also inhibit species movement (physical dispersal or pollination processes) and, hence, diverse genetic exchange within the adaptation to a changing environment (Donati et al., 2022). As an intrinsic element in the very definition of NBS (Cohen-Shacham, 2016), ‘biodiversity’ still faces challenges in practical implementation due to a lack of understanding of how to tailor certain solution to the variety of ecological design factors (e.g., Donati et al., 2022) and human needs. This is further complicated by the need to consider both terrestrial and aquatic ecosystems as well as their potential linkage (e.g., amphibian species spend their life cycle in both – Cayuela et al., 2020), think about flora and fauna as well as symbiotic relationships or predatory-prey relationships, the presence of invasive species (e.g., invasive neophytes – Hejda et al., 2009) and how broader climatic impacts are likely to impact these ecological processes and dynamics in future.

Early efforts of incorporating ‘biodiversity’ into NBS planning and design relied on suitable proxy indicators that represent habitat quality or landscape connectivity (Nguyen et al., 2021). More explicit and recent insights can be loosely grouped around regional-scale efforts to understand landscape connectivity, and local-scale efforts to decide on design elements of NBS (e.g., plant selection) to improve local biodiversity and habitat quality (Prodanovic et al., 2019a, b). These are often supported by the economic assessment to make a ‘business case’ around ecosystem improvement (Locatelli et al., 2020).

Studies have investigated ways to improve the spatial connectivity relationships between the urban and natural environment (Donati et al., 2022; Molné et al., 2023), highlighting that understanding the process requires broader regional considerations before local-scale approaches of ecological management practices and urban NBS options can be tailored. Methods for quantifying ecological connectivity are reasonably well-established and modelling approaches are based on either island biogeography, metapopulation dynamics, graph theoretical approaches such as least-cost path, or more recently established circuit theory (Dickson et al., 2019). Extensive reviews in the ecological scholarship have investigated a plethora of different approaches for predicting connectivity as well as their shortcomings (e.g., Tischendorf et al., 2000; Moilanen et al., 2002). Irrespective of method choice, this regional-scale approach enables us to understand how urbanisation leads to pinch points in ecological corridors and what can be done to protect or restore connectivity through local-scale NBS interventions (i.e., the active use of vegetation and water elements). They enable testing of how such interventions can act as ‘stepping stones’ to bridging fragmented landscapes (Donati et al., 2022) and in what order their implementation should be prioritised (Molné et al., 2023).

Urban NBS have the potential to enhance the variety and abundance of animal and plant species in a city, which define urban biodiversity (Guerry et al., 2021), often providing greater habitat quality compared to regular urban greening (Filazzola et al., 2019). In designing a specific urban NBS system (e.g., bioretention, green roof/facades), the plants play a role as an important element for assimilation of pollutants from stormwater runoff or in providing evaporative cooling for heat mitigation and as an entry point for improving local biodiversity (Dagenais et al., 2018). Nevertheless, floral species (native vs. non-native) planting strategies and other design elements and objectives require a careful balancing act that is an active area of research.

In parallel to local- and regional-scale efforts, we must also understand the economic consequences and benefits around biodiversity loss and improvement. Valuation studies have approached urban NBS in a number of ways as also detailed in the previous section, but some estimated broad monetary ranges for habitat provision and improvements to urban ecosystems include 15% of land value (Locatelli et al., 2020) or up to 55.24 €/m2 (Johnson et al., 2019). Whilst this pales in comparison to the estimated financial impact of global biodiversity loss ranging in the trillions of dollars worldwide (Costanza et al., 1997) and the limitations of putting monetary units on species extinction, these economic studies (Johnson et al., 2019; Locatelli et al., 2020) at least provide an initial leverage to more deeply embed biodiversity considerations in active planning and design of urban NBS.

Actionable research on the design of urban NBS for biodiversity will continue to emerge in the coming years. Coupled with local-scale efforts to create local habitats, we will start see how our urban NBS can be improved for more local biodiversity benefits. Despite its critique (Suganuma et al., 2022) and seemingly loose guidance, the Field of dreams hypothesis, which states ‘build it and they will come’ appears to currently be the best modus operandi, particularly in the case of wetland restoration projects (Palmer et al., 1997) much like opportunistic approaches of urban NBS for other uses (Kuller et al., 2021) and a potential pathway towards consciously incorporating biodiversity in urban NBS planning. Much like the examples of ‘happy accidents’ in Rosenzweig (2003) reflection of successful reconciliation ecology examples, we must learn by doing at the local-scale and guide the implementation strategically with regional-level understanding, a clear economic ‘business case’ and, finally, good governance frameworks to support ongoing efforts (Xie et al., 2020).

