Distribution of Plants on Terrain

Possible ways to describe plant distribution

• Raster: Using a coverage bitmap
  • discretely - type information per element of the elevation grid, or any overlapping grid
  • pro: easy to draw in a paint program, easy to vary density
  • con: doesn't work well for fine boundaries, e.g. along the edge of a road
• Polygons
  • structured - polygonal bounding regions indicate areas of plant coverage
  • pro: more precise placement, works well for crops
  • con: would have to create graphical tool for defining the areas, or rely entirely on existing polygonal data sources
• Points: Placing individual trees, or lines of trees (e.g. along a road or fence)

Producing plant distributions

• Many academic papers cover algorithms which extract plant data from imagery, or simulate ecosystems - see the Plant Papers page for significant papers like Tree detection from aerial imagery (2009)

Sources of Data

• The GlobeLand30 (GLC30) LULC Dataset
  • has 30m coverage globally, released 2014 by China's National Administration of Surveying, Mapping and Geoinformation (NASG)
  • apparently it is online, browsable and downloadable at globallandcover.com
  • "ten major land cover classes": Water bodies, Wetland, Artificial Surfaces, Tundra, Permanent snow and ice, Grass lands, Barren lands, Cultivated land, Shrub lands, Forests
• USGS LULC (land use and land cover) files (see below)
  • includes classification code such as "Herbaceous Rangeland" and "Evergreen Forest Land" which very roughly description vegetation type
  • U.S. only
• USGS Global Land Cover Characterization
  • huge raster files, in a raw format, with 1km coverage of the whole earth
  • has much better classification than LULC, example of a single type: "Needleleaf Forest (Western Hemlock, Sitka Spruce, Douglas Fir)"
  • the main drawback is the very low resolution (1km) which makes it useless for visualization at a human scale
• European Space Agency (ESA) Ionia GlobCover
  • GlobCover Land Cover V2.3 (Dec. 2010) is the highest resolution (300 meters) Global Land Cover product ever produced. Based on ENVISAT MERIS data at full resolution covering 2 periods: December 2004 - June 2006 and January - December 2009.  Labeled according to the UN Land Cover Classification System.
  • Version 2 was a 436 MB GeoTIFF file.
• The GLCF has global vegetation data
  • Tree Cover Continuous Fields
    • based on 1km AVHRR satellite data
  • MODIS Vegetation Continuous Fields
    • percent tree canopy cover, gathered 2000 - 2001, from the MODIS satellite data
• USGS NLCD (National Land Cover Database)
  • NLCD 1992 is a 30-meter US-only land-cover classification (21 classes) based on an unsupervised clustering algorithm run on LandSat data
  • distributed as huge files, raw, 1-byte, in Albers projection, or as GeoTIFF
  • the The Multi-Resolution Land Characteristics (MRLC) Consortium was working on a newer version, NLCD 2001
• WWF Ecoregions is an amazing global map of ecoregions
  • the map data is available (25 MB zipped Shapefile), only requires submitting an email address
• USDA vegetation maps?  7.5-Minute DLG Vegetation Overlay?
• a great many people in the GIS field use polygonal areas with attributes (e.g. in a Shapefile) to describe vegetation coverage
  • unfortunately, there are no standards for the number and meaning of the attributes
  • therefore, there is almost never a consistent way to exploit these data sources, or to exchange data
• OpenTreeMap
  • A set of web software (built on Geoserver, Django, PostGIS, etc.) which provides a front-end to a GIS dataset of individual trees in a city, in use in San Francisco and Philadelphia as of 2011
  • Has a standard set of attributes per tree, including species name, location, height
  • Can export tree lists as KML or CSV files.

