When applying a data cube paradigm to analyse a time series of remotely sensed data, one assumes that one can directly compare temporally-separated images. But when one wishes to cover a large region, then achieving seamless wide-area coverage can be challenging. By merging information from multiple medium- or wide-swath sensors into a single 2D regional coverage, one can simplify many analysis streams. Composite products generated from optical imagery that cover wide regions have become common place over time. Such products are often necessary in the case of optical sensors due to the problem of persistent cloud cover. Only composite products can offer cloud-cleared data, often a pre-requisite for wide-area analysis.

Seamless composite wide-area radar backscatter images have until now not been available within data cubes. Although cloud-clearing is not necessary for radar data, other effects provide strong motivation for considering composite products as an intermediate analysis ready data layer. The influence of topography is much stronger on radar than on optical imagery. The geometry of all measurements is shifted by the topography due to the ranging nature of radar data – that can be handled by applying range-Doppler geocoding (orthorectification). However even after that is done, the radiometry of the radar images is significantly modulated in the case of hilly landscapes, and utterly dominated in the case of mountainous terrain. Foreshortening, layover, and shadow each play their role and must be treated in their own way. Radiometric terrain correction as proposed in [1] corrects for both geometric and radiometric effects. However, even after those corrections, variations in the local resolution of the backscatter measurements remain as residual distortions in the flattened backscatter maps. To resolve issues with those residual effects, combination of multiple tracks including ascending/descending geometries was proposed and demonstrated in [2]. This technique weighs contributions from multiple backscatter observations taken with diverse geometries within a narrow time window to generate a backscatter composite with coverage across all included tracks. The local resolution of each observation is used to weigh the strength of its contribution to the composite – which then has lower noise than any individual acquisition combined with higher spatial resolution, as the backslope observations with the higher local resolution are assigned higher weights.

We show wide-area time-series of such composite radar backscatter for multiple regions, and demonstrate analyses based on time-series of this ARD product. We show forest-type classification at full Sentinel-1 IW-mode resolution, as demonstrated in [3]. Wet snow signatures are visible in the attached image, overlaying three such ARD products within the 2018 spring melt season, as the melting began at lower elevations, moving successively higher into the summer. The ARD product has also shown utility in relatively flat regions: melt-onset mapping in sea-ice regions within the Canadian arctic are demonstrated by integrating backscatter from ESA’s Sentinel-1A, -1B, and CSA/MDA’s Radarsat-2 via local resolution weighting. Daily temporal resolution over a wide region was achieved for the first time using these three satellites. With the launch by the Canadian Space Agency of the Radarsat Constellation Mission planned in early 2019, three more C-band satellites will become available, widening the scope for combinations and potentially further narrowing the time-window necessary to achieve full coverage also at more temperate latitudes. We demonstrate how synergetic multi-sensor backscatter integration enables the generation of ARD products that would not be possible using only a single radar satellite constellation. Collaboration across space agencies is required to ensure common standards for geometric and radiometric calibration as well as synergetic acquisition patterns.

[1] Small, D., “Flattening Gamma: Radiometric Terrain Correction for SAR Imagery,” IEEE Trans. Geosci. Remote Sens., 49(8), 3081–3093, Aug. 2011.
[2] Small, D., “SAR backscatter multitemporal compositing via local resolution weighting,” in IEEE International Geoscience and Remote Sensing Symposium, Munich, Germany, 2012, 4521–4524.
[3] Rüetschi, M., M. Schaepman, and D. Small, “Using Multitemporal Sentinel-1 C-band Backscatter to Monitor Phenology and Classify Deciduous and Coniferous Forests in Northern Switzerland,” Remote Sens., 10(55), 30p., Jan. 2018.


Contains modified Copernicus Sentinel data (2018) -- Red: Feb. 24-Mar. 7, Green: Apr. 1-12, Blue: May 1-12

Rapidly growing stacks of SAR data over Australia acquired by the Copernicus Sentinel-1 constellation are enabling applications that require long time-series, including Interferometric Synthetic Aperture Radar (InSAR). Geoscience Australia is developing a scalable workflow to process this large volume of Sentinel-1 SAR data being acquired every 12 days on descending passes over the whole Australian continent. This paper describes the operational InSAR workflow that is being developed for routine production of a suite of Analysis Ready Data (ARD) products suitable for InSAR applications and made available through the Digital Earth Australia (DEA) initiative. Geoscience Australia will use the derived InSAR products from this processing to generate a crustal deformation model for Australia that will enable users to determine local deviations relative to the national geodetic height datum. The InSAR-derived deformation model will also support other geoscience applications, such as subsidence monitoring.
Since becoming a widely used remote sensing technique in the 1990’s and 2000’s, differential InSAR processing has reached a level of maturity where most expert users implement the same procedures to turn a stack of SAR single look complex (SLC) products into a network of geocoded unwrapped interferograms. The methodology is therefore an ideal candidate for standardisation and routine generation of ARD products for open dissemination to the community, including non-experts. The usual procedures implemented in differential InSAR processing include co-registering the stack of SLCs, forming a differential interferogram, removing the orbital and topographic contributions of the phase signal, filtering, unwrapping, and finally, geocoding. The output product of this differential InSAR processing (a geocoded unwrapped interferogram) is then typically the input to further time series processing to derive the displacement history and average displacement rate (velocity) for certain pixels in the SAR image stack. Time series processing can be implemented via the well-used persistent scatterer (e.g. implemented in the StaMPS software) or SBAS techniques (implemented in the PyRate software).
Our InSAR processing workflow makes use of the GAMMA InSAR processing software. The compiled binary programs of the GAMMA software are linked together and parallelised via a series of scripts for operation on the petabyte-scale raijin system of the Australian National Computing Infrastructure. Leveraging the amassing SAR data and the available compute resources will enable up-scaling to continental product generation in Australia. As standard output of the workflow, we produce a network of geocoded unwrapped interferograms (for a given SLC stack), together with geocoded flattened interferograms (with wrapped phase) and interferometric coherence. The definition of an ARD standard for these products will form part of Geoscience Australia’s contribution to the global effort to define a set of well-defined ARD products under the CEOS Analysis Ready Data for Land (CARD4L) framework. Furthermore, SAR backscatter is generated according to the existing CARD4L SAR backscatter product definition. Examples of the InSAR and backscatter products together with a progress report on the development of the ARD product definitions will be given in this paper.

Access to well calibrated and standardised Earth observation data is a fundamental requirement for effective and rapid development of information products under the Digital Earth Australia (DEA) initiative. Funded by the Australian Government, DEA is an analysis platform for observations of all forms, but particularly those from satellites which have unique potential and pose particular challenges for their full exploitation. DEA will enable the processing, interrogation and presentation of petabyte-scale data from satellite observations for the government’s priority information needs and also provide businesses with access to reliable, standardised satellite data that can be used to build new applications for commercial purposes.

Analysis Ready Data (ARD) are satellite data that have been processed to a minimum set of requirements and organised into a form that allows immediate analysis with a minimum of additional user effort and interoperability both through time and with other datasets. The DEA Surface Reflectance (SR) dataset is a foundational ARD product that enables dense time-series analyses and rapid generation of other information products. Field based validation of the SR product is critical to characterise uncertainties that determine its fitness for purpose, the generation of reliable quality indicators and its applicability in downstream product development.

In this paper, an approach for continental scale validation of the DEA surface reflectance products from Landsat and Sentinel-2, using field-based measurements that are near-synchronous to the satellite observations over multiple sites across Australia will be presented. Best practice measurement protocols governing the acquisition of field data, including field instrument calibration, sampling strategy and approach for post-collection processing and management of field spectral data will be outlined.

Details of a nationally coordinated, collaborative field data collection campaign in Australia including the identification of permanent field sites, to support validation efforts within the broader EO community for continental scale products will also be presented. The approach is expected to serve as a model for coordinated ongoing validation of ARD products at continental to global scales.

The generation of analysis ready data (ARD) has become important for operational earth observation programmes relying on the processing of large volumes of satellite imagery. This is particularly relevant where quantitative image analysis techniques are used for automated processes. A range of preprocessing and normalisation techniques are considered beneficial or even necessary for the generation of ARD. Each of these techniques intends to reduce error in order to make significant inference of change presented by the data. These techniques include ToA reflectance normalisation, atmospheric correction, topographic normalization, and BRDF correction.
In this research, the error budgets have been assessed between a time series of Landsat 8 and Sentinel 2 imagery over four study areas covering a range of climatic zones: Vancouver Canada, New Orleans USA, Lofoten Norway and Tzaneen South Africa. For each study area at least 40 Sentinel 2 and Landsat 8 scenes are used acquired during the vegetation period of 2018. For Vancouver and New Orleans, Level 2 and ARD Landsat products supplied by the USGS were used for independent validation.
The reduction of error is quantified for the following normalisation techniques:
1. ToA reflectance using both mean solar geometry versus pixel based solar geometry.
2. Atmospheric correction using LaSRC and Sen2Cor in an automated fashion versus manually tuned atmospheric correction parameters via Atcor as implemented in PCI Geomatica 2019.
3. Topographic normalisation using the Lambertian correction, C correction, Minnaert correction, generic models and a BRDF sky light adaptive technique.
4. NBAR BRDF correction comparing global techniques to surface specific corrections using class specific average correction coefficients versus measured BRDF coefficients from the MODIS product MCD 43.
Various matrices are used to quantify the reduction of error, varying from the use of pseudo invariant features to class / spectral signature correlation and inter-scene variability.
This research has provided insight to the relevant conditions that justify the application of certain image normalisation and pre-processing technique for ARD data. It provides guidance on when simple ToA reflectance is sufficient for analysis, which atmospheric correction technique provides the best and most consistent results, when topographic normalisation is required and what technique produces reflectance values no longer correlated to local incidence angles, and which BRDF correction technique provides the best NBAR results. Stability and trade-offs for all of these techniques, as well as their suitability for operational use in automated processing chains are discussed in detail.

