Aerial Analytics

Introduction

The preceding two articles in this site, demonstrates the utility and process of collecting and extracting aerial imagery for the desired location and time. ESA’s Sentinel Missions have made the process, very straightforward, to retrieve rich cloudless imagery which can be used for a multitude of studies. It uses the power of ML and Deep Learning (U-Net) to extract pixel level insights, translating into real world land use computed in near real time. The process offers tremendous potential for spatial analysts like GIS experts & practitioners, Policymakers, Urban planners, and Architects to understand and map the changing land use patterns over time.

For example, Heathrow Airport, London operates at 99% capacity for over a decade now without addition of infrastructure like terminals and runways. Increase in demand for aviation through increased economic activity (GDP numbers worldwide) exerts spatial pressure on urban infrastructure which possess the most impact on aviation as it being the single largest contiguous entity in any urban settlement. Data driven and intelligence powered analysis using AI and satellite imagery holds promising potential in being able to derive high-frequency advanced and macro level forecast for the context of aviation development and urban congestion.

In this segment, we will be exploring the potential applications of a robust yet flexible workflow for generating useful insights for urban stakeholders. It will offer potential benefits like reduction of time, effort and hence translating into saved cost and energy. More specifically in the context of aviation and urban planning, we use data to answer the following questions:

1. How does land use around airports impact its ability to expand?
2. What’s the distribution of airports in the USA according to its states?
3. How expandable are airports in the USA?

To answer the above questions, we look at the following variables in this study.

• Population
• Passenger Footfall
• Land use composition
• Percentage Built-up area
• Percentage developable land
• Percentage water

Firstly, we aim to understand the macro scale land use around the airport’s infrastructural fabric using our semi-automated workflow described in the previous section. To achieve this, we focus on assigning land use classes to our dataset of 1000 airports, which will be specifically tested against the airports in the United States. The reason for choosing US as our baseline is, that from the economic standpoint, US is one of the matured countries in the world in terms of Population and GDP, with its aviation network starting to saturate.

From the previous articles, we have seen pixel level expansive calculations on measuring the land “developability” around airports by direction. This involves measuring the predictions from the ML which were classified as “urban” (red) to be built-up and “vegetative” (green) to be developable. As per the above figure, it can be clearly seen that the top 9 most crowded airports in the USA (from top left: Atlanta, Los Angeles, O’Hare, Dallas, Denver, John F. Kennedy, San Francisco, Seattle, and Las Vegas) sorted decreasing according to their passenger footfall numbers in 2018 along with their levels of buildable land around the airport. It is to be noted that water was not considered as an easy area to develop/expand airport due to the cost involved in land reclamation as well as surrounding geography. The level of expandability shows the level of saturation of American airports, with majority of the top city airports having little to no available land to expand, either in the form of a new terminal building, runway or supporting services (cargo, amenities etc).

Airports being the single contiguous and large entity in any urbanscape, as seen from above, is constantly at pressure due to a range of factors notably aviation demand, passenger traffic, economics, and politics. Moreover, for growing economies and in democratic systems, it's always a pressing question and a challenge for urban stakeholders to choose between expanding the current airport or going for greenfield development. For the former, it is highly crucial that the surrounding geography and economics at the airport plays an active role in answering the question on whether it is feasible (whether it is possible, in many cases) for the airport to support such expansion.

Secondly, for our detailed study, it is significant to understand the distribution of airports by states for the US. From the larger world dataset, we identified around 375 airports in the US currently in operation. Initial inferences like economic activity correlating with aviation can be easily drawn from the above insight looking at California, Washington, New York and Texas being thriving hubs of economic activity. Although this doesn’t offer conclusive evidence, it is sufficient to make a specific observation of the airports that could potentially be in a space crunch.

Based on the predictions from the U-Net trained on the world airport images dataset, a pie chart was constructed to show the distribution of the various land uses around the airport. This was plotted into a map of the USA as seen from the above figure with the charts located at the approximate locations of the airport themselves. Airports of particular interest in states like California and Texas were found to be relatively highly urbanised whereas traces of arid land were seen in the dry states of Arkansas. Moreover, coastal airports were easily distinguishable by the prominence of the blue, especially for the airports located at the shore or in vicinity of the coast. Spatially crowded and congested airports could be easily spotted by the trained eye, for instance, O’Hare, Los Angeles and airports in the state of Texas, which are dominated by red pixels indicating urbanisation.

In order to the posed research questions, we considered additional variables such as passenger footfall, to understand if congested airports are facing a potential “choke” to the city. Passenger footfall numbers for 2018 are plotted against the built-up ratio (logarithmic scales) to derive urbanisation and passenger movement relationship, for every individual airport in the US. First and obvious inference drawn is that with increasing urbanisation, aviation numbers can be expected to see an increase. Additional inferences such as low-urbanised high-congested and high-urbanised low-congested zones are also populated from this plot.

The interesting findings offer notably important insight to city planners to rethink the future of airports that are facing a crunch/pressure. Factors such as increasing economic activity (urbanisation) and therefore increased passenger movement, promises a novel and data-driven approach for macro level city analysis. In order to translate the findings to real world conclusions, 4 different categories are devised based on the 4 zones from the urbanisation vs traffic plot, namely:

• • Build – Airports have potential to grow and are low urbanised
• • Monitor – Airports with high footfall and high urbanisation
• • Remove – Airports have low passenger movement but are spatially congested
• • Leave – Airports have little to no impact either on urbanisation or on passenger footfall

Although inexhaustive, the 4 planning recommendations offer potential value to city stakeholders in being able to plan in terms of aviation and urban infrastructure at the higher level. It is interesting and intriguing to note that over 50% of the airports in the US fall into the “Monitor” zone, which calls for an important course of action, to rethink aviation and urban metrics. Rare, but a few airports are seen to be in the “Remove” zone which may not necessarily mean ceasing operations but thinking on increasing the options for reconfigurability, or merging operations with nearby airports, which will improve efficiency in the aviation network as well as produce positive impact in terms of sustainability.
Overall, the nature of this semi-automated AI based workflow to automatically predict land usage around the airport offers various levels of promising insights to different stakeholders translating data into real world actionable insights. This study also raises important questions pertaining to the context of aviation and urbanism specifically touching areas of sustainability like noise levels, global airport network analysis and the city-airport synergy.

In the upcoming part of this study (Chapter 2), we will be looking at an upgraded technique of retrieving aerial images which is faster and efficient, thus allowing us to extrapolate this study globally to about 1800 airports. Specific questions related to expansive capacity and distance between airport and city’s CBD, changing use cases of airports and aerotropolises will be focused on.

Stay tuned.

If you are interested in this work please get in touch with us.