Personal tools
Home » Members » perry » Extracting vector basin statistics from raster-based data.
Document Actions

Extracting vector basin statistics from raster-based data.

by Matthew Perry last modified 18-07-2006 13:32

A number of important watershed charachteristics such as climate, slope, land use and land cover are represented as a raster grid. Using a software packaged called StarSpan, we were able to develop a technique for automated exctraction of basin-level statistics from these raster grids. This document also describes the processing steps to create the raster inputs.

Prepping the input rasters

Temperature

Fertilizer

(See Mapping Fertilizer and Pesticide Distribution)

Pesticide 

(See Mapping Fertilizer and Pesticide Distribution)

Water Balance

Vegetation

Elevation

Slope

Soil Order

Soil Erodibility

Wilderness

Tree Cover

Percent Irrigation

Humidity

Glaciers

Lakes and Water Bodies

Human Induced Soil Degradation

Potential Evapotranspiration

Precipitation

RUSLE

(see Modelling Global Erosion Potential by River Basin - A GIS based RUSLE model)

Protected Areas

Human Footprint

Resevoir Depth



StarSpan software overview

  Specialized GIS Utility written at UCDavis to extract raster summaries from polygon layers.

 

Please cite the software in published papers as follows (modifying the formatting as appropriate):

Rueda, C.A., Greenberg, J.A., and Ustin, S.L.  StarSpan: A  Tool for Fast Selective Pixel Extraction from Remotely Sensed Data. (2005). Center for Spatial Technologies and Remote Sensing (CSTARS), University of California at Davis, