Published February 20, 2023 | Version 1.0.0
Dataset Open

Sentinel-1 Global Harmonic Parameters: A Seasonal Model for Flood Mapping and More

Description

This dataset was generated by the Remote Sensing Group of the TU Wien Department of Geodesy and Geoinformation,  within Framework Contract (No. 939866-IPR-2020) as part of the provision of an automated, global, satellite-based flood monitoring product for the Copernicus Emergency Management Service (CEMS) managed by the European Commission. The Global Flood Monitoring (GFM) product is integrated within the user interface of the Global Flood Awareness System (GloFAS) of the CEMS. Open use of the dataset is granted under the CC BY 4.0 license.

The Copernicus Sentinel-1 constellation is a highly-capable monitoring mission and provides one of the most comprehensive global archives on satellite imagery. The satellite sensors acquire Synthetic Aperture Radar (SAR) images, and as such, they observe regardless of weather conditions and daylight. The regular and systematic observations generate rich information on the global land surface and its dynamics, which is used for---but not limited to---terrestrial applications like e.g. land cover mapping, flood detection, or drought monitoring.

The complete Sentinel-1 time-series dataset is challenging to analyze, primarily due to its sheer data volume of at the (global) scale of Petabytes. As a user-friendly alternative, this dataset provides a Harmonic (Fourier) series model that reduces the SAR backscatter seasonality to a relative small number of GeoTIFF files holding the harmonic coefficient values.

This dataset publication provides a temporal Sentinel-1 model for most of the world's land masses. Seven coefficients computed using (harmonic) least squares regression, along with the standard deviation of residuals and number of observations, comprise the harmonic parameter set. The parameters are being operationally used to determine the expected SAR backscatter signal for any day of the year as part of the TU Wien's method contributing to GFM's ensemble flood monitoring effort (Bauer-Marschallinger et. al, 2022). The Global Harmonic Parameters (HPARs) were derived from the whole Sentinel-1 VV temporal stack for the period 2019-2020 by least squares regression with a harmonic model formulation, running three sinusoidal iterations (k=3).

The model describes the typical seasonal Sentinel-1 backscatter variation on a 20 m pixel level. It was designed as a smoothed time-series approximation, removing short-term perturbations, such as speckle and transient events (like floods for instance). Hence, the model is suited to discern the seasonal changes brought about by varying water content, e.g., inundation or soil moisture, and progression of vegetation structure.

We encourage developers from the broader user community to exploit this extensive and functional data resource. In particular, we promote the use of these Sentinel-1 HPARs in models for various applications dealing with land cover, seasonal water mapping, or vegetation phenology.

For the datasets' theoretical formulation and primary use case as a non-flooded backscatter reference model, please refer to our peer-reviewed article. Additionally, the software used, computation process, and outlook are discussed in this conference paper.

Data record

The parameter sets are provided per Sentinel-1's relative orbit to account for geometric effects. The parameter files are sampled at 20 m pixel spacing, georeferenced to the Equi7Grid, and divided into six continental zones (Africa, Asia, Europe, North America, Oceania, and South America. For portability and easier downloads, further sub-divisions into continental parts are done, resulting in 12 compressed bundles (please refer to coverage map).

The parameter sets are provided per Sentinel-1's relative  to account for geometric effects. The parameter files are sampled at 20 m pixel spacing, georeferenced to the Equi7Grid, and divided into six continental zones (Africa, Asia, Europe, North America, Oceania, and South America);. For portability and easier downloads, further sub-divisions into continental parts are done, resulting in 12 compressed bundles (please refer to coverage map).

The data itself is organised as square tiles of 300 km extent ("T3"-tiles). Note that the parameters are generated for each orbit, resulting in several orbit-sets per tile. Given this structure, a total of 98910 files for the 10990 tiled orbit-sets, comprising overall a compressed disk size of 3.7 TB.

The datasets follow the Yeoda filenaming convention (documentation here) where the core meta information is embedded. Notably, the file name is prefaced by the product name 'SIG0-HPAR-' and the particular parameter codes:

  • M0 - effective mean of the time series stacks, also called the harmonic residual in other literature.
  • Cn - cosine component coefficients, where n = 1, 2, or 3.
  • Sn - sine component coefficients, where n = 1, 2, or 3.
  • STD - standard deviation of residuals, a proxy for model goodness of fit.
  • NOBS - number of observations used for the least squares regression, also
    an indicator of solution quality.