The production of Earth Observation (EO) Analysis Ready Data (ARD) can only be assessed as fit for purpose if there is a rigorous assessment of the accuracy of the data. The Digital Earth Australia (DEA) initiative aims to assemble national scale remotely sensed ARD. To assess its accuracy, the DEA EO data is being assessed by comparison with ground surface reflectance data obtained as part of a national scale field validation program. The surface reflectance data are being collected at ~10 field sites across Australia representing a range of surface types and observations are being acquired across a range of atmospheric and solar geometry conditions. Following an established protocol (Malthus et al. 2018) spectroradiometric measurements of solar radiance reflected from homogeneous 100 x 100m ground sites are referenced to a measurement of radiance reflected from a calibrated Spectralon plate.
When high spatial resolution field reflectance measurements are used to validate satellite EO data they are typically averaged over an area equivalent to the scale of the satellite sensor pixel. The statistical variability of the ground data is then accepted as a good estimate of the ground reflectance uncertainty. However, the uncertainty in the individual ground reflectance spectra should also be considered. The quality of the ground reflectance measurements is affected by factors such as sky conditions and changing solar angle encountered during the field campaign, placement of the Spectralon plate in the field, characterisation and calibration of the Spectralon plate, and various ASD operation factors. We present here a description and analysis of the factors leading to uncertainty in the individual field spectra and their relative importance. A good understanding of these uncertainties is critical in fully characterising the ground surface reflectance uncertainty, and in turn, assessing the validity of the ARD.

Reference:
Malthus, TJ, Ong, C, Lau, I, Fearns, P, Byrne G and Thankappan, M., Chisholm,L., Suarez, M., Clarke, K., Scarth, P., Phinn, S. (2018) A community approach to the standardised validation of surface reflectance data. A technical handbook to support the collection of field reflectance data. Release version 1.0. CSIRO, Australia. ISBN: 978-1-4863-0991-7

We are currently experiencing an exciting new era of Earth Observation, wherein multiple, freely available remote sensing systems provide us data at unprecedented spatial, temporal and spectral resolutions. This regular data influx might enable us to achieve sustainable development goals by closely monitoring environmental status and change at relevant scales. However, this flood of data can easily be overwhelming, both in terms of volume and usage complexity. While the first point is merely a technical burden that can be leveled out with enough processing power and investments in storage, EO data still need to be processed to a considerable degree before being adequate for most analyses.
To alleviate this, the Framework for Operational Radiometric Correction for Environmental monitoring (FORCE) is being developed to be an ‘all-in-one’ solution for the mass-processing of medium-resolution satellite image archives to enable large area and time series applications. FORCE is capable of processing Landsat 4–8 and Sentinel-2 Level 1 data to Level 2–4 products, which are essentially different degrees of Analysis Ready Data (ARD). Such data can be analyzed with a minimum of additional user effort, which will empower much broader user groups in using mass data from Landsat and Sentinel-2.
Currently, the term ARD is mostly used to describe data that are radiometrically and geometrically consistent, and that include cloud and other poor quality observation flags that allow filtering data prior to analysis. This definition is well aligned with FORCE Level 2 ARD, which are essentially radiometrically corrected Bottom-of-Atmosphere reflectance products. FORCE Level 2 data are accompanied by pixel-based quality information and are stored in a gridded data cube using a single projection across large areas.
Whilst these data can indeed be directly analyzed, usage is complex nonetheless. Data volume is still similar as before, and Level 2 ARD are spatially incomplete (due to observation characteristics, and quality screening). Therefore, FORCE provides a Level 3 Processing System that generates temporal aggregations of ARD to provide seamless, gap free, and highly Analysis Ready Data (hARD) over very large areas. This includes Best Available Pixel (BAP) composites as well as Spectral Temporal Metrics (STM, e.g. average reflectance within a defined time period). Unlike ARD, hARD are considered the optimal input for many machine learning applications, e.g. for large area land cover /change classification purposes.
Level 4 data are defined as “model output or results from analyses of lower level data, e.g., variables derived from multiple measurements” (Asrar and Greenstone 1995). To this end, FORCE provides the Time Series Analysis (TSA) module that is capable of generating and analyzing quality-controlled time series from spectral bands, indices or unmixed fractions using a number of aggregation and interpolation techniques. Annual Land Surface Phenology (LSP) metrics can be derived, and change and trend analyses can be performed on any of the generated time series. Many outputs of FORCE TSA are considered as highly Analysis Ready Data plus (hARD+), meaning that generated products can be opened in a GIS and be directly analyzed to fuel your research questions without any further processing.
Asrar, G., and R. Greenstone, eds (1995). MTPE EOS Reference Handbook. Greenbelt, MD, USA. NASA/Goddard Space Flight Center.