Advancements in AI urban planning and biodiversity protection

Recent progress in data science, in conjunction with Internet-of-things (IoT) and remotely sensed data, are boosting the application of Artificial Intelligence (AI) in biodiversity monitoring and predictive modelling. It is attributed to the fact that AI offers substantial potential for optimal and, to some degree, autonomous biodiversity protection and utilization of ecosystem resources in a fast-changing and resource-limited world (Silvestro et al., 2022). For instance, the automated processing of satellite imagery can yield maps showing the distribution of individual species or groups of species within the large spatial extent, levels of species richness, and biodiversity list specific to each location analysed (Nunes et al., 2020). New technologies for documenting and monitoring species occurrences (e.g., Bolliger et al., 2020) as well as public engagement and open data repository platforms such as the Global Biodiversity Information Facility (GBIF) are providing a plethora of available and location-specific data points for detailed modelling and assessment.

Biodiversity is also an important objective in urban planning broadly recognizing two schools of thought: the first advocates for planning support systems, whereas the second promotes generative design systems, which leverage computing to create innovative and efficient spatial designs (Koumetio Tekouabou et al., 2023). The perspective of generative design systems takes into account the advancements in computational capabilities and data accessibility. Consequently, it emphasises the use of AI-driven solutions in urban planning, particularly focusing on the application of AI-based optimization algorithms for enhancing urban water planning processes (e.g., Fig. 1). A novel approach in urban planning, known as the Smart Design Framework, is thus gaining prominence (Koumetio Tekouabou et al., 2023), where AI-based methodologies are used to support and advance (in most cases to speed up), the urban design process. AI is becoming a key approach for addressing real-world challenges in NBS planning. Within this field, NBS planning uses three main forms: black box, white box, and hybrid planning approaches. The black box approach is particularly valued in NBS urban planning for its efficiency in uncovering hidden patterns and accurately predicting complex nonlinear interactions (Stojković et al., 2017). In contrast, the white box approach, notably case-based reasoning, utilizes previous experience as a foundation for its intuitive methodology (Shiu et al., 2004). This approach seeks out patterns similar to past situations and applies these insights to new NBS projects. Meanwhile, the hybrid approach integrates case-based reasoning with other AI techniques, thereby facilitating NBS planning on larger urban scales. These models are effective for monitoring NBS performance through ground-based measurements and remote sensing (Kumar et al., 2021) as well as for real-time control of NBS by adjusting them to measured and forecasted data (Brasil et al., 2021).

Fig. 1
figure 1

Illustration on linking AI and NBS in urban flood prediction (Horizon EU 101,159,480 ARTIFACT project framework; authors archive)

Incorporating NBS is essential for long-term urban planning, especially under the evolving climate change conditions. Urban biodiversity within NBS is particularly effective in reducing flood risks by diminishing surface runoff and peak flow levels (Wübbelmann et al., 2023). AI offers valuable tools to assist decision-makers in responding to urban flooding and in strategizing for the long-term. While AI has commonly replaced theory-driven and physically-based models in certain areas, its deployment in rapid urban flood prediction (e.g., Fig. 1) is still emerging and developing (Leitao et al., 2018; Berkhahn et al., 2019). This integration of AI in urban planning signifies a pivotal advancement in tackling environmental challenges more effectively. The few existing AI flood models are based on a simple black-box approach and therefore lack generalization (Bentivoglio et al., 2021). However, recent advancements have been made with several AI methods being introduced for rapid urban flood prediction (Burrichter et al., 2023), including a site-based model using Long-Short Term Memory (LSTM), a 1-D model employing Graph-Neural Networks (GNN), and a 2-D model utilizing Generative Adversarial Networks (GAN). A significant challenge with these models is the lack of simulated data, often necessitating a comprehensive output from physically based hydrodynamic models for accurate 1-D and 2-D urban flood predictions. Additionally, integrating new AI techniques into flood models presents difficulties for both water resource engineers and AI experts, highlighting the need for a joint effort across these disciplines.

The variability associated with climate projections poses a significant challenge to the planning of NBS (Tian et al., 2023). The climate variability that is transmitted through regional climate models, statistical postprocessing tools, and the hydrological models amplify the overall uncertainty in the projected water movement over the landscape, which is crucial for the design of NBS (Stojkovic et al., 2020). Addressing these climate uncertainties necessitates an alternative methodology, which involves creating a deterministic-stochastic approach (Stojković et al., 2017; Stojkovic et al., 2020). This model should incorporate both long-term climate variability and climate change scenarios to accurately assess the impact of climate change on hydrological responses. Furthermore, this challenge is compounded by the necessity for accurate predictions regarding biological responses, which are crucial for the effective conservation of biodiversity in urban areas (Urban et al., 2016). The inherently complex and unpredictable nature of urban ecosystems necessitates the exploration and implementation of innovative methodologies to enhance current urban planning practices (Mannucci et al., 2023). Furthermore, the increasing disparity between the sophistication of predictive models and the available data required to forecast the most adverse effects of climate change on terrestrial life is a substantial concern (Urban et al., 2016). Employing a dynamic adaptation strategy that responds to real-time shifts in system variables and identifies the appropriate time for transitioning between planning strategies becomes crucial (Tian et al., 2023). Consequently, it is necessary to develop various decision-making strategies under deep uncertainty to assess the effectiveness of NBS to adapt to uncertainties arising from future climate variability (Babovic et al., 2018).

