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

Contact us: data at  |  Disclaimer: Please read
Product pages » Woody Extent and Overstorey Foliage Projective Cover (FPC)

Woody Extent and Overstorey Foliage Projective Cover (FPC)

Last modified by Peter Scarth on 2012/03/23 17:11

Woody Extent and Overstorey Foliage Projective Cover (FPC)

Statewide Landcover and Trees Study (SLATS) 2009, 2009 Woody Extent and Overstorey Foliage Projective Cover (FPC), Landsat SLATS Scene Series (Code:ANZQL0132002906)


Foliage projective cover (FPC) is the percentage of ground area occupied by the vertical projection of foliage. The SLATS FPC mapping involves an automated decision tree classification technique with a nominal accuracy of 85%. Landsat image dates used in the classification were all dry season (May to October). The field data used to calibrate the imagery / FPC relationship was mostly collected over the 1996-1999 period. A number of corrections were also applied including topographic, cloud, cloud shadow and water removal, and regrowth. This product has more precise FPC predictions than previous versions. Limitations of the product can include regrowth areas classified as non-woody, woody clearing classified as woody, and errors due to topographic effects. The analysis has been completed on each of 87 satellite scenes. These have been joined together in the MGA zone applying to the majority of each scene to give three mosaic products. 

Future development

2010 release in July 2012.



The SLATS project was founded in 1995 to investigate the woody vegetation cover and extent of land clearing in Queensland. The project uses Landsat TM and ETM+ satellite imagery acquired from Geoscience Australia. This imagery has a nominal ground resolution of 30 metres. The imagery is orthorectified to 25m spatial resolution and converted to reflectance units based on calibration results. It is standardised for atmospheric and directional variations between dates and scenes using an empirical radiometric correction (Danaher 2004).


The SLATS FPC mapping uses an automated decision tree classification based on a time series of Landsat images. The classification used dry-season (May-October inclusive) image dates from 1986 to 2007. The use of multi-date imagery helps remove some of the seasonal effects (e.g. green herbaceous FPC) present in any individual image date, producing a longer-term view of woody FPC. Decision thresholds were optimised using a genetic algorithm on a field derived training dataset collected over the 1996-1999 period. Further details are provided in Armston et al. (2008) and Armston et al. (2004). The final classification and model had a Kappa statistic of 85.12%. An auxiliary code layer image was created to identify the method used to calculate the FPC value for each pixel.

Water, topographic shadow, incident angle, cloud and cloud shadow masks were created for each scene using spectral classifications and indices as well as visual interpretation and were applied to each individual date. A multi-date watermask was used to remove persistent areas of water that had been incorrectly classified as woody. The topographic shadow and incident angle masks were used to remove commission errors associated with the water mask over steep terrain.

A crop mask was applied to cropping areas that were incorrectly classified as woody and an additional regrowth correction has been applied to correct areas were regrowth has been misclassified.


The final woody extent classification model had a Kappa statistic of 85.12%. Comparison of overstorey FPC estimates with independent field estimates of perennial FPC acquired for a range of vegetation types over Queensland, show excellent agreement (y = 1.017x - 0.305, r2=0.84, RMSE=8.95, N=47). Independent regional scale comparisons with airborne Lidar and MODIS estimates of FPC are forthcoming.


1. Commission Error 

Some regrowth areas have been classified as non-woody, usually due to multiple clearing/regrowth events in the time series causing a low robust minimum and high standard error. Some areas of tree crops have been classified as nonwoody based on the classifications used from the QLUMP 1999 data. A regrowth correction that works by finding the optimal split in the time-series using a reverse-timeseries analysis t-test algorithm, has been used to correct areas classified as plantation by the QLUMP 1999 Land Use Classification Project. Additionally, a regrowth correction applied in a similar way to the plantation areas was used to correct areas of the state that were not classified as crop or plantation in the QLUMP 1999 Land Use Classification Project. More advanced analysis of the Landsat time-series aimed at mapping regrowth and other long-term trends is currently underway. 

