contact  |  support  |  links
basis function map
We have constructed a basis function map(file and readme) for the TransCom 3 experiment that reflects both geographical and mechanistic elements. The terrestrial and oceanic portions of the world were each broken into 11 separate source/sink regions. The terrestrial boundaries were constructed following certain land use type classifications with some additional smoothing of the boundaries. A full explanation of the methodology can be found through the Boundary Construction section. Below we present both the original unsmoothed boundaries and the final smoothed basis map.
We have included a sea ice mask in the oceanic basis regions. Because of the substantial seasonality in sea ice extent, this imparted seasonal spatial structure to four of the oceanic basis regions. Follow this link to the sea ice mask discussion.
Terrestrial regions, no smoothing applied
terrestrial and oceanic regions with 2 smoothing passes
Constructing initial source/sink region boundaries
The starting point for the boundary construction is the 1 x 1 land cover type classification as used by SiB2. This classification map contains 12 terrestrial vegetation categories and comes from the International Satellite Land Surface Climatology Project (ISLSCP). This data set is derived from the normalized difference vegetation Index (NDVI), a product of AVHRR satellite measurements. The methods used to construct land cover types from satellite data can be found in DeFries and Townshend 1994.
Note that the DeFries and Townshend 1994 paper provides an explanation of a 1 x 1 land cover map with 11 categories. This map was translated into a land cover map with 12 categories - 'wooded grassland' in the DeFries and Townshend 1994 classification are broken in types 6 and 8 in the SiB2 classification map (12 categories) used here.
The 12 vegetation types are as follows:

1. Broadleaf evergreen
2. Broadleaf deciduous
3. Broadleaf and needleleaf
4. Needleleaf trees
5. Needleleaf-deciduous trees
6. Broadleaf trees with groundcover
7. Ground cover (maize optical)
8. Broadleaf shrubs with ground cover
9. Broadleaf shrubs bare soil
10. Tundra, dwarf trees, ground cover
11. Low latitude deserts
12. Broadleaf deciduous trees over wheat cover

The three other categories in the map are: land ice (13), sea ice (14), and ocean (0).
The goal was to generate approximately 10 spatially-simple regions that enclose vegetation of similar seasonal structure and carbon exchange. The goal was to be as objective as possible in the construction of these source/sink regions. A simple rule structure was followed.
Boreal regions included vegetation types 4, 5, and 10
Tropical regions included vegetation types 1, and 6.
Temperate regions included vegetation types 2, 3, 7, 8, 9, 11, 12 and sometimes 6 (this broad classifications can represent tropical - i.e. SE Asia - or temperate - i.e. non-Amazonian Brazil and Western Argentina - vegetation types).
Regions with little or no NPP were included with an adjacent region. For example, The Saharan desert was included with the rest of Africa.
Some subjective divisions were imposed as well. They are as follows:
  • A division at 60 East latitude was imposed to separate Europe from Asia. This division was further extended as a straight line from the Aral Sea to Istanbul. Though some of Europe is boreal when considering vegetation type, this was not separated as such.
  • Africa was split at the equator due to the different phasing of seasonality between hemispheres. The North African region was separated from the Asian regions at the Red Sea.
  • Some high-latitude landmasses were left unclassified: Greenland, Iceland, Novaya Zemyla, Svalbard.
Smoothing
Given the 'ragged' nature of the boundaries derived with the above methodology, a spatial smoother routine was used to generate source/sink regions with smooth, continuous boundaries. The smoothing routine can be run repeatedly, achieving greater amounts of smoothness with a successively larger number of iterations.
References
DeFries, R. S. and J. R. G. Townshend, 1994, "NDVI-derived land cover classification at a global scale," International Journal of Remote Sensing, 15: 3567-3586.
 
site design by Micek Design
All Images on this Website are the property of the stock.xchng, all rights reserved.
Copyright 2011, Arizona State University all rights reserved.