Satellites are now routinely used for measuring water and land surface reflectance and hence parameters such as aquatic chlorophyll a concentration and terrestrial vegetation indices. Since these remotely sensed parameters may be significantly affected by errors in top of atmosphere calibration and atmospheric correction, radiometric validation at the bottom of atmosphere is needed for quality control, ensuring that data reaching end-users are of known quality and that quality issues are reported to satellite data providers for improvement.

For each satellite mission, radiometric validation is needed for all spectral bands and covering all typical conditions where the satellite data will be used, including various water/land types, sun angles, atmospheric conditions (aerosols, absorbing gases, scattered and semi-transparent clouds), surface altitudes, spatial heterogeneity, etc.

The users of in situ measurements of water and land surface reflectance are identified as:

• Satellite operators, including international and national space agencies and “Newspace” commercial data providers—see Supplementary Material S1 for a list of target missions.

The user requirements for in situ measurements of water and land surface reflectance are identified as (Goyens et al., 2018):

• Hyperspectral coverage 380–1020 nm for water and minimally 380–1700 nm but preferably 380–2500 nm for land.

In addition to the reflectance measurements, users request pictures from cameras and information on aerosols and direct/diffuse irradiance ratio.

Atmospheric applications are specifically excluded here because aerosol parameters are already well-measured by AERONET (Holben et al., 1998) and because measurement of absorbing atmospheric gases (Verhoelst et al., 2021) generally requires much finer spectral resolution, e.g., sub-nanometre.

The user needs of Section 1.2 are partially met by the existing networks, AERONET-OC for water and RadCalNet for land.

The AERONET-OC 12 band instrument (Zibordi et al., 2021) matches well the spectral bands of “ocean colour” sensors, but less well the wide bands of “land” sensors such as Sentinel-2/MSI, Landsat-8/9, Planet Superdoves, etc.—see Figure 1. Validation of the new generation of hyperspectral sensors by multispectral ground measurements is also insufficient (Giardino et al., 2020; Braga et al., 2022). While aerosol correction can be assessed by a limited set of multispectral bands, the new potential of hyperspectral satellite data, e.g., spectral curvature/derivative algorithms (Dierssen et al., 2020; Lavigne et al., 2022), will require hyperspectral in situ data.

RadCalNet (Bouvet et al., 2019) is designed to provide hyperspectral radiance at top of atmosphere for the purpose of satellite vicarious (in-flight) calibration. RadCalNet sites are located in optimal locations with horizontal homogeneity of the surface and clear, stable atmosphere and do not cover the more difficult surfaces required for a complete sensor validation plan. Some sites do not use hyperspectral instruments. RadCalNet provides only nadir-viewing data.

While both AERONET-OC and RadCalNet must continue to provide vital data for calibration and validation of satellite missions, it is our viewpoint that a new in situ measurement network is needed to fully satisfy the user requirements for multi-mission water and land surface reflectance validation. This was the motivation for setting up HYPERNETS, a federated network of sites running autonomous hyperspectral radiometer systems to provide validation data at all spectral bands in the range 380–1700 nm for all satellite missions with a surface reflectance product.

The land and water optical remote sensing communities have traditionally operated very separately with different data processing chains, atmospheric correction algorithms, etc. This separation is driven by the different user groups and data products, and by different physical processes, particularly those relevant for atmospheric correction.

Water is generally much darker than land, exacerbating challenges for aerosol correction, adjacency effect correction, vicarious calibration, absorbing aerosol detection/correction, etc. Removal of air-water interface reflection is an additional challenge, particularly as regards sunglint removal for near nadir-viewing sensors. However, there are algorithms that can apply over both water and land surfaces, e.g., iCOR (Keukelaere et al., 2018) and ACOLITE/DSF (Vanhellemont, 2019).

Land surfaces may have higher spatial heterogeneity at short length scales (<30 m) and high angular but low temporal variability of upwelling radiance.

