How Earth observation and AI are opening up space data
By Ioanna Lykiardopoulou
For decades, Earth observation was largely removed from everyday decision-making. Satellite imagery and space data were mainly used by governments, research institutes, and specialised agencies, with limited reach beyond expert communities.
Today, that is changing. Satellite-derived insights are increasingly used in practical contexts, from agriculture and infrastructure planning to air-quality modelling and urban management – often by people with no background in space or geospatial analysis.
This shift reflects the broader evolution of the new space economy, where space data is becoming accessible to a wider range of users and applications. To explore how this transition is playing out in practice, we spoke with:
Parya Pasha Zadeh, Program Manager at NL Space Campus, Project Officer EO Science at the University of Twente, and Director and co-founder of Marmoris.
Thijs Oosterhuis, Solutions Architect for Geo-Information and Remote Sensing at Provincie Zuid-Holland.
Opening Earth observation: data, access, and AI
The shift towards wider access to space data and Earth observation (EO) applications results from three main changes over the past decade, according to Pasha Zadeh.
The first one is the availability of public space data. Especially under Copernicus, the EU’s Earth observation programme, Sentinel satellites operated by the European Space Agency began providing systematic satellite data from the mid-2010s onwards, making large volumes of EO imagery openly and reliably available.
The second change is the rise of commercial satellite operators – alongside the boom of satellite constellation deployment – that not only provide frequent captures of EO data, but also offer higher resolution images.
From automation to insight
The third key shift is AI, Pasha Zadeh adds. While machine learning began to be adopted at scale in Earth observation in the early 2000s, recent advances in AI are significantly expanding the range of possibilities, such as enabling more data-driven insights.
“Take spatial multicriteria analysis for site selection as part of a conservation project,” she says. “Traditionally, you would use different factors (like time series EO data and end user perspective), combine them as criteria in a GIS decision-support tool, and come up with a site-selection recommendation.”
“What we are now exploring is using AI to learn through the different layers of data which factors are more important, and weigh those criteria to make a decision.”
In a spatial planning context, this could mean combining different data sources over time – like air quality, land use, demographics, or income – to reveal trends such as increasing housing demand that can help explore where to build new houses, adds Thijs Oosterhuis, Solutions Architect Geo at Provincie Zuid-Holland.
“By adding AI to the mix, you can potentially explore scenarios from the opposite perspective. For example, how building new housing or an industrial area in a certain location might affect water quality, land use, or socio-economic patterns over time. The possibilities of modelling our surroundings can become endless,” he emphasizes.
Another, already visible, benefit of modern AI is the higher speed of satellite data processing. Satellite imagery is extremely data-heavy, with each pixel containing tens or even hundreds of measured values, and much of it obscured by cloud cover.
“Earlier, we used to receive all the data and did the pre-processing on the ground – cloud masking, removing water bodies, and then analysing only the usable parts,” Pasha Zadeh explains. “Now, newer satellites use on-board AI and edge computing to pre-process the data and send only what’s relevant, which saves a significant amount of storage and bandwidth.”
Bridging gaps and lowering barriers
AI could also help bridge gaps in Earth observation data, particularly when satellite imagery is captured at different spatial and temporal resolutions, adds Oosterhuis.
“Sentinel data, for example, is available at around 10-metre resolution with regular coverage every few days,” he explains. “Data from commercial satellites can be available at three metres with near-daily coverage, or at sub-metre resolution but only a few times a year and at less predictable moments. That means you don’t always have the same level of detail at every point in time.”
“Helping bridge some of the temporal gaps could be a promising added function for AI,” he says.
Both Pasha Zadeh and Oosterhuis see emerging opportunities in using LLMs to allow end users to interact with space data directly. “If we can make a common user interact with spatial data in a conversational way, we're opening more doors for the uptake of this type of information,” Pasha Zadeh explains.
Earth observation markets and applications
The Earth observation market has traditionally been dominated by public-sector users, but this is gradually changing. While governments and public institutions remain the largest consumers of EO data, a broader set of application domains and emerging markets is now taking shape, according to both Pasha Zadeh and Oosterhuis.
