Scheduled maintenance: 8-10 am ACT/NSW each Wednesday. Expect shutdowns and restarts.

Contact us: data at  |  Disclaimer: Please read
Outreach » Nationally coordinated approach to ground cover mapping

Nationally coordinated approach to ground cover mapping

Last modified by Matt Paget on 2013/06/17 18:43

Up-scaling: Report from the breakout group


During the workshop entitled “Nationally coordinated approach to ground cover mapping’ held in Canberra on Monday and Tuesday 23-24 November 2009, it was agreed that there would be four ‘break-out’ working groups that would document discussion from the workshop and present a way forward.  The four groups are:

(1)   Implementing MODIS-based fractional cover mapping operationally for the Nation;

(2)   Field Data Collection;

(3)   Up-scaling; and

(4)   Monitoring Land Management Practices with Remote Sensing.


Most members of the workshop were mapped into one of the four ‘break-out’ working groups, with some individuals mapped into two of the four groups.  Members of the up-scaling break-out group (listed alphabetically by surname) who contributed to this document follow:

Tim Danaher (NSW Department of Environment, Climate Change and Water);

Juan-Pablo Guerschman (CSIRO Land and Water);

Leo Lymburner (Geoscience Australia);

Tim McVicar (CSIRO Land and Water);

Lucy Randall (Bureau of Rural Science); and

Michael Schmidt (QLD Department of Environment and Resource Management).


Working Definition: What is Up-scaling?


Up-scaling is a way of taking information based on observations that have a small spatial domain (or temporal domain) and extending the spatial domain to a larger geographic area (or time period).  For example, a relationship between field observations of fractional cover and Landsat resolution data (25 m pixel size for images that are ~180 km length east-west and north-south, so covering an area of ~32,400 km2), maybe up-scaled to the entire Murray-Darling Basin (1,000,000 km2 using either 250 m, 500 m or 1,000 m pixel size MODIS data, depending on the spectral bands).  In the above example the different resolutions (and hence corresponding spatial extents) of the remotely sensed imagery is clearly seen, and crossing a resolution (or scale) boundary is a key feature of up-scaling and is what distinguishes from aggregation (which is the process of adding data from smaller sub-regions – for example crop yield data may be know for each SLA which are then added to provide a State-wide or National estimate).  Of course the dates of field work and image acquisitions can not be too far apart or else the vegetation characteristic being measured and estimated remotely may have changed – for example the crop may have started to senesced or more dramatically may have been harvested.


Remotely sensed based examples of up-scaling?


The following schematic shows how an ‘up-scaling’ methodology can be used to extend the spatial domain of measurements initially made over a much smaller area than the final relationships are used for.  In this example, 1 m2 field observations of Leaf Area Index (LAI) in homogeneous herbaceous (i.e., cropping and grazing) fields were destructively sampled, and Step 1 [Tech Report 96.14 McVicar, et al., 1996a] involved related this site data to Landsat TM data, this enabled us to produce a Landsat TM based LAI (Leaf Area Index) image.  Step 2 [Tech Report 96.15 McVicar, et al., 1996b] involved developing relationships between the AVHRR data and Landsat TM based LAI image. The AVHRR LAI estimate could then be estimated over the entire Murray-Darling Basin.


A similar methodology has been implemented in mixed tree-grass grazing systems (i.e., tropical savannah) to evaluate MODIS for groundcover and biomass/feed availability estimations in such environments (Milne et al., 2007).  In this example the SLATS ‘star-transect’ field layout was used to collect observational data which was then related to the Landsat scale imagery (centred on Charters Towers and the Brisbane Valley), which then has the potential to be used to transfer the information to the MODIS resolution across all of Queensland [Milne, et al., 2007].



Figure 1. Schematic illustrating a remotely-sensed ‘up-scaling’ method to spatially extend data beyond the isolated field measurement sites.


In another recent example [Guerschman, et al., 2009] linked field level spectral radiometric observations (hyperspectral field measurements made with an instrument called an ASD) to Hyperion 30 m resolution (similar to Landsat) hyperspectral satellite data (Landsat is broad-band, only having 6 reflective bands, where as Hyperion has ~150 reflective bands).  End-members relating to the pure components of photosynthetically active (or green) vegetation, senescent (or non-photosynthetically active or yellow) vegetation and bare soil could be distinguished in the both the hyperspectral data (i.e., the ASD and the Hyperion image).  However, to up-scale to the MODIS resolution as this is also a broadband instrument like Landsat (both do not have all the hyperspectral bands of the ASD and Hyperion) they had to develop spectral surrogates using an optimally selected set of MODIS bands to capture the information content in the hyperspectral imagery.  Doing so enabled them to unmix two MODIS indices to map the fraction of photosynthetically active vegetation, senescent vegetation and bare soil at the MODIS resolution.


What is needed to acquire suitable data for up-scaling ground cover?