Despite these differences, the HYPERNETS consortium decided to develop instrumentation and processing for both water and land surfaces, benefitting from the economy of scale for radiometer, host system and processor development and offering a larger customer base for the new radiometer. The commonality in data processing and distribution for the land and water network will also facilitate validation of land and water surface reflectance simultaneously and open up new opportunities to study complex interactions between water and land environments.

In this Perspective Article we discuss recent advances in optical radiometry and describe the HYPERNETS federated network concept (Figure 2). HYPERNETS integrates two branches for the different surface types: WATERHYPERNET (Ruddick et al., 2024) and LANDHYPERNET (De Vis et al., 2024a). Both branches use the newly developed HYPSTAR^® radiometer system, which is the main focus of this article. The PANTHYR/TriOS radiometer system (Vansteenwegen et al., 2019) is also included in WATERHYPERNET and in the summary of results of Section 3.

For water reflectance measurements, automated above water radiometry (Zibordi et al., 2009) has proven to be cost-effective. HYPERNETS adopts the same approach as AERONET-OC for measurement of water-leaving radiance, L w , but makes direct measurement of downwelling irradiance, E d , using a flat diffuse collector with approximately cosine angular response instead of sun photometry—see Ruddick et al. (2024).

For land surface reflectance measurements, protocols for automated radiometry are less mature. Most measurements are supervised and use a reflectance based method with a reflectance standard as a reference (Slater et al., 1987). An autonomous multiangle surface reflectance protocol was developed by Meygret et al. (2011) for measurement of Hemispherical Conical Reflectance Factor (HCRF) with a multispectral instrument. This protocol could not be implemented by HYPERNETS due to the hyperspectral measurement time, power and data transfer requirements. HYPERNETS developed a sequential acquisition protocol (De Vis et al., 2024b), measuring E d before and after a series of upwelling radiance, L u , measurements at nadir and azimuth angles corresponding to typical viewing geometries of polar-orbiting satellites—see Supplementary Material S3.

While the basic water and land reflectance measurement scenarios are sufficient, complementary measurements can be added, e.g., sunglint pointing for estimation of wave slope statistics (Goyens and Ruddick, 2023); direct sun radiance measurement, potentially supplemented with various sky radiance measurements, for comparison with the direct E d measurement; a more complete measurement of land surface HCRF, etc.

The user requirements of Section 1.2, particularly the spectral requirements and the need for a pointable radiometer could not be met by any Commercial Off The Shelf (COTS) radiometer—see Kuusk et al. (2024). The HYPERNETS consortium therefore designed a new hyperspectral radiometer, the HYPSTAR^®. This innovative design (Kuusk et al., 2024) is based on spectrometers covering 380–1020 nm at 3 nm FWHM with 0.5 nm sampling interval and (land units only) 1020–1680 nm at 10 nm FWHM and 3 nm sampling interval. The spectrometers are multiplexed between alternative optical paths, one for radiance and one for irradiance, reducing cost and minimising impact of thermal sensitivity and some absolute calibration uncertainties on reflectance products. One innovative feature of the HYPSTAR^® radiometer design is the integration of an external LED source, which can be used at night for monitoring the long term stability of radiometric calibration. Another useful feature, first suggested in the OSPREY design (Hooker et al., 2012), is an embedded RGB camera. This camera has proven extremely useful for troubleshooting equipment failures and strange radiometric data—at some sites it is not unusual to see a bird sitting on the radiometer, partially obscuring the field of view!

The HYPSTAR^® system integrates the radiometer with a pointing system, a PC, a power source and data transmission hardware and associated software/firmware (Doxaran et al., 2024; Kuusk et al., 2024).

For autonomous systems deployed with minimal maintenance requirements, and in potentially hostile environments (hot, cold, wet, salty), remote connection capabilities should be ensured and all components must be selected for reliability. HYPSTAR^® has only two moving parts, the radiometer multiplexer and the pointing system.

Over the last decade, technological improvements and mass COTS production have been significant for: pointing systems (now ubiquitous for security cameras), data transmission (4G), photovoltaic power and solid state drives.

The HYPSTAR^® system (Kuusk et al., 2024), integrates a Will-Burt Bowler RX pointing system with a Cincoze rugged PC and a custom-made board integrating relays for efficient power management. Components were selected for an operating temperature range of −25°C to +45°C and housings are rated with Ingress Protection IP66, IP67 or IP68.

