Fractional Cover

Other Names:

fCover

Description

Fractional cover refers to estimating the proportion of an area that is covered by each member of a pre-defined set vegetation or land cover types. For mapping fractional cover, the proportions of the different classes should sum to one. In terms of remote sensing, the area being considered is generally a pixel (although fractional cover estimation can be applied to object-based image analysis), and the estimation of fractional cover is considered a type of “spectral unmixing” or sub-pixel classification.

There are many different techniques for estimating vegetation fractional cover from remotely-sensed data. Fernandes et al. (2004) and Scanlon et al. (2002) described the following subpixel mapping techniques applied to estimating fractional cover:

  • Conventional “Hard” Classification – “Hard” classes are rigidly defined in terms of their composition and cover, and are generally found in land cover maps. For example, a land cover map may have a Douglas-fir class, a ponderosa pine class, and a mixed Douglas-fir/ponderosa pine class. The implicit assumption here is that the “pure” classes have very little cover of the other species, and that the mixed class has roughly equal proportions of both. With such an approach, techniques like supervised classification or unsupervised classification are employed to try to directly classify or map the “hard” classes.
  • Linear Modeling – linear modeling approaches attempt to relate reflectance data recorded by a sensor to field measurements of fractional cover using linear regression techniques. Such approaches may correlate field-measured fractional cover with sensor reflectance bands, or to vegetation indices like NDVI.
  • Spectral Unmixing Models – spectral unmixing is based on the theory that the observed reflectance of a pixel (or object) is a function of the proportion of the different cover types within the pixel as well as other factors like the relative brightness of the cover types. If spectral signatures are known from “pure” representations of each cover type, then the proportion of the cover types in the pixel can be figured out.
  • Artificial Neural Networks (ANN) – ANN are networks of simple processes, decisions, or algorithms applied to data that are good at analyzing data from non-linear and non-parametric systems. With remote sensing, ANN has typically been used for “hard” classifications, but it has also been used to derive fractional cover (e.g., Foody et al. 1996, Moody et al. 1996).
  • Physical Models – Physical models use principles of how light energy is absorbed or reflected from different surfaces to estimate fractional cover. Biophysical models incorporate parameters related to how light interacts with processes like photosynthesis, evapotranspiration, stress, and decay of plant material. Geometric-optical models consider how the spectral reflectance of different surfaces interact to influence the values recorded by a sensor. Physical models require making assumptions of how systems reflect light and estimating or measuring model parameters to achieve accurate results. Many physical models require estimation of a bi-directional reflectance distribution function – a function that defines how much light is reflected from a surface depending not only on the surface’s reflectance properties, but also on the angles of the incoming and reflected light. Because of this, physical model approaches to estimation fractional cover have, to date, been applied mostly to fine-scale images over small areas.

Similar Methods

Output

Fractional cover estimation requires that the classes to be mapped be defined ahead of time. Accordingly, the output of fractional cover estimation is an estimate of the proportion (i.e., percentage of the total) of each class for each pixel (or object).

Successful Rangeland Uses

Fractional cover mapping has been applied in a number of rangeland systems, but mostly as test cases for research, and not as management applications. In many cases, these were experimental techniques that may or may not be suitable for deriving data layers for rangeland management.

  • Chopping et al. (2008) described a method of using bi-directional reflectance distribution functions of MISR imagery for predicting fractional cover of shrubs in desert grasslands.
  • Marsett et al. (2006) described a method for converting data from Landsat TM imagery into estimates of total vegetation cover. This is probably the simplest rangeland application of fractional cover, and could produce management-ready information with a moderate amount of effort.
  • Scanlon et al. (2000) used a spectral unmixing model with NDVI and rainfall data to predict fractional cover of vegetation in southern Africa.

Application References

  • Chopping, M., L. Su, A. Rango, J. V. Martonchik, D. P. C. Peters, and A. S. Laliberte. 2008. Remote sensing of woody shrub cover in desert grasslands using MISR with a geometric-optical canopy reflectance model. Remote Sensing of Environment 112:19-34.
  • Marsett, R. C., J. Qi, P. Heilman, S. H. Beidenbender, M. C. Watson, S. Amer, M. Weltz, D. Goodrich, and R. Marsett. 2006. Remote sensing for grassland management in the arid southwest. Rangeland Ecology and Management 59:530-540.
  • Scanlon, T. M., J. D. Albertson, K. K. Caylor, and C. A. Williams. 2002. Determining land surface fractional cover from NDVI and rainfall time series for a savanna ecosystem. Remote Sensing of Environment 82:376-388.