One of the most established areas for EO is, naturally, climate monitoring and prediction, as Pasha Zadeh explains, where the focus is less on fine spatial detail and more on consistent, long-term measurements. Closely linked is also environmental monitoring, including biodiversity and ecosystem services.
“There are several AI-powered EO products that can be used for detecting deforestation, monitoring land-use and land-cover change, identifying habitat loss, and detecting pollution,” she says. Alongside these uses, EO also underpins long-standing application areas such as disaster risk management and food security.
Novel applications in environmental impact assessment
Pasha Zadeh’s startup, Marmoris, specializes in environmental impact assessments for coastal and aquatic habitats. As part of a recent project, Pasha Zadeh is exploring a new EO application: using satellite data to have an indication of biodiversity in water – for inland water bodies, the coast, or ports.
“We don’t know if this is possible and we’re currently in the research phase,” she says. The approach combines satellite-based observations with eDNA samples, water quality indicators, and different environmental parameters.
If successful, Pasha Zadeh argues, this could save time and money, enable biodiversity monitoring to scale, and provide early warnings of threats. Potential use cases include conservation and nature-based projects, as well as identifying damage from activities such as construction or tourism.
EO in policy and regional governance
Oosterhuis’ work illustrates how EO is used in practice at the regional policy level. As a solutions architect for geo-information and remote sensing at Provincie Zuid-Holland, his focus is on bringing together many different datasets. “The main goal is to combine a lot of different data sets,” he explains. “That can be satellite data, but also data measured on the ground, policy data, airborne data, or other geographic data.”
He describes this as assembling different puzzle pieces to form a more complete picture that can support policy makers with information ranging from vegetation monitoring dune areas to tracking degradation of road markings over time.
At the core of this approach is a lakehouse, where all these data sources are brought together before more advanced analysis or AI is applied. To structure all the data, the team is also building a data cube, which overlays a hexagonal grid across South Holland. Essentially, each hexagon has a unique index and contains multiple data layers for a specific location. These layers are stored in the cube with a time dimension, allowing multi-year comparisons and trend analysis.
“It’s about using the right pieces together to see what’s actually happening and to identify what kind of extra pieces might be needed,” Oosterhuis says.
More broadly, this represents a shift towards more data-driven policymaking, Oosterhuis argues. Instead of relying on point-based measurements or static reports, policy makers can work with continuously updated data across space and time, allowing them to better understand patterns, trade-offs, and long-term effects of their decisions.
Emerging markets and commercial applications
Both Pasha Zadeh and Oosterhuis see signs of growing commercial interest. Pasha Zadeh points to ESG compliance, sequestration monitoring, and mining as areas where EO is increasingly relevant, bringing more private companies into the market.
“We’ve seen several acquisitions of Earth observation companies by larger investors, as well as increasing investment in EO startups from outside the space sector,” she says. ““That’s a strong indication of the growth potential in the space data and applications market.”
Oosterhuis highlights agriculture and commodities as another example, where satellite data can be used to monitor crop growth and estimate yields at scale, feeding into market analysis. He also points to longer-term possibilities such as biodiversity credits – instead of carbon credits – measured with satellite data, as a future direction for more market-driven EO applications.
How does the Netherlands fare in Earth observation?
Asked about the Netherlands’ position in Earth observation, both Pasha Zadeh and Oosterhuis point to a strong ecosystem.
“For the size of the country, we have quite a lot of companies working in EO, and some of them have been around for a long time.” While the Netherlands is often associated with space hardware because ESA’s ESTEC is based there, she stresses that the country is also strong on the data side, particularly in terms of EO uptake.
That uptake is particularly visible across public institutions, from national ministries to provinces and municipalities, which have long been among the main users of geodata and Earth observation.
Oosterhuis also sees the Netherlands as having a strong position in Earth observation, but frames it less as a centre of innovation and more as a connector. “The Netherlands is very good as a kind of broker,” he says, “a reseller of innovative ideas rather than a breeding ground of innovation in itself.”
This is why he stresses the need for continuous investment. For Oosterhuis collaboration is also central. “Ideally, it’s a triangle,” he says, referring to government, companies, and universities. “The government brings practical use cases, companies bring technology, and universities bring knowledge about new opportunities.”
