What to tell using AI and satellite imagery? Here are some ideas
What stories can be told with a combination of satellite imagery and AI? In this second part of this two-part article, you will find some story ideas to put into practice.
[If you'd prefer some tips to get you started instead, click here to read part one]
Many of these stories can be told using satellite data alone, without using AI; the main advantage of algorithms is quantifying phenomena, identifying patterns, showing changes or finding a "needle in a haystack" across large territories or different time periods. Or to automate a process: since satellites produce recurring data, you can build, for example, a platform that automatically detects changes in the size of forests.
I adopt here the framework developed by Paul Bradshaw for data journalism, which recognises eight types of stories: scale, change, ranking, variation, exploration, exploration, relationships, stories about data and stories through data.
Some of the stories I propose follow existing examples; others are paths not yet travelled that are worth exploring. The underlying idea, in any case, is that they will serve as inspiration for journalists interested in exploring these techniques further.
These are stories that allow us to quantify a phenomenon, either to provide context or to understand the scale of a problem. Satellite information is especially useful where there is no official data, which is common in the global south, or when dealing with illegal activities.
These involve a time dimension and tell how something has (or has not) changed. Decades-old satellite data, such as from the Landsat programme, are very powerful for looking at the consequences of climate change. Some current providers, such as Planet, produce daily images of the same location, allowing even very small changes to be monitored.
These are stories that establish an order, revealing which is the best, worst or "different" in a series. Algorithms and computational power on satellite imagery make it possible to compare very large territories, even entire countries. Using the same dataset (the same satellite) provides an extra advantage and overcomes the problems that arise when comparing data collected in different jurisdictions, which are not always standardised.
Show differences in phenomena where, a priori, there should not be any, evidencing inequality. As with ranking histories, always using the same dataset of satellite images allows different areas to be compared using a single criterion.
These are stories that encourage the user to get involved in the exploration of a phenomenon, with or without interactivity. Satellite imagery can be used to build large maps where each reader can explore what is happening near their home.
These pieces explore or analyse the relationship (or lack of relationship) between two different phenomena, with Bradshaw's key caveat: correlation never implies causation. Satellite imagery can be used to account for the two phenomena being compared; or for one of them and use other information (official data already available, for example) to account for the other.
These are stories about the data itself: whether it is missing, inaccurate or flawed. Satellite imagery can be used to verify or question the quality of official datasets; missing or censored imagery is also saying something.
In these stories, data is the means to unveil or discover something. Particularly useful here is the use of AI to find patterns, infrastructure or objects in a large amount of satellite imagery.