2022
Environment
Challenge provided by Bristol City Council

Identification of dark ecological corridors

By implementing these solutions, Bristol City Council could increase bat activity by 10% while saving about 185 tons of emitted CO2.

Artificial Light at Night (ALAN) is increasingly recognized as a major threat to global biodiversity [1]. It alters the amount, quality, and connectivity of available habitats for species. Light pollution causes habitat fragmentation by making areas harder to pass through and thus creating spatial barriers, evolutionary changes, and distorting normal growth patterns, among other negative effects. Lighting affects species differently depending on when their feeding and mating seasons occur. 

One of the most affected species is horseshoe bats [2], as they are sensitive to light, and ALAN can greatly reduce their feeding area and activity. As a solution, dark ecological corridors and spaces can be created inside the city so bats can thrive and move around different areas. Public lighting systems today enable temporarily activating elements of the dark ecological network. The Bristol City Council (BCC) plans to install a Central Management System (CMS) within the next two years to allow dimming regimes to be implemented in certain locations.

Goal

The goal of this challenge was to reduce the impact of ALAN on the bats in the city of Bristol and therefore reduce its adverse effect on the ecosystem. To accomplish this, the teams needed to create an optimization algorithm connecting the bats' natural habitats (feeding, roosting, and breeding sites). 

United Nations SDG 

GOAL 15: Life on Land

  • Target 15.5: Protect biodiversity and natural habitats

Datasets
  • Records of recently spotted bats (within the last 10 years) with 1km resolution for security purposes. Bat records older than 10 years were provided at full resolution, provided by the Bristol Regional Environmental Records Center (BRERC).
  • Data on the occurrence of moths (a major food source for bats) in Bristol, provided by BRERC.
  • West of England Habitat GIS Map with Priority Habitats, potential Priority habitats and other habitats, provided by BRERC.
  • Green Alleys GIS, provided by BRERC.
  • Wildlife Corridors dataset containing sites that help link up and buffer Sites of Nature Conservation Interest (SNCIs) and the City Green Belt, provided by BCC.
  • Ordinance survey open data green space, provided by BCC.
  • Public lighting data – location of street lights and type of lights, provided by BCC.

Data

Besides the data provided, most teams complemented the datasets with information on POIs and transportation services from OpenStreetMaps. Some datasets from the Bristol Open Data Portal, such as traffic accidents and criminality rates per ward, were also used. This data helped to understand where lightning might be more important compared to other places.

One team used satellite information provided by the NASA Visible Infrared Imaging Radiometer Suite that quantifies the amount of light reflected from Earth. Using this dataset, it was possible to understand the areas where it might not be possible to turn off the lights (e.g., billboards). 

Regarding the data provided, most teams pointed to the quality of the observational datasets. While the data goes many years back, the observations were not consistent; for example, some years had only one datapoint measured. Although it made yearly modeling possible, it posed a challenge for seasonal modeling. Another issue found with the observational data was that categories such as  “several”, “present”, and “abundant” had no numerical definition. It was also pointed out that making the crime dataset (available on the city's open data portal)  more detailed regarding the location would help the model include more safety parameters. 

Methods and Techniques

The teams divided the map into a grid (most used hexagons). Some teams started by using an HDBSCAN and X-means for clustering areas of bats or prays that needed to be connected. One team also clustered together the location of bats and prays and applied the Lotka-Volterra equations to describe the dynamics of each cluster. 

Some teams used Dijkstra's algorithm to identify the dark corridors to find the shortest path considering the number of bats, moths, street lights, and penalties for switching the lights off. One team, on top the Dijkstra's algorithm, used Markov chain processes on graphs to mimic the bat’s behavior and an Agent-Based Model to achieve an overall optimal street lighting configuration.

Another approach was to use a genetic algorithm to identify the dark ecological corridors.

Main Insights from Data

It was noticed that the number of horseshoe bats in the last decade has been decreasing in the city of Bristol, therefore highlighting the importance of this work. One team also observed that horseshoe bats could not be compared to other bats, as they were observed in different locations and also have a different diet and therefore it is important for the techniques to be scalable to other species. 

Regarding seasonality, some teams noticed that the location of the clusters changed during different seasons, and most sightings happened during summer, which is the breeding season. The change in the clusters across seasons can be seen in Figure 1.

Figure 1 - Observation of bats in the city of Bristol in different seasons.

Product

Most of the teams proposed a tool to assist the city officials in finding the location where the lights should be turned off. The proposed main activity was to create suggestions on where to turn off the lights to form a dark ecological corridor through the input of different species and locations. Other features presented by the teams were predicting the change in the population, the output of seasonal corridors, the input of new restrictions, and the parameterization of the different constraints and restrictions. The resulting output would be a map of the ecological dark corridors and a list of lights to be turned off and/or dimmed. An example of a dashboard can be found in Figure 2.

One team suggested creating an informative website for the population of the city explaining the importance of the project and a map with the lights projected to be turned off. On this website, the citizens could also provide feedback to the city council and request to turn on the lights in the case of an emergency. 

Figure 2 - Example of a dashboard with proposed dark corridors and the impact on the lights.

Social Impact

The main outcome identified by the teams was to mitigate the impact of ALAN and increase the city's biodiversity without impacting humans. It was proposed to measure the following impact metrics: 

  • Number of horseshoe bats observed in the city and its surroundings
  • Artificial light pollution by measuring the brightness of the sky 
  • Saved CO2 from shutting down lights
  • Financial savings on electricity from shutting down lights

Besides these, safety metrics should be monitored to ensure that there is no negative impact from the implementation of the solution.

The teams estimated that by implementing this proposal the bat activity could be increased by 10%, saving about 185 tons of emitted CO2 and £25,000 in electricity from shutting down lights. 

References

[1] Kévin Barré, Arthur Vernet, Clémentine Azam, Isabelle Le Viol, Agathe Dumont, Thomas Deana, Stéphane Vincent, Samuel Challéat, Christian Kerbiriou, Landscape composition drives the impacts of artificial light at night on insectivorous bats, Environmental Pollution, Volume 292, Part B, 2022, 118394, ISSN 0269-7491, https://doi.org/10.1016/j.envpol.2021.118394.

[2] Bo Luo, Rong Xu, Yunchun Li, Wenyu Zhou, Weiwei Wang, Huimin Gao, Zhen Wang, Yingchun Deng, Ying Liu, Jiang Feng, Artificial light reduces foraging opportunities in wild least horseshoe bats, Environmental Pollution, Volume 288, 2021, 117765, ISSN 0269-7491, https://doi.org/10.1016/j.envpol.2021.117765.

Top 5 Solutions
Open-source code

More about this category

World Data League - a competition for data scientists
World Data League @Copyright 2022