2. Omission Error

Some areas of woody clearing between 2006 and 2007 imagery have been classified as woody, due to higher than expected 2007 FPC observations resulting in a lower standard error and higher robust minimum. Causes could be regrowth and/or green pasture cover present after the clearing event, or the presence of green foliage on trees that have been pushed over and left. Some areas of Triodia spp. (spinifex) are often classified as woody or have overestimated FPC due to the presence of perennially green understorey foliage. In some cases, where the 2006 image value is significantly different from the predicted value, the individual date 2006 value has been used instead of the predicted. Additionally, some areas of crop are still erroneously classified as woody where the Landsat time series is not representative of the cropping cycle. 

3. Topographic Effects

FPC values can be underestimated on eastern slopes and overestimated on western slopes due to topography and incident sun angle. A Landsat-5 TM and Landsat-7 ETM+ topographic correction based on the incident angle has been implemented and significantly reduces error. Future research is underway to develop a topographic correction for the original reflectance values from Landsat-5 TM and Landsat-7 ETM+ in order to avoid this problem entirely. 

Key specifications


Image value = FPC + 100 (i.e. FPC of 5% = image value of 105). Range is 100-200 which is equivalent to 0-100% FPC.

The pixel values in this data set represent the predicted foliage projective cover (FPC) values for wooded areas (1% FPC (very lightly wooded) to 100% FPC (closed canopy). FPC is the percentage of ground area occupied by the vertical projection of foliage. In the imagery, the FPC values are scaled by a factor of 100 (1% FPC = 101 in the image). Additionally, values erroneously predicted above 100% or below 0% have been classed as 200 and 100 respectively.

Code layer value

0 = Null values

1 = FPC value from the 2007 image 

2 = Predicted value from the robust regression

3 = Non-woody by regression thresholds

4 = Predicted through topographic correction

5 = Predicted through plantation correction

6 = Water in the 2007 image

7 = Persistent water through multi-date watermask

8 = Regrowth correction

9 = Crop correction

10 = Fill in using the 2007 FPC value

11 and onwards = further fill in values using available dates

The code layer image values reflect the latest correction or prediction applied to each pixel. This includes: prediction or classification by thresholds, individual date values, topographically corrected values, regrowth corrected values, and areas filled in by individual date pixels. 


Quantitative validation of the final product showed good agreement with field (r2 0.78) and lidar (r2 0.93) estimates of FPC.

All the data described here has been generated from the analysis of Landsat Thematic Mapper (TM) data, which has a resampled spatial resolution of 25m. The imagery is rectified using control points measured with a differential GPS ensuring a maximum root mean square (RMS) error of 20 metres at these control points. However, it is possible that errors up to ±50 meters occur between these control points. The imagery has been corrected for height displacement using a digital elevation model (DEM) from the Shuttle Radar. It is not recommended that these data sets be used at scales more detailed than 1:100,000.


Tim Danaher, Peter Scarth, John Armston, Lisa Collett, Joanna Kitchen, and Sam Gillingham (2010), Remote Sensing of Tree–Grass Systems: the Eastern Australian Woodlands. In Hill, M. J. and N. P. Hanan, Eds. Ecosystem Function in Savannas: measurement and modeling at landscape to global scales. Boca Raton, CRC Press.

Armston, J. D.; Denham, R. J.; Danaher, T. J.; Scarth, P. F. & Moffiet, T. N. (2009), 'Prediction and validation of foliage projective cover from Landsat-5 TM and Landsat-7 ETM+ imagery', Journal of Applied Remote Sensing 3(1), 033540--28.

de Vries C, Danaher T, Denham R, Scarth P, Phinn S (2007) An operational radiometric calibration procedure for the Landsat sensors based on pseudo-invariant target sites. Remote Sensing of Environment 107, 414-429.

Kitchen J, Armston J, Clark A, Danaher T and Scarth P (2010) Operational use of annual Landsat-5 TM and Landsat-7 ETM+ image time-series for mapping wooded extent and foliage projective cover in north-eastern Australia. In '15th Australian Remote Sensing and Photogrammetry Conference'. Alice Springs. (Eds Sparrow, B and Bhalia, G)

Scarth P, Armston J, Danaher T (2008) On the relationship between crown cover, foliage projective cover and leaf area index In '14th Australian Remote Sensing and Photogrammetry Conference'. Darwin. (Eds E Edwards and R Bartolo)

Created by Peter Scarth on 2012/03/23 17:11

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