Technical References

  • Carlson, T. N., E. M. Perry, and T. J. Schmugge. 1990. Remote estimation of soil moisture availability and fractional vegetation cover for agricultural fields. Agricultural and Forest Meterology 52:45-69.
  • Carlson, T. N., and D. A. Ripley. 1997. On the relation between NDVI, fractional cover, and leaf area index. Remote Sensing of Environment 62:241-252.
  • Fernandes, R., R. H. Fraser, R. Latifovic, J. Cihlar, J. Beaubien, and Y. Du. 2004. Approaches to fractional land cover and continuous field mapping: a comparative assessment over the BOREAS study region. Remote Sensing of Environment 89:234-251.
  • Foody, G.M., Lucas, R.M., Curran, P.J., and M. Honzak. 1996. Estimation of the areal extent of land cover classes that occur at a sub-pixel level. Canadian Journal of Remote Sensing 22:428-432.
  • Moody, A., Gopal, S. and A.H. Strahler. 1996. Artificial neural network response to mixed pixels in coarse-resolution satellite data. Remote Sensing of the Environment 58:329-343.
  • Morsdorf, F., B. Kotz, E. Meier, K. O. Itten, and B. Allgower. 2006. Estimation of LAI and fractional cover from small footprint airborne laser scanning data based on gap fraction. Remote Sensing of Environment 104:50-61.
  • North, P. R. J. 2002. Estimation of fAPAR, LAI, and vegetation fractional cover from ATSR-2 imagery. Remote Sensing of Environment 80:114-121.
  • Olthof, I. and R.H. Fraser. 2007. Mapping northern land cover fractions using Landsat ETM+. Remote Sensing of the Environment 107:496-509.
  • Qi, J., R. Marsett, M. S. Moran, D. Goodrich, P. Heilman, Y. H. Kerr, G. Dedieu, A. Chehboundi, and X. X. Zhang. 2000. Spatial and temporal dynamics of vegetation in the San Pedro River basin area. Agricultural and Forest Meterology 105:55-68.
  • White, M. A., G. P. Asner, R. R. Namani, J. L. Privette, and S. W. Running. 2000. Measuring fractional cover and leaf area index in arid ecosystems: digital camera, radiation transmittance, and laser altimetry methods. Remote Sensing of Environment 74:45-57.

Limitations

One limitation of fractional cover estimates is that all of the classes for which cover proportions are to be estimated must be defined prior to the analysis. This generally is only a problem in cases where you encounter a different cover type that you weren’t expecting. Also, the number of classes that you define and predict fractional cover of can affect the accuracy of the result – especially if the spectral reflectance of those cover types is very similar or highly variable.

Data Inputs

The data inputs for estimating fractional cover will vary depending on the method used.

Software/Hardware Requirements

Like the data inputs, the software and hardware requirements for estimating fractional cover will also vary depending on the methods used. Any attempt to produce fractional cover layers will require remote-sensing/image processing software (e.g., ERDAS Imaging, ENVI). Calculating fractional cover is normally a complex exercise, and all but the most simple implementations of it (e.g., Marsett et al.’s (2006) conversion of SATVI to fractional vegetation cover) will also require at least statistical software and maybe additional specialized software.

Existing datasets

  • The USGS Land Process Distributed Active Archive Center (LPDAAC) serves up a yearly MODIS fractional cover product at 500m resolution called the Vegetation Conversion-Continuous Fields Yearly L3 Global 500m product. This dataset, free to download, contains proportional cover information for life form (proportion of woody vegetation, herbaceous vegetation, or bare ground), leaf type (needleleaf or broadleaf), and leaf longevity (evergreen or deciduous). More information can be found at: https://lpdaac.usgs.gov/products/modis_products_table/mod44b

Comments are closed.