Discussion and the way forward

From the presented literature, it is clear that convenience-based or ‘opportunistic’ approach (Kuller et al., 2021) to the design of mostly human-centric NBS is currently dominant in the urban planning, where peoples’ benefits are considered over biodiversity outcomes (Ignatieva et al., 2023). Cases where we aim at larger biodiversity outcomes in our cities, and follow-ups to assess the success or failure of the project are sparse. Initial results on how regional connectivity can be assessed and created through targeted NBS strategies that leverage key environmental variables for habitat creation, stepping stones and promoting species movement across large extents, were demonstrated in Donati et al. (2022). However, achieving such outcomes may also require larger-scale and long-term transformations of the urban fabric, where gains may only be seen far into the future. While the economic justifications for such biodiversity projects exist (Johnson et al., 2019; Locatelli et al., 2020), and there are diverse approaches and methods for biodiversity assessment (Tischendorf et al., 2000; Moilanen et al., 2002), it seems that human-centric application is still the dominant form of NBS design in urban environments due to human-led project perspective, i.e., there is still some bias towards solving human issues first. To address the underrepresentation of ecological aspects in NBS we need an unbiased (or less biased) approach. This is where integration of machine learning or AI could be used to enhance the NBS urban planning process and balance the field between human and ecological benefits (see Fig. 2).

Fig. 2
figure 2

Design pathway for implementation of AI in balanced NBS design between human- and eco-centric priorities

Firstly, current modelling tools for urban planning are often too slow for real-time planning and adaptive changes, due to computationally and time restrictive process-based methods (Prodanovic et al., 2022b). AI approaches have already been utilised to speed up flood extent predictions with Convolutional Neural Networks (CNN) (Leitao et al., 2018) with new research showing promise in the application of predictive AI for fast water pollution modelling (Zhang et al., 2019). With more AI-based methods and tools for optimisation of specific NBS benefits (e.g., water management, climate management, economic evaluation, etc.), there is a potential for integration of all these AI-based models under a single framework or tool designed for urban NBS exploratory planning and allowing for co-creation approaches to gain dominance. Fast NBS planning would allow the broader public to be engaged in the decision-making process, which would bring together a critical mass of varied interests, cultures, and expertise, allowing for a more ‘fair’ design process (Naserisafavi et al., 2022).

Secondly, for balanced NBS design, AI can be utilised to enhance the communication between communities, professionals, urban planners, and decision-makers (Hawken et al., 2021). Due to the involvement of many stakeholders with differing objectives, communication breakdown (miscommunication or misunderstanding) could occur between any of above-mentioned levels, which can cause poor or unbalanced NBS design. On the most fundamental level, Natural Language Processing (NLP) models (e.g., Large Language Models (LLMs) can be utilised to create unified NBS project criteria, assignments and documentation which would prevent misunderstandings in the initial project setup phase. While general data and knowledge bases are used to ‘train’ AI models, having a dedicated LLM for technical NBS criteria would be a useful next step. LLMs could also be utilised as a part of NBS planning tools for unifying set criteria for balancing NBS benefits across different stakeholders (both human and non-human). It could potentially overcome not only language barriers of experts across the globe but also interdisciplinary nuances associated with the term NBS and often haphazardly related terms (e.g., Sponge City, Water Sensitive Urban Design (WSUD), Low Impact Development (LID) that share similarities and are found across different scientific disciplines such as green space planning, urban ecology and stormwater management (Matsler et al., 2021). With the rise of visualisation methods, LLMs present a unique opportunity to increase communication pathways through better NBS planning visualisation, and the generation of new ideas. The emerging field of ‘Generative AI’, encompassing LLMs, generative adversarial networks (GANs), variational autoencoders (VAEs) and diffusion models can also yield photo-realistic representations of the urban environment that can be changed through preselected prompts to the model (e.g., Wijnands et al., 2019; Seneviratne et al., 2022). This can enable smoother communication throughout the planning process (Wijnands et al., 2019), but also allow communication of results to the broader public, aiming at informing and educating. Hence, NLP as a broad AI category has the potential to become a multi-faceted approach for equalising knowledge, opportunity, and access to urban NBS design, benefiting lacking biodiversity aspects, and bridging a major challenge in current siloed NBS planning, i.e., lack of communication.

