Forest probability circa 2000
and forest cover clearing between 1990s and 2000s

----------------------------------------------------------------------------

8-bit data
47144 pixels by 30888 lines
Data format: GeoTiff. This data is importable to any image viewing software. 

 Projection                     :    	Sinusoidal
 Earth Ellipsoid                :    	Sphere, rad 6370997 m
 Upper Left Corner (m)          :    	903459.765 X  833962.86 Y       
 Pixel size (m)                 :    	57 X 57 Y  

----------------------------------------------------------------------------

Values
0-100	forest probability
253	forest clearing between 1990s - 2000s
250	water
254,255	no data 


----------------------------------------------------------------------------

Basin-wide, multi-spectral Landsat composites were created for nominal 1990
and 2000 date epochs. A total of 475 Landsat acquisitions were obtained over
the study area and segregated by year into 1990s (1986  1996) and 2000s
(>1996  2003) epochal datasets.  The Landsat sensor has a nominal at nadir
pixel size of 28.5 x 28.5 meters.  In this study, Landsat pixels from bands
4, 5, 6, and 7 were re-sampled to 57 by 57 meters and geo-referenced to a
common GeoCover base.  For each path/row and time period, a single cloud-free,
radiometrically normalized composite was created from multiple acquisitions.
Each scene was processed individually and normalized for inter-scene
sun-sensor-geometry and atmospheric variation.  Quality assessment (QA)
values were assigned to each pixel based on likelihood of cloud cover,
cloud shadow or haze.  The single best pixel (e.g. the least likely affected
by cloud, cloud shadow, or haze) in each epochal composite was then used in
the final image composites and, subsequently, for classifying forest cover
and loss.   

Forest cover is defined per pixel as a continuous variable between 0 and 100
quantifying the probability that the given pixel is represented on the ground
by dense forest cover. Forest cover clearing (value 253) is defined as
complete removal of the forest overstory.  Forest cover was characterized
using classification and regression tree methods employing training data
derived from the MODerate Resolution Imaging Spectroradiometer (MODIS)
Vegetation Continuous Fields (VCF) product.  Discretized maps of forest cover
can be created by applying thresholds to the forest probability variable and
labeling each class accordingly.  For example, forest likelihood values less
than 50 generally indicate areas of non-forest and values 50 or greater
generally indicate areas of forest cover.  

----------------------------------------------------------------------------

This data is a part of the Congo Basin Forest Monitoring Project and is a
contribution to the Central African Regional Program for the Environment
(CARPE) and is provided by the Geographic Information Science Center of
Excellence (GIScCE), South Dakota State University

Other participating partners include the University of Maryland, NASA, and
the United States Agency for International Development.

The project's web site at SDSU:
http://globalmonitoring.sdstate.edu/projects/congo

CARPE home
http://carpe.umd.edu

----------------------------------------------------------------------------

Provided data are available for use for valid scientific, conservation,
and educational purposes as long as proper citations are used. We ask that
you credit the Congo Basin Forest Monitoring data as follows: 

Hansen, M., Roy, D., Lindquist, E., Justice, C., Altstatt, A., 2008.
A method for integrating MODIS and Landsat data for systematic monitoring
of forest cover and change in Central Africa, Remote Sensing of Environment,
vol. 112, pp. 2495-2513.

Lindquist, E., Hansen, M., Roy, D.P., Justice, C.O., The suitability of
decadal image data sets for mapping tropical forest cover change in the
Democratic Republic of Congo: implications for the mid-decadal global land
survey, International Journal of Remote Sensing, vol. 29, pp. 7269-7275. 

----------------------------------------------------------------------------

For further information, please contact:

Dr. Matthew Hansen
Geographic Information Science Center of Excellence - SDSU
Phone: (605) 688-6848
Matthew.Hansen@sdstate.edu
