At Adatos, we specialise in using satellite imagery to train our AI models and perform monitoring and predictions for clients. We stay updated with the latest technologies and evaluate each satellite platform to ensure we use the most effective tools for specific use cases. We also prioritise the preprocessing of the data to ensure it is of a high quality before being input into models. Adatos believes that for most use cases in carbon and agriculture, free 10 m resolution data can produce sufficiently accurate analysis.
The development of remote sensing for agriculture and carbon applications has revolutionised the way that we monitor and manage these critical areas. Satellite data offers unique advantages over handheld sensors, drones, and aerial platforms, including scalability and an extensive archive of historical imagery. By understanding the different types of satellite data available, choosing the best platform and carefully preprocessing the data, we can unlock a wealth of valuable information that can help us to better understand and manage our planet's natural resources.
Source: Media from Wix
Remote sensing has been used in agriculture and forestry for many decades. However, the development of satellite technology has brought remote sensing to a new level, allowing us to monitor vast areas more frequently and accurately. Satellites have significant advantages over other remote sensing platforms:
Scalability: cover vast areas in a short amount of time, making them ideal for monitoring large-scale agricultural activities such as crop growth, land use changes, and deforestation at low cost.
Archive: store large amounts of data, which can be useful for historical analysis and time series analysis.
Frequent monitoring: monitor the same area multiple times a day or week, allowing us to track changes in vegetation, land use, and other environmental factors over time.
Seeing the Unseen: Understanding the Types of Satellite Data
Satellites can see beyond what the naked eye can see, each specialising in a different kind of ‘vision’. Optical satellites can detect beyond the visible spectrum, often including infrared wavelengths. Active radar (SAR) and laser (LiDAR) can see physical properties such as texture, elevation and volume. Atmospheric satellites often operate in the microwave, infrared and UV ranges. Each type of satellite data has its unique function and advantages.
1. Optical - Multispectral and Hyperspectral
Optical satellites capture information across different wavelengths of the electromagnetic spectrum, not just the visible spectrum. Multispectral platforms are usually able to detect near-infrared (NIR) and shortwave infrared (SWIR) on top of visible wavelengths. The NIR band is sensitive to chlorophyll levels and water quality, making it useful for vegetation monitoring and water management applications. The SWIR band is used to identify land cover and monitor the condition of vegetation and crops using vegetation indices, as it can map plant moisture and soil composition. Overall, the range of wavelengths in multispectral satellite data provides valuable information for various applications in agriculture and environmental management.
Unlike multispectral satellites that capture data across a certain number of bands, hyperspectral satellites capture data across hundreds of narrow and contiguous spectral bands. The higher spectral resolution theoretically allows more detailed analysis such as detecting nutrient deficiencies. Unfortunately, based on our experience so far, its limited archive, restricted access and unresponsive tasking systems have rendered hyperspectral products difficult to use in practice.
Advantages of using optical satellite data:
Spectral resolution: Optical satellite data captures information across different wavelengths of the electromagnetic spectrum, providing detailed spectral signatures that can be used to identify land cover and monitor the condition of vegetation or crops using vegetation indices.
Interpretation: Optical satellite images provide a clear visual representation of the Earth's surface, which can be easily interpreted by human analysts. This makes it easier to identify changes in land use or vegetation cover.
2. Synthetic Aperture Radar (SAR)
SAR is an active radar technology that operates at microwave frequencies to detect changes in the Earth's surface. SAR data can identify height differences and penetrate through clouds and vegetation, making it useful for measuring biomass and volume, as well as creating digital elevation models. Additionally, SAR is sensitive to texture and moisture, which makes it helpful for identifying land covers and monitoring plant conditions. SAR data is often combined with optical data for more comprehensive insights as it is all-weather.
Advantages of using SAR data:
Cloud penetration: SAR data can penetrate through clouds, making it useful for monitoring changes in areas with frequent cloud cover.
Day/night imaging: SAR data can be acquired day or night, making it useful for monitoring changes in areas with limited daylight.