The members of the up-scaling breakout group recommend, based on discussions held by all participants of the workshop, that the following protocols be implemented to ensure successful up-scaling strategy. 


1)      The SLATS ‘star-transect’ field layout be used to collect field data – of course these locations be accurately located using differential GPS technology.

2)      Field crews in each State be trained by the staff from the Queensland Department of Environment and Resource Management, and any modifications to the SLATS field technique is that additional data are collected.  In other words, the SLATS field technique specifications are the agreed minimum and all data must be collected, if other agencies wish to collect additional data this is up to them.  Hopefully the cal/val breakout working group will make a similar recommendation.

3)      Field data be stored in a (spatially enabled) relational database that is compatible with the current Queensland database (PostGIS - to quickly generate a National database.

4)      Field sites be located in both homogenous herbaceous land-covers and in more heterogeneous tree-shrub-grass mixed systems; within a field site the land-cover needs to be fairly homogenous and needs to be representative of wider conditions.  The field data needs to be ‘industrial strength’ suitable for a national monitoring system, not just to provide calibration opportunities of remotely sensed imagery which tend to shy away from heterogeneous field sites.

5)      To assess the impact of different soil colours some additional Hyperion imagery to perform the backbone of the spectral transformation from ASD through Hyperion to MODIS needs to be purchased to improve the MODIS fractional cover product based on Guerschman et al. [2009].

6)      The field work strategy should be designed to allow for coincident SLATS ‘star-transect’ field collection and Hyperion data acquisition over a wide range of soil colours i.e. white kaolinite clays, black cracking clays, red desert sands and yellow earths.  Hyperion is pointable hyperspectral sensor that acquires visible near-infrared [(VNIR) 400–1000 nm] and shortwave infrared [(SWIR) 900–2500 nm] spectra with 155 stable bands yet has a narrow swath width (7.65-km) so exact site locations need to be specified prior to tasking [Datt, et al., 2003].  The overpass is coincident with Landsat-7 data.

7)      To determine the impact of soil colour some coincident ASD spectra will also need to be acquired to compliment the SLATS ‘star-transect’ field collection method.

8)      Given the large number of SLATS ‘star-transect’ field validation data collected for Queensland, the emphasis will be on collecting such data for the various natural and agricultural systems in other states, noting that the black clays of central-western Queensland will likely need some intensive field data collected there.

9)      The technical details on the optimal upscaling algorithm are yet to be determined.


Finally, members of the up-scaling break-out group note that the data (both field and satellite) and associated analysis that enable upscaling of field observations will contribute to both:

1. the validation of an operational 500 m fractional cover product (a CfoC goal); and

2. the on-going improvement of a 500 m fractional cover product (a TERN goal).


It is anticipated that resources from both funding sources (i.e., CfoC and TERN) will be used to support a more comprehensive field work and data analysis campaign leading to on-going improvements i.e. version 1, version 2 etc of CSIRO's 500m fractional cover product.




Datt, B., T. R. McVicar, T. G. Van Niel, D. L. B. Jupp, and J. S. Pearlman (2003), Pre-processing EO-1 Hyperion hyperspectral data to support the application of agricultural indices, IEEE Trans. Geosci. Remote Sens., 41, 1246-1259.

Guerschman, J. P., M. J. Hill, L. J. Renzullo, D. J. Barrett, A. S. Marks, and E. J. Botha (2009), Estimating fractional cover of photosynthetic vegetation, non-photosynthetic vegetation and bare soil in the Australian tropical savanna region upscaling the EO-1 Hyperion and MODIS sensors, Remote Sens. Environ., 113, 928-945.

McVicar, T. R., D. L. B. Jupp, P. H. Reece, and N. A. Williams (1996a), Relating LANDSAT TM vegetation indices to in situ leaf area index measurements, Technical Memorandum, 80 pp, CSIRO, Division of Water Resources, Canberra, ACT 

McVicar, T. R., D. L. B. Jupp, and N. A. Williams (1996b), Relating AVHRR vegetation indices to LANDSAT TM leaf area index estimates, Technical Memorandum, 33 pp, CSIRO, Division of Water Resources, Canberra, ACT 

Milne, J., T. J. Danaher, P. Scarth, J. O. Carter, J. Armston, B. Henry, N. Cronin, R. Hassett, G. Stone, P. Williams, R. Denham, M. Byrne, and S. Gillingham (2007), Evaluation of MODIS for groundcover and biomass/feed availability estimates in tropical savannah systems final report to Meat & Livestock Australia (Project Code NBP.330), Queensland Department of Natural Resources and Water, Climate Impacts and Natural Resource Systems, Indooroopilly, Brisbane 


Created by Peter Scarth on 2012/10/19 16:51

This wiki is licensed under a Creative Commons 2.0 license
XWiki Enterprise 4.5.3 - Documentation