Authors / CoAuthors
Mueller N. | Roberts D. | Emma Ai | Miller, J. | Jorand, C.
Abstract
<div>The DEA Geometric Median and Median Absolute Deviation products use statistical analyses to provide information on variance in the landscape over a given year. They provide insight into the “average” conditions observed over Australia in a given year, as well as the amount of variability experienced around that average. These products are useful for monitoring change detection, such as from cropping, urban expansion or burnt area mapping. </div><div><br></div><div>Satellite imagery allows us to observe the Earth with significant accuracy and detail. However, missing data — such as gaps caused by cloud cover — can make it difficult to create a complete image. In order to produce a single, complete view of a certain area, satellite data must be consolidated by stacking measurements from different points in time to create a composite image. </div><div><br></div><div>The Digital Earth Australia GeoMAD (Geometric Median and Median Absolute Deviation) data product is a cloud-free composite of satellite data compiled annually over each calendar year. </div><div><br></div><div>Large-scale image composites are increasingly important for a variety of applications such as land cover mapping, change detection, and the generation of high-quality data to parameterise and validate bio-physical and geophysical models. A number of compositing methodologies are being used in remote sensing in general, however, challenges still exist. These challenges include mitigating against boundary artifacts due to mosaicking scenes from different epochs ensuring spatial regularity across the mosaic image and maintaining the spectral relationship between bands. </div><div><br></div><div>The creation of good composite images is especially important due to the opening of the United States Geological Survey’s Landsat archive. The greater availability of satellite imagery has resulted in demand to provide large regional mosaics that are representative of conditions over specific time periods while also being free of clouds and other unwanted visual noise. One approach is to ‘stitch together’ multiple selected high-quality images. Another is to create mosaics in which pixels from a time series of observations are combined (using an algorithm). This ‘pixel composite’ approach to mosaic generation provides more consistent results than with stitching high-quality images due to the improved colour balance created by combining one-by-one pixel-representative images. Another strength of pixel-based composites is their ability to be automated, hence enabling their use in large data collections and time series datasets. </div><div><br></div><div>The DEA GeoMAD product can be used for seeing how an area of land usually looks rather than only viewing it at a single point in time. Hence you can assess the land cover and land use on a general basis rather than at a specific date. It can also be used to assess how much an area changes over time. You will notice areas like bare rock that are very stable versus those like cropping areas that change dramatically. </div><div><br></div><div>The DEA GeoMAD product combines the Geometric Median and the Median Absolute Deviation algorithms in a single package. The Geometric Median output provides information on the general conditions of the landscape for a given year. Meanwhile the Median Absolute Deviation output provides information on how the landscape is changing in the same year. </div><div><br></div>
Product Type
dataset
eCat Id
149305
Contact for the resource
Resource provider
Point of contact
Cnr Jerrabomberra Ave and Hindmarsh Dr GPO Box 378
Canberra
ACT
2601
Australia
Point of contact
Digital Object Identifier
Keywords
- ( Project )
-
- DEA – Digital Earth Australia
-
- Satellite Images
-
- Geomedian
-
- Geometric Median
-
- Median Absolute Deviation
- theme.ANZRC Fields of Research.rdf
-
- Earth Sciences
-
- Published_External
Publication Date
2024-08-08T23:09:20
Creation Date
2024-03-06T01:00:00
Security Constraints
Legal Constraints
Status
completed
Purpose
This product provides statistical tools that utilise DEA’s time series Earth observation data to provide annual images of general conditions and how much an area changes in a given year. The geometric median part of the product provides an 'average' cloud-free image over the given year. The geometric median image is calculated with a multi-dimensional median, using all the spectral measurements from the satellite imagery at the same time in order to maintain the relationships between the measurements. The median absolute deviation part of the product uses three measures of variance, each of which provides a 'second order' high dimensional statistical composite for the given year. The three variance measures show how much an area varies from the 'average' in terms of 'distance' based on factors such as brightness and spectra. The three variance measures are: Euclidean distance (EMAD), Cosine (spectral) distance (SMAD), and Bray Curtis dissimilarity (BCMAD). They provide information on variance in the landscape over the given year and are useful for change detection applications.The GeoMAD product is useful for the following: - Land cover mapping.- Change detection and classification (such as for burn-area mapping, crop mapping, and urban area mapping).- General variance and change so that it can be used in machine learning for change detection. - Environmental monitoring.
Maintenance Information
annually
Topic Category
geoscientificInformation
Series Information
Lineage
<div>The GeoMAD is derived from Landsat surface reflectance data. The data are masked for clouds and shadows to increase clarity and ensure the best data is used in the median calculation. </div><div><br></div><div>The three MAD layers of the GeoMAD are calculated by computing the multidimensional distance between each observation in a time series of multispectral (or higher dimensionality such as hyperspectral) satellite imagery versus the multidimensional median of the time series. The median used for this calculation is the geometric median corresponding to the time series. </div><div><br></div><div>The GeoMAD is calculated over annual time periods on Earth observations from a single sensor by default (such as the annual time series of Landsat 8 observations); however it is applicable to multi-sensor time series of any length that computing resources can support. </div><div><br></div><div>For the purposes of the default DEA product, GeoMADs are computed per calendar year per sensor for Landsat 5, Landsat 7, and Landsat 8 until 2021. GeoMADs are computed from combined sensors Landsat 8 and Landsat 9 from 2022 onwards. GeoMADs are computed using terrain-illumination-corrected surface reflectance data (Analysis Ready Data).</div><div><br></div><div>For details of the scientific algorithms implemented, see:</div><div>Roberts, D., Mueller, N., & McIntyre, A. (2017). High-dimensional pixel composites from earth observation time series. IEEE Transactions on Geoscience and Remote Sensing, 55(11), 6254-6264.</div><div>Roberts, D., Dunn, B., & Mueller, N. (2018). Open data cube products using high-dimensional statistics of time series. In IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium (pp. 8647-8650).</div><div><br></div>
Parent Information
Extents
[-44, -9, 112, 154]
Reference System
Spatial Resolution
Service Information
Associations
Downloads and Links
Source Information