Power supply is site-specific, preferably grid power, but otherwise photovoltaic or wind power with associated battery and controller. Data transmission is generally by cellular or cabled internet, but may be manual for some locations, e.g., Gobabeb and Antarctica.

System components are attached to mounting structures, which vary from large platforms to standalone masts. The choice of platform and mounting location is important for data quality and should be made to limit optical perturbations of the radiometric targets by the structure and maximise acceptable viewing angles—see Section 3.3 of Ruddick et al. (2024).

A validation network should include sites covering a wide range of surface types and atmospheric conditions, in particular including the “difficult cases” where atmospheric correction may fail or produce poor results.

Water sites should cover a range of turbidity, phytoplankton biomass and species, Coloured Dissolved Organic Matter absorption, sun zenith angle, clouds, aerosols, etc. Sites at high altitude and with strong adjacency effects are also needed.

Land sites should cover a range of substrates and vegetation including non-vegetated, snow/ice, grassland, agricultural with various crops and practices, forest with various species, age and canopy cover, etc.

When comparing satellite and matchup in situ measurements it is necessary to consider additional uncertainties associated with the different wavelength, space, time and angular coordinates.

A key question for each validation site is the spatial heterogeneity. Each validation site should be characterised in terms of spatial variability to give the additional uncertainty involved in a matchup comparison as function of length scale, e.g., Dogliotti et al. (2015), Dogliotti et al. (2024), Doxaran et al. (2024) for water and Morris et al. (2024) for land.

Temporal variability at a validation site, also of importance in satellite/in situ matchups, can be characterised from time series of the in situ radiometer measurements (Doxaran et al., 2024).

For land sites, the angular variability of upwelling radiance is very high for many surface types, particularly for vegetated surfaces and shadowing surfaces. The BRDF needs to be modelled for each site (Schunke et al., 2023), using measurements at multiple sun and viewing zenith and azimuth angles. The separation of spatial and angular variability may be complex, but should be feasible if data are collected over sufficient time and for multiple sun and viewing angles.

The list of currently operated HYPERNETS validation sites is provided in Supplementary Material S2.

Network data processing and quality control (De Vis et al., 2024a) must be centralised and automated to ensure efficient operations and reliability of data provided to users. Measurements are acquired at the validation sites every 15–30 min during daylight. Raw data and metadata are transmitted to computer servers, one for water and one for land processing.

Identical processing is applied for water and land sites up to the level of calibrated radiances and irradiances with associated uncertainties. The derivation of water and land surface reflectance then uses different measurement functions, but within a common software framework. This integration of land and water processing facilitates joint use of quality control tests, e.g., comparison of E d with a clear sky model to check for clouds and obstructions (including birds).

An important feature of the HYPERNETS processor, of relevance to Fiducial Reference Measurements (Donlon and Zibordi, 2014; Goryl et al., 2023), is the propagation of measurement uncertainties from their sources to the final reflectance data using a Monte Carlo approach (International Standards Organisation ISO, 2008). The implementation, using the open source CoMet toolkit, ² preserves temporal and spectral uncertainty covariance. The covariance information (De Vis et al., 2024b) provided with the output reflectance ensures that the uncertainty of downstream products, such as Normalised Difference Vegetation Index (NDVI) or aquatic chlorophyll a concentration, can be correctly evaluated.

For optimal user uptake, data from the network will be publicly distributed through an interactive user interface, and an Application Programming Interface (API). Data are provided in NetCDF format with relevant metadata, following the INSPIRE directive, and are covered by a CC-BY-ND open data licence.

HYPERNETS provides support for both site operators and data users. Site operators receive installation and troubleshooting advice, share experiences on site operations and data exploitation in scientific discussions and receive radiometer calibration and characterisation services.

A common request from space agencies is for a list of validation site coordinates to be used in prioritising acquisitions, e.g., during a commissioning phase. The coordinates of HYPERNETS sites are provided in Supplementary Material S2.