Abstract

It is the best of data times and it is the worst of data times. Although we have an unprecedented breadth of data types available at previously unimaginable granularity and scale, data shortcomings abound (e.g. ethics, access, representativeness, or in terms of validation). In some ways, more data only means more weakness to overcome. A skim of several domain journals – this one being no exception – certainly confirms increasing interest in newly available data sources. The era of Big Data (e.g. Kitchin, 2014) has come and gone, superseded first by data science and most recently artificial intelligence (AI). However, the attraction of novel ways of measuring cities remains, even though these data do not seem to get us much closer to resolving the fundamental challenges that have puzzled urban researchers for decades. Why is this? In our opinion, part of the reason why challenges in urban analytics and city science have not been successfully addressed by new data is that we have not paused to consider exactly how they could do so. We, the authors, have been thinking a lot lately about the challenge of one particular data source – satellite imagery – and how to make it more ‘useful, usable, and used’ for urban and social research and policy (Imago, 2026). Based on this experience, we have a few ideas about how our data fall short, how imagery can help (both as data source and exemplar), and how to help create the data ecosystem we need.
What do we mean by new urban data and what would desirable urban data look like? For one thing, it is important to acknowledge that the current attraction of ‘emerging’ data types such as mobile phone traces, text from social media and the web, 1 eye-level images, sensors and, of course, satellite imagery data is only the most recent iteration of a longstanding, and possibly unavoidable, fascination with novel ways of measuring and collecting information about the world. After all, there was a time when tabulating census data for neighbourhood units was shiny, new, and help open pathbreaking possibilities for urban research. Similarly, transport surveys and time use diaries, now old hat when it comes to understanding urban mobility behaviours, were once state-of-the-art urban data. Indeed, where satellites are concerned, a few sub-disciplines looking at cities have long devoted attention to this source. For them satellite imagery is mainstream (e.g. land use and environmental fields). We believe satellite imagery has so much more potential to offer to an even broader constituency of urban researchers. Satellite technology has also radically changed and advanced in the last few years, to the point that, although some satellite missions have been in orbit for over half a century, we think it is more helpful to frame it as part of the new wave of urban data.
There are several reasons why novel urban data in any era inevitably captures the imagination. One, more cynical, is that it is novelty, not necessarily utility, that draws attention and enables publication. A more generous perspective is that cities are constantly evolving and mechanisms for capturing those changes must also adapt. There is also the undeniable fact that existing data never completely serve the purpose; there is always some key element that is lacking, forcing us, as researchers, to continue the hunt for a more perfect urban data source. 2 There are several characteristics the ‘more perfect’ urban data source should ideally possess. The most basic and obvious is utility: data need to measure the things we want to measure at the requisite temporal and spatial scales, whether place or individual characteristics. Coverage matters too. Ideally data provide information for all people and spatial units, or at least a representative sample. In addition, we count on data to be reliable, with provenance and processing we can evaluate and trust. And, naturally, data must be accessible. This means a level playing field for acquiring, paying for, and working with data. Underpinning all these desired data characteristics are ethics, including consent and privacy.
The current landscape of urban data is far from ideal. Whether place characteristics or individual behaviours, data often do not directly measure what we need (for example, what kinds of people are travelling to what sorts of places). Our data also often lack the temporal and spatial resolution required to understand processes and behaviours that operate at fine scales, such as hour or day, land parcel or street segment, hindering understanding of, for example, urban heat vulnerability and adaptation. Perhaps most importantly, much of the data used in urban research is privately owned, with disparate access privileges that can be easily revoked, high costs, and opaque inputs and production algorithms. If we think of our data as infrastructure, cracks and weaknesses emerge where data integration, privacy, and reliability are concerned. Lastly, and just as important, the emphasis placed on novel data distracts attention and resource from traditional data sources, even though most newer forms of data are still reliant on old data for validation, benchmarking, and context. We should attend to the considerable opportunity costs and tradeoffs.
Among novel urban data types, satellites offer advantages that no other data can. They provide an always-on observatory that records the Earth consistently over space, time, and quality. By their nature, satellite sources offer global coverage in almost all relevant cases with similar cadence, a comparable sensor, and at the same resolution. All those dimensions vary across types of satellites: some offer higher resolution, some offer higher refresh rates, some offer more variety of data. The overall picture is consistent: compared to almost any other source of data, it is hard to match satellite reliability, coverage, utility, and accessibility. How do these traits fit satellite imagery within the ‘mosaic of urban data’? Imagery can be combined with other sources that provide denser representations in sparser areas to ‘stretch’ them and extract more value. We can also blend them with deep but slow-to-update data about cities to build inter-release updates that track otherwise invisible changes. Satellite imagery can contribute to fused data where the sum is greater than the parts.
