Updated: Nov 18, 2021
Hosted 26th May 2021
To watch the recording of the whole event, click here.
We were joined in May by Jasmine Muir. Jasmine’s event was inspired by her work at FrontierSI, and answers a very important question for Earth observation businesses: How do we get people to trust Earth observation data?
Jasmine Muir has worked in Earth observation across government, university and the private sector for close to 20 years. She is passionate about promoting the continued growth of the Earth observation sector, encouraging and training the next generation of scientists and the integration of research into industry. Jasmine currently works at FrontierSI as Earth Observation Technical Lead. In this role she works to combine her passion for end-user engagement, to promote the use and uptake of Earth observation for decision making. Jasmine obtained her PhD from the University of Queensland, with her research focused on the use of terrestrial lidar for the assessment of tree structure diversity as an indicator of landscape condition.
FrontierSI is a leading not-for-profit that aims to connect people in research, industry and government, and help solve problems with spatial information. Having completed market research studies on how Earth observation is and isn’t being used, Jasmine was able to outline that trust in Earth observation data is one of the most important things required to help grow the market for the industry.
Setting the Scene
Currently, the Earth observation sector contributes over AUD$5 billion to the Australian GDP, however there’s huge potential for that number to increase if the market is made aware of what different types of Earth observation data is available. There are lots of satellites planned to be launched at the moment, and the numbers in the sky are expected to grow exponentially in the next few years.
“But there are a lot of new satellites planned to be launched, and hopefully Australia will have its own satellite up there one day soon.” - Jasmine Muir
The graph below shows a huge number of Earth observation launches in the private sector, many which are constellation launches and satellites not limited to traditional multispectral satellites. Thermal satellites can be used to find the temperatures of objects on the ground and fine scale radar satellites and hyperspectral images can also be used to receive different information from traditional satellite images. These differences help collect data for a variety of unrealised uses that may not be currently available with our current technology.
(Launched vs. Planned satellite launches in the government and private sector, Source: Jasmine Muir)
What is Trust?
With these new innovations in Earth observation technology, customers need to know that the data they are receiving are of good quality. For Earth observation data to become widely used, users need to have trust in that data – but for different users and use cases, the required level of trust can vary.
Jasmine drew a very relevant comparison to users driving their car. Consumers use their car every day, and trust that it will work and be safe without knowing all the details about how a car functions. This is because they’ve either used it before, have seen others use it before, and have an understanding that it is designed to adequately fulfil its job.
So how do we go about building this trust – especially for something consumers may not understand like EO data?
The definition of trust is the belief in the reliability, accuracy or ability of something. When we think of Earth observation data, trust fundamentally lies in the belief that the Earth observation is reliable and will output the data and results customers are looking for. If this trust is absent, users are unlikely to use the data to its fullest extent, leading to a decrease in productivity.
Making sure Earth observation data is reliable is all based on the end user’s needs. Businesses must make sure that it is perceived to be useful, easy to use and accessible to users with no prior knowledge to effectively be converted to actual system use.
How Is Trust Created for Satellite Data?
Trust for earth observation satellite data can be created through formal calibration and validation, which are well documented and can be carried out through insight infrastructure and specific validation campaigns for assessing the accuracy of the product. Another way would be through its performance, with customers gaining an understanding of how the data will work for them, either through formal performance indicators or word of mouth often in the form of a user story. Finally, industry engagement is important to help customers understand how Earth observation data can be incorporated into strategies to solve problems. Jasmine noted it is important to use a push & pull market strategy, where ‘push’ encourages uptake growth in users already in the sector through methods such as product development, while ‘pull’ educates users who may not already be involved to build the product that is required.
Calibration and Validation
Calibration and validation happen across the lifecycle of a satellite. Calibration is crucial as sensors must be thoroughly tested before launch as there is no chance to modify the sensors after launch. There are different types of calibration, such as geometric positioning by lining up features on the ground to make sure they are in the correct position, and radiometric calibration to make sure colours are true values (or calibrated to the true values) of what is reflected on the ground. Validation defines the accuracy of the product retrieved from satellite Earth observation. This is further outlined in the AusCalVal report here.
Several important considerations are required for selecting calibration and validation requirements. This includes the users interacting with the product, who will have different experiences and thus require different types of communication to gain trust. Other important factors are the interactions of user maturity, use cases, satellite characteristics and the lifecycle phase.
(Earth Observation User Maturity Levels table from Jasmine’s event, Source: FrontierSI, read more here.)
The Cal/Val Data Decision Matrix provides a great guideline for assisting the calibration and validation process. For example, the type of acquisition changes the calibration and validation process; continuous satellites give more opportunity for calibration and validation while tasking does not, and requires specialised field campaigns to collect appropriate data. Jasmine went into further detail to consider different end-uses, and the differences in those situations.
(Cal/Val Data Decision Matrix from Spiral Blue event, Source: Jasmine Muir)
Barriers to Trust
While there are many ways outlined to increase the trust in Earth observation data, Jasmine also noted it is important to keep in mind the barriers that can erode trust placed in the data.
There is a common assumption that satellite data is objective, trustworthy and not influenced by humans. Mislabelling can create users’ over expectations and can also cause products to be oversold. This can happen if the accuracy or method is not reported or assessed appropriately, or there is omission of information. Ideally, satellite data is accurate and precise, but often it can fall short in one or both areas. These problems can allow businesses to improve the product if they are appropriately addressed and communicated.
Miscommunication is also a prevalent issue with satellite data, especially with what Jasmine called a symbology problem. Very commonly, quantitative symbology with colour ramps can affect how people perceive and potentially misinterpret data. The image below is also an example that can spread misinformation across social media, where symbols ineffectively used and inaccurately portrayed how much of Australia was experiencing bushfires.
(Map of 2019 Australian bushfires, Source: Time Magazine)
In a section Jasmine titled “Garbage In, Garbage Out”, she highlights that good training data is crucial to obtaining a good model. In these situations, experienced operators are essential in obtaining the correct data. For example, in her own experience with crop data collection, inexperienced field operators collecting data on the ground misidentified barley and wheat. Because the ground operators were unable to differentiate between the two, the crops were indirectly classified in the data. Working with agronomists in this case would have avoided this situation and resulted in usable and accurate output data.
Finally, ‘deep faking’ satellite imagery and photos can now easily be achieved with modern technology, whether that is creating new satellite images or altering existing photos. This can significantly erode trust for users and the public if used to create hoaxes, and can discredit stories that are based off or reliant on real satellite data and imagery.
Trust in satellite data is crucial to take into consideration for the Earth observation industry, especially for users who need full confidence in the data to make important decisions or financial evaluations. It is also important to keep in mind that in other cases, a high level of trust is not as necessary and less accurate data can be tolerated.
Jasmine closed on this note: It is best to adopt a “Zero Trust Architecture” – assume the data cannot be trusted, before it proves itself for use.
About Spiral Blue
Spiral Blue is a Sydney SME focused on building the next generation of Earth observation services with artificial intelligence and Space Edge Computing. Spiral Blue technology has applications in defence, city planning, utilities, and other industries. Founded in 2018, the company has recently launched its first Space Edge Zero prototypes to orbit, and is now awaiting results of this in-orbit demonstration.