Polygonal Data Formats

• Standard GIS formats (ESRI Shape, MapInfo...) are commonly used to represent polygonal data.  However, there are no standards at all for the per-polygon attributes that go along with these datasets, so importing data (e.g. with VTBuilder) requires a manual step of mapping one set of attributes onto another.
• Example of polygonal data with attributes: the GIS Data at the Charles Lathrop Pack Forest
  • a good, free, detailed dataset includes polygonal coverages for soil, lakes, and forest stands
• There are some domain-specific polygon data formats, such as USGS LULC, which define a standard for their attributes and can therefore be understood directly by software.

USGS LULC (land use / land cover) files

• 37 land usage types
  • detailed description of each classification code is described in:
    Anderson, J.R., Hardy, E.E., Roach, J.T., and Witmer, R.E., 1976, A land use and land cover classification system for use with remotely sensed data: USGS Professional Paper 964, 28 p.
• LULC does not contain any information about species or ecosystem, so it's primary usefulness is to describe vegetation density
• for example, it can tell which areas are agricultural, but not which crop is growing there
• the 1:100,000-scale format is a 30' x 60' quadrangle, UTM-based
• can be read by VTBuilder and the the vtdata library
  • it turns out GIRAS (the underlying format of LULC) is not difficult to read
  • however, the map coordinates are in a strange format, "local UTM"
  • how to correctly convert these coordinates to lat-lon or actual UTM is unclear
  • one way is to use the four LULC corners as a guide, defining a "perspective transform"
    • found an algorithm for this in the book Digital Image Warping, pages 52-56
• format description
  • the entire format is documented in USGS Data Users Guide 4 (converted by me to HTML for your reading pleasure)
• As of 2010, classic polygonal LULC is out of favor; most landcover data is now in raster form which is more efficient for high detail and more directly output from remote sensing classification.

General Land Cover Classification

• Besides placing vegetation, land use/cover classification like "pasture" or "pavement" can be use to generate synthetic Geotypical, Ground Cover Texture or Detail Textures.
• Many software tools exist to classify space/aerial imagery
  • Feature Analyst ($1k-$16k depending on license type, ArcGIS plug-in)
  • MaterialMAP from TerraSim ($?) is specifically for the Vis-Sim field (output for VBS2, Steel Beasts, OneSAF OTF, and JSAF CTDB)

Classification of Vegetation

• there is very little standardization
• some people treat classification of vegetation as a subset of general land-use classification (e.g. the USGS and USDA) but this does not well describe either the physical appearance or the scientific description, e.g. species
• one fairly common physical measure of trees in the forestry industry is their radius
  • this is usually stated as "Diameter at Breast Height" (DBH)
  • see Simplified guide to measuring DBH
• "There are some standards, but surprisingly few, and almost always non-visual. Some ecologists have worked on structural vegetation classifications, since it turns out that structure is very important to birds, for instance. But from what I've seen, this is usually for trees only, mostly based on age, "site class" (soils/slope/aspect), and DBH. I've never seen anything which would cross-walk a meadow from Siberia to Florida. You might get part of the way there with a structural description of vegetation form ranges (bounding sphere x age?), plus geographic range (by bioregion?). You could then pick the closest vegetation form from the closest broad bioregion. If you are in a forest in the NE but have no trees for that area, fallback might be trees of same size from a Michigan forest." - Michael Flaxman of Geodesign

Descriptions of Plant Distribution

• GENETICS Automated Vegetation Generation
  • part of an academic project in 2005, describing how to use terrain slope and land cover classification (LCC) to distribute vegetation
  • it was later rolled into the Delta3D engine - for programmers, see the VegetationDecorator class and their tutorial
• ECOSim - Ecosystem Simulation: site is no longer online, but available as a textbook (in French), COQUILLARD P., HILL D. Modélisation et Simulation des Ecosystèmes, 1997
• The VTP software includes distribution of vegetation - see VTBuilder Vegetation Layers

Game Programming:

• "We simply have the artists create a density map for the terrain and then randomly (with a seed) place/ render vegetation depending on the density map. Then between min and max distance we either shrink objects to size 0 or fade them out using alpha. This approach works fine for small stuff like grass or ferns." - Daniel Vogel, Epic Games