Orbit sets are distinguishable by orbit direction, i.e. (A - ascending and D - descending) and relative orbit number, for example: 'A175', 'D080'.

File naming scheme is as follows:

SIG0-HPAR-NNN_YYYYMMDD1_YYYYMMDD2_VV_OOOO_TTTTTTTTTT_GGGG_V02R01_S1IWGRDH.tif 

*bold faced items are fixed for this product version.

  • NNN - product name which indicates parameter code.
  • YYYYMMDD1 - start date of time series processed.
  • YYYYMMDD2 - end date of time series processed.
  • VV - polarization of product.
  • OOOO - relative orbit.
  • TTTTTTTTTT - Equi7grid tile name.
  • GGGG - Subgrid. Contains continent code and sampling size.
  • V02R01 - Product version.
  • S1IWGRDH - Sensor type.

For example:
'SIG0-HPAR-STD_20190101_20210101_VV_D111_E102N066T3_SA020M_V02R01_S1IWGRDH.tif'

The parameters' file format is an LZW-compressed GeoTIFF holding 16-bit integer values, with tagged metadata on encoding and georeference. Compatibility with common geographic information systems such as QGIS or ArcGIS, and geodata libraries as GDAL is given.

This repository provides all parameter sets per orbit for each tile and is organized in a folder structure per (sub-)continent. With this, twelve zipped dataset collections per (sub-)continent are available for download.

Code Availability

We suggest users to use the open-source Python package yeoda, a datacube storage access layer that offers functions to read, write, search, filter, split and load data from this repository as an HPAR datacube. The yeoda package is openly accessible on GitHub at https://github.com/TUW-GEO/yeoda.

Furthermore, for the usage of the Equi7Grid we provide data and tools via the python package available on GitHub at https://github.com/TUW-GEO/Equi7Grid. More details on the grid reference can be found in this publication .

A day-of-year estimate reader tool based on the packages above is likewise available on GitHub at https://github.com/TUW-GEO/hpar-reader

Acknowledgements

The authors would like to thank our colleagues: Thomas Melzer of TU Wien for his invaluable insights on the parameter formulation, and Senmao Cao of Earth Observation Data Centre GmbH (EODC) for his contributions to the code base used to process dataset.

This work was partly funded by TU Wien, with co-funding from the project "Provision of an Automated, Global, Satellite-based Flood Monitoring Product for the Copernicus Emergency Management Service" (GFM), Contract No. 939866-IPR-2020 for the European Commission's Joint Research Centre (EC-JRC), and the project "Flood Event Monitoring and Documentation enabled by the Austrian Sentinel Data Cube" (ACube4Floods), Contract No. 878946 for the Austrian Research Promotion Agency (FFG, ASAP16).

The computational results presented have been achieved using the Vienna Scientific Cluster (VSC). We further would like to thank our colleagues at TU Wien and EODC for supporting us on technical tasks to cope with such a large and complex dataset.

Files

HPAR_Example_Figure.png
Files (3.7 TiB)
Name Size
md5:ed85d8cce21a5236a151d78e4be3c9e7
5.4 MiB Preview Download
md5:2d6cae1f59c7559e154475922c88d7d8
275.6 GiB Preview Download
md5:35c214f0c71cf8b127e74a3ac83eec6d
271.1 GiB Preview Download
md5:3a7646d00e203063e1d457b26a67a316
344.2 GiB Preview Download
md5:3bf458d714bceceae5c15aeb3374922f
335.4 GiB Preview Download
md5:060d07fe8e5a5e4e53599dff12488547
378.3 GiB Preview Download
md5:ecbbb424ddb3b65652e5ce06fb3668d9
440.2 GiB Preview Download
md5:fcf83c70e06f07a8f6e8308f55e0759f
276.6 GiB Preview Download
md5:f78f3fdb0298307585620db512ccec77
258.1 GiB Preview Download
md5:ae973c6c0ae87ef99bfca144881ace01
203.0 GiB Preview Download
md5:da3d83777dc2ec39ad037cea970d5544
425.9 GiB Preview Download
md5:38f524ee96e1809fd8f76ad1b65f511b
306.7 GiB Preview Download
md5:e47fe976dee150db5cce14737b7958a8
237.0 GiB Preview Download
md5:fb2a749b48d2b66ccb831e4535602f93
761.1 KiB Preview Download

Additional details

Created:
March 13, 2023
Modified:
October 17, 2023