Plus, there is everything else that has happened over recent decades that has made satellites an even more valuable and powerful source of urban insight. Cheap storage and compute; powerful algorithms, from the deep learning revolution of the early 2010s (LeCunn et al., 2015) to the recent geospatial foundation models (GeoFMs, Janowicz et al., 2025) that bring state-of-the-art approaches to geographic data; and the wide democratisation of access to these advances that open source has brought in (Rey, 2018). The confluence of these three trends, although superficially detached from satellites, has rendered them a new proposition for urban researchers. It is also important to highlight that satellites, as a term, encompass a rather large and diverse set of technologies and associated data sources, many of them comparatively novel. When we refer to satellites in this context, we are thinking of a broad and vibrant ecosystem that involves both government and industry led initiatives, including very old (e.g. the first Landsat mission in 1972) as well as more recent than this editorial, and capturing visible ‘colors’ (optical), many more invisible wavelengths (hyperspectral), plus temperature (thermal), and more. This is to say, readers who have not recently considered the satellite imagery possibilities are in for a (pleasant) surprise.
If there is so much potential, what is holding satellite imagery back in city research? We can think of at least three key reasons. First, satellite data are big. The notion of ‘big’ is constantly being redefined by the tension between the size of the data and the power of the computer to process them. By any common metric, however, the former outpaces the latter in this case; to the extent, it is an important factor that steers many research decisions. Second, satellite data are hard. Most urban data are non-trivial to understand. How sampling is set in a survey is constrains its downstream applications. How geographies are encoded in coordinates or geometries determines what one can and cannot do with the data. But satellite data rarely, if ever, record directly the phenomena of relevance for urban researchers. Instead, most sensors capture wavelengths of light with known physical relations to some outcomes of interest (e.g. vegetation). Establishing the mapping between the two in useful ways thus requires an additional degree of understanding and cognitive load. Third, satellite data are different. This, to us, is one of the reasons why satellites are much less common in urban studies than other disciplines such as climate science. Urban researchers are typically exposed to a fair amount of quantitative training, but virtually all of it is based on tabular data (networks being a common exception). Images are typically stored in data structures (e.g. multi-dimensional arrays) that are fundamentally different from tables, where every row is an observation and every column is an attribute. Thus, the techniques to work with these data are also different from those traditionally taught in urban curricula. This can make satellite data feel intractable, difficult, and foreign to researchers interested in substantive urban questions.
Zooming back out, for us, a big question is how we collectively go from the current state of playful data experimentation to one where we focus our play on the questions we need answers for. This is a large topic for a small editorial, so we will limit ourselves to sketching out a few general directions. For one thing, our view is that there is no such thing as the one dataset to rule them all – not even satellites! No single data source will meet all the requirements we have outlined above. What is more, we seriously doubt whether one ever will. Instead, what we propose is question-driven, tasteful triangulation of varying subsets of everything that is available (paired with advocacy for creating the data we need). Every source has its advantages and also its weak spots. One of the core roles of urban data scientists and quantitative urbanists alike going forward is not working with data (machines increasingly do that) but instead curating data for each project, each question, and each scientific exploration. Each challenge will require a different recipe of needs, and it is our role to find the data ingredients that will address it. To make sure we have as many, and as prepared as possible, cooks, we also need to take proactive action. We need to rethink curricula so that the researchers of the future feel native when working with all these new forms of data. And, in the meantime, we need to build bridges that feel like on-ramps to accessing and using these sources. Only by bringing the data to interested researchers in ways they can engage with will we fully realise their potential (and change the world for better).
What does this all mean for readers and authors at Environment and Planning B? We hope this editorial is read in ‘the right way’. Not as a cheap critique of the world as it is, but as a hopeful aspiration to the one we would like. We recognise the promise of the new wave of urban data, with satellites being a particularly exciting one. We also understand the challenges of delivering the vision we have laid out. Above all, we are acutely aware of the lack of easy solutions. For that reason, we would like to end this editorial as the start of a broader, longer, and more sustained conversation. One that takes place in venues of actual conversation (conferences, workshops, live events), but also in the pages of journals like this one.
Footnotes
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the UK Research and Innovation (UKRI) through funding of the SDR UK Imagery Data Service for Sustainability, Prosperity and Wellbeing (IMAGO) project [grant number ES/Z504208/1], a consortium with teams at the University of Liverpool, Newcastle University, Harvard University, and University of Manchester.
