Global AW3D30
Introduction
The “ALOS World 3D 30 m mesh (AW3D30)" dataset is a recent digital surface model (DSM) with global coverage. The version 2.1 of the dataset was released in early 2018 – and it holds a global coverage in a 30m (1-arc-second) resolution. The dataset was created with the aim for a target vertical accuracy of 5m RMSE – and a horizontal accuracy of 5m. AW3D30 was generated from data obtained from the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) - an instrument that was carried onboard the Japanese Advanced Land Observing Satellite (ALOS). The ALOS satellite was launched on January 24th, 2006; it had an expected target life of 5 years. After a successful mission it ended its operations during May 2011 due to a power anomaly. This dataset was produced by JAXA - Japan's Aerospace Exploration Agency.
The AW3D30 dataset is expected to be better than SRTM (however not without issues):
At launch time for the dataset - JAXA released a press-statement stating that: "The data set is published based on the elevation data set (5-meter mesh version) of the “World 3D Topographic Data”), which is the most precise global-scale elevation data at this time, and its elevation precision is also at a world-leading level as a 30-meter mesh version."
Dataset Overview
- Spatial Coverage: Global - coverage is most of land bodies on earth
- Resolution: 1 arc-second (30m)
- Data-type: Digital Surface Model (DSM)
- Coordinate system: Geographic WGS (EPSG:4326)
- Version: WindPRO holds the 2.1 version of the AW3D30 data
- Vertical Accuracy: See below - section 'Accuracy of AW3D30'.
Recommendations
The AW3D30 dataset is very interesting for wind-resource application due to its accuracy and resolution. However, since the dataset was obtained using optical sensors, then many of the satellite images used in the DEM-creation could be covered with clouds or have low contrast – such as ice and snow. Such areas are problematic for an optical-stereography approach. If such problematic errors exists in the imagery, then these cloud or snow/ice areas have been void filled with data from alternative sources – using a data from SRTM3, DEM10 data in Japan or PRISM DSM. As such, the EMD recommendation is currently to use the AW3D30 dataset as a supplement to other sources such as SRTM1 or other global/regional datasets - and to validate if any void-fills have been carried out in your model domain.
The AW3D30 quality mask is available from within the windPRO area object, and we recommend that you inspect it (look for void-filled areas) in the following way:
1. Open your windPRO project and create a new area-object
2. Set purpose to 'Roughness map based on closed lines'
3. Go to the - 'Import/Export' tab - and press the 'online-data' button
4. Choose the 'AW3D30 Elevation Model - Quality Mask' - and press 'OK' to load the data
5. Go-to the 'presentation' tab - and set checkmark in 'Transparent' for the 'Boundary Line' - then press 'OK'
6. Right-click the area-object icon - remove the 'edit-mode' and choose to 'show legend'
Availability from within windPRO
The DSM-data and the associated quality mask are available directly from within windPRO. The DSM elevation data is delivered as gridded data with a spatial resolution of 1 arc second. As a part of the processing of the DSM elevation data from our EMD server, the user has an option of choosing whether the data downloaded should be received as either raw gridded data or converted to height contour lines. In order to use the data for energy yield calculations, the data must be converted to contours. These data be accessed from the online-services in the following three objects:
- DSM-Data in the line object (with purpose to height contour lines)
- DSM-Data in the elevation grid object
- Quality mask from AW3D30 is accessed from the area-object
Accuracy of AW3D30
As stated above, the AW3D30 is expected to have a world-leading accuracy - however is still expected to have issues - due to its optical origin as well as data-downstream limitations over some regions of the world. Here we link to a few studies - our recommendation is to gain more knowledge on the data accuracy - and currently to use these data as a supplement to SRTM1 - or to use local data - if these are available. In general, digital elevation models are evaluated using ground control points (GCP’s). For the AW3D30, several evaluations exist:
- A preliminary study by Tadono et al showed that the AW3D30 dataset has a vertical accuracy of 4.40m (RMSE) and a bias of -0.44m – considering 5121 independent ground control points (GCP’s) with global distribution. The same study reported that SRTM v. 3 had a RMSE of 7.50m and a -1.00m bias at the same locations.
- A study by EMD at 1145 turbine locations (see pages 18 and 20). Here, high-quality LiDAR DTM's were available and were compared to AW3D30 and other DEM-sources. The analysis showed that the AW3D30 dataset had a vertical accuracy (RMSE) of 3.7m and a bias of 0.7m. Using SRTM1 at the same sites showed a lower RMSE of 2.0m but with a larger bias of -1.6m. The results are differing quite much from country to country (see the table on page 20).
- The above study by EMD also looked at the impact at the annual energy production at 766 turbine locations (see table on page 21). Here AW3D30 came out with best when comparing to 'ground-truth' values (bias at -0.2% and std.dev. at 0.6%) - while SRTM1 is at bias -0.9% and std.dev. 1.4%.
- A study by Takaku et al, based on 4628 GCP's showed that AW3D30 had a RMSE at 3.28 m and an average bias of -0.30m.
- A study by Santilland and Makinano-Santillan compared performance of AW3D30, SRTM and ASTER at 274 ground control points in the Philippines. They showed that AW3D30 had the lowest RMSE at 5.7m with SRTM at 8.3m and ASTER-GDEM at 12.0m.
- A study by Becek et al "On the Vertical Accuracy of the ALOS World 3D-30m Digital Elevation Model" did conclude that the AW3D30 data is, in terms of accuarcy, comparable to a commercial product such as WorldDEM (tm). The paper is available in a open-access, pre-print version - (here). The peer-reviewed paper is aviailable from the journal "Remote Sensing Letters" - (here)
Acknowledgement, Credits and License
Acknowledgements:
- JAXA is acknowledged for the development and release of this free dataset - and thus for aiding the development of renewable energy and wind energy in particular.
- Integration of this dataset into EMD services was co-supported through the InnoWind project (www.innowind.dk) which is co-funded by the Danish Innovation Fund
Copyright: The Alos World 3D 30m data is provided by JAXA which also holds copyright of the data.
License: When using the AW3D30 data, you are bound by the JAXA licensing conditions, which makes the dataset available and ready-to-use with no charge under the following conditions:
- When you provide or publish any products and services to a third party using this dataset, you are kindly requested to display that the original data is provided by JAXA.
- When you publish the product(s) using this dataset, you are kindly requested to show the copyright (©JAXA) and the source of data.
- JAXA does not guarantee the quality and reliability of this dataset and JAXA assume no responsibility whatsoever for any direct or indirect damage and loss caused by use of this dataset. Also, JAXA will not be responsible for any damages of users due to changing, deleting or terminating the provision of this dataset. We (EMD) recommend the following statement to comply with the JAXA conditions:
Acknowledgement: Contains AW3D30 elevation data from the JAXA - Japan Aerospace Exploration Agency (©JAXA). Data processing and distribution through EMD and windPRO.
External Links
- More information on ALOS DSM project is available at http://www.eorc.jaxa.jp/ALOS/en/about/about_index.htm
- AW3D30 - Product description from JAXA (pdf-format) - here
- A commercial, higher resolution (5m) version of the AW3D30 dataset is available from NTT-data and Restec - see https://www.aw3d.jp/en/