Provides additional, often complementary data to optical satellites
3. Light Detection and Ranging (LiDAR)
LiDAR uses laser pulses to measure the distance between the instrument and the Earth's surface and is a well established technology, however expensive and historically required a plane to take measurements. Satellite-based LiDAR is a relatively recent development, with only a handful in operation. Similar to SAR, some uses of LiDAR are canopy height and biomass mapping.
Advantages of using LiDAR data:
Day/night imaging: As LiDAR is also active imaging, data can be acquired day or night. However, LiDAR cannot penetrate through clouds.
Centimeter level accuracy: The lasers of LiDAR can focus on smaller spots, giving more precise measurements.
LiDAR's high resolution comes with a tradeoff of smaller footprints, which limits its coverage. GEDI, for example, measures in 25 m diameter footprints spaced 60 m apart along 8 tracks. This results in uncertainties in the gridded 1 km x 1 km aboveground biomass (AGB) product, which is inferred from the footprint samples. Moreover, there are still significant spatial gaps in between gridded products. Although GEDI provides important information for estimating aboveground biomass (AGB) at regional or global scales, its limited spatial coverage and low resolution can make it challenging to accurately estimate AGB at smaller scales using only GEDI data.
4. Atmospheric satellites
Atmospheric satellites are designed to monitor various atmospheric parameters, such as weather patterns, air pollution, greenhouse gases, and ozone layer depletion. These satellites collect data using a variety of sensors, including visible, infrared, and microwave sensors.
NASA recently discovered that the OCO-2 and OCO-3 satellites, designed for monitoring CO2 at global and continental scales, could detect specific source emissions from the Bełchatów power plant in Poland. The measurements showed a high correlation of 0.885 when compared to reported power generation. This highlights the potential benefits and accuracy of space-based measurements. Read our full assessment of OCO-2 and OCO-3 satellites and their potential to monitor nature-based solutions here.
Despite such breakthroughs, further development is still needed to optimise satellites for monitoring emissions with high accuracy at scale. Atmospheric study satellites are often either low spatial resolution (e.g. Sentinel 5P has global coverage but 5.5 km x 3.5 km resolution), or limited in spatial coverage (e.g. GHGSAT has ≤30 m resolution but only captures small tasked areas). These tradeoffs make it difficult to pinpoint sources of greenhouse gas emissions over large areas. Upcoming constellations, such as MethaneSAT aim to bridge this gap.
Choosing the Best Platform for Your Use-Case
Apart from selecting the most appropriate type(s) of satellite data for your application, here are some other important factors:
Resolution: Higher resolution provides more detailed information, but it also comes with higher costs and limitations in terms of coverage area. It may also not be required for your use case.
Cloud cover: There are ways to work around cloud cover with techniques such as compositing or masking and stitching images close to the time of interest.
Revisit: The frequency at which a satellite platform can revisit an area is important for applications that require frequent monitoring or the detection of rapid changes. High revisit is also useful for image compositing or stitching in cloudy areas.
Archive: Historical training data and time series analysis require access to archives of satellite data. Some platforms offer access to extensive archives of historical data, but tasking satellites or more recent platforms often have limited archives.
Cost: In many cases there is no need to pay for imagery. High-resolution and tasking satellites are usually expensive particularly for large areas. It's important to consider the long-term costs of satellite data acquisition, including subscription fees and data processing/storage costs.
Through our extensive work on precision agriculture and nature-based solutions projects, we have consistently identified Sentinel-1 and Sentinel-2 as the optimal satellite options. These satellites offer a high degree of granularity with a resolution of 10 m, a rich archive of data, a revisit cycle of 5-6 days, and a wide spectral range for multispectral data. Furthermore, they provide these capabilities at no cost, making them an attractive choice for a range of applications.
Connect with Us
From agriculture and forestry to climate change and disaster management, satellite imagery is playing an increasingly important role in collecting data and monitoring our planet. At Adatos, we are committed to staying at the forefront of this technology and leveraging it to provide our clients with the most accurate and actionable insights possible. If you are interested in learning more about our services, please don't hesitate to message us on LinkedIn. We would love to connect and discuss how we can help you achieve your goals using the power of satellite imagery and AI.