Fractional cover - MODIS, Monthly median composites, Australia coverage

Fractional cover - MODIS, Monthly median composites, Australia coverage

Figure 1: Fractional cover monthly median composite for January 2014

Link to the data

Data licence and Access rights

Abstract or Summary

Each image is a composite of all MODIS fractional cover images for the month. The input images are version 2.2 or version 3.0.1 of the CSIRO fractional cover product (Guerschman et.al. 2009, Guerschman et.al. 2012). The medoid method of Flood (2013) was used to create the composites.

These data are used in support of the NSW DustWatch project, http://www.environment.nsw.gov.au/dustwatch.

Spatial and Temporal extents

Point of contact

Credit [optional]

We thank CSIRO for making the fractional cover data publicly available.

Keywords

There are three main thesauri that AusCover recommends:

1. Global Change Master Directory (http://gcmd.nasa.gov)
2. Climate and Forecast (CF) convention standard names (http://cfconventions.org/standard-names.html).
3. Fields of Research codes (http://www.abs.gov.au/ausstats/abs@.nsf/0/6BB427AB9696C225CA2574180004463E?opendocument).

Data quality

Fractional cover data quality is subject to the data quality data quality constraints of the MODIS Fractional Cover product.

Potential data quality issues arising from the medoid method are described in Flood et al 2013.

Validation status

Validation of the fractional cover data is subject to the validation performed for the MODIS Fractional Cover product.

Validation of the medoid method is described in Flood et al 2013.

Related products

References

Algorithm summary

Updated algorithm information can be found in the README file, ftp://tern-auscover.science.uq.edu.au/aus_modis_monthly_fractional_cover/README.

Three products are provided for each version. These are identified by a three-character product 'stage' code in the file name.

For version 2.2 the stages are:

• ba2 - monthly composites
• ba4 - gap filled composites
• ba5 - a code layer for the gap-filled images.

For version 3.0.1 the stage codes are:

• bb1 - monthly composites
• bb3 - gap filled composites
• bb4 - a code layer for the gap-filled images.

For ba2, bb1, ba4 and bb3 (monthly and gap-filled composites) the following Band labels apply:

• Band 1 = percent bare (BS)
• Band 2 = percent photosynthetic vegetation (PV)
• Band 3 = percent non-photosynthetic vegetation (NPV)
• Null Pixel value = 255

Stages ba2 or bb1: Monthly fractional cover

The medoid method of Flood (2013) was used to create the composites. The medoid has two steps. For each grid cell:

1. Find the median cover value for each layer (BS, PV, and NPV). This is a multi-dimensional median.
2. Find the pixel from the input images that is closest to the multi-dimensional median.

This method has the advantage that the output pixel is one of the input pixels. This is important for the fractional cover as the sum of the percent covers is normally 100, and we wish to retain this in the composites. We compute a medoid where there is at least one non-null pixel from the input images.

For each month number below, the day of year (doy) images listed were used to create the composites:

Stages ba4 or bb3: gap-filled composites

The monthly composites may contain missing values. This is likely for cloudy areas where there are no valid observations within the month. Depending on the application, it may be desirable to have images without data gaps.

The gap-fill algorithm fills null values with a resampled pixel at a coarser resolution. The resolution of the resampled image is made progressively coarser until all pixels are filled. The resampled pixel sizes are the original resolution multiplied by a power of 2. For example, if res is the original pixel size then the resampled pixel sizes are: res*2¹, res*2², res*2³, ..., res*2ⁿ.

While this method is simple, it has the disadvantage that the sum-to-one constraint is lost, i.e. the sum of the percent covers may no longer be 100 where a pixel has been filled.

Stages ba5 or bb4: A code (flag) layer corresponding to the gap-filled composites

The gap-filled code image identifies those pixels in the gap-filled product that had to be filled. A non-zero pixel value represents a filled pixel. The value itself is the resolution of the resampled image used to fill the pixel, expressed as the power of 2 used. For example if the resolution was res*2⁴, then the code value is 4. A value of zero represents a pixel that was not filled. The null value (ocean pixels) are set to 255.

Product version history

Other items [optional]

Metadata history