Finally, in addition to enhanced communication, Generative AI has the potential to inform about the results of NBS projects. Most NBS projects are very active in the planning and construction phases, still the monitoring and maintenance phase tends to be less pronounced (Hawken et al., 2021; Prodanovic et al., 2022a). However, it is the post-construction phase that demonstrates the success or failure of the NBS project (Hawken et al., 2021) and that allows the communities to be directly involved in the use of the NBS asset. Currently, there has been a marginal success in maintaining NBS projects through a post-construction phase, usually through monitoring technical parameters (e.g., (Beryani et al., 2021), while biodiversity outcomes have been scarce. The biggest barrier to this has been consistent funding after the project implementation (Hawken et al., 2021). However, Generative AI could generate visual or text-based monitoring states of the project NBS through little or no guided inputs, and (almost) free of charge. This could further be enhanced with real-time, low-cost sensor technology, as well as community reporting or social media campaigns, which would feed the Generative AI models for more up-to-date (or real-time) generation. In this way, biodiversity, water quality, or air quality improvements, could be presented more easily to all stakeholders, providing a comprehensive understanding of what the NBS is achieving, and what negative effects it is preventing (e.g., urban flooding, drought, heat wave, etc.). This application of Generative AI in this context is in its infancy with some early work conducted by Wijnands et al. (2019). Seneviratne et al. (2022) already demonstrated how image generation models in their base-trained state can generate convincing and representative urban planning concepts. Nevertheless, more investigation is needed, not only into the technology but also its responsible use (Sætra, 2023). We see this potential being explored, both in countries with long-standing experience in designing with nature such as Australia and the United States of America, and in countries that are actively embracing the NBS practice such as Switzerland, where biodiversity, heat mitigation, and sustainable urban drainage management (BAFU/ARE, 2022) are central targets for broader NBS adoption and where anecdotal evidence already points to active experimentation in practice with AI image generation for visualizing potential future NBS design outcomes.

It is worth briefly discussing the limitation and scepticism around AI application. AI is often limited by the information provided for its learning. While AI can adapt and create some unique features (as seen in Generative AI models), this too is based on the styles of previous teachings. Nevertheless, databases used for training of these models are vast and ever-expanding. Even though AI is not expected to reach human-level of creativity in the near future, there is an ever-growing number of uses that the application of AI can contribute in NBS urban planning, and as discussed above, to the benefit of urban biodiversity. While AI could provide a balanced approach between human and eco-centric NBS design, the decision-making that is utilised in AI is largely context-independent and purely based on the statistical nature of its training data (Sætra, 2023). The latter can sometimes be considered a limitation as the human operator cannot uncover underlying logical explanations. Unsupervised learning of AI is the reason for its speed and accuracy, but physically-based phenomena are not always represented well, hence new directions in physics-informed AI are gaining popularity (Berkhahn et al., 2019). However, it is arguable if this approach is beneficial for biodiversity outcomes in NBS planning due to the randomness in the living patterns exhibited by most living organisms. While this randomness is challenging for traditional process-based models, and there are new connectivity models such as Circuit Theory (Dickson et al., 2019) that attempt to simulate it, the AI has the potential to better encapsulated the randomness of living things due to its deep neural network structure.

Conclusions

This work focused on current and future directions in urban NBS human-centric and biodiversity-centric design, highlighting the potential of rapidly advancing AI practices for more balanced outcomes. The work showed a current inkling towards human-centric design, with better technical aspects and economic justification for such an approach. This approach still struggles cultural and historical considerations in NBS design, it often doesn’t facilitate co-creation with multiple stakeholders (including decision-makers) and is often opportunistic rather than strategic. On the other hand, eco-centric design (non-human design) is less pronounced and highly segmented between different methods. With less pronounced economic justification and follow-up monitoring, it requires more promotion to become more competitive to human-centric design. With the AI rise in modern urban planning and engineering science, multiple machine learning models show promise in utilisation for more balanced (between human and non-human) NBS design. With current technical AI solutions for climate change, water quality and quantity prediction, rapid visualisation, and language processing, there are three distinct applications where AI could be utilised to enhance NBS urban planning: (1) embedded into urban planning tools for fast exploratory modelling, (2) as an effective tool for better communication between stakeholders and across different project phases, and (3) as a dissemination and project results presentation tool for better project outcomes visualisation. Further research is needed in all three of these areas, however, some preliminary studies show potential. The uncertainty about AI decision-making could present an obstacle, however, given the randomness of biodiversity phenomena (both human and non-human), AI-driven stochasticity could be able to provide even better estimate compared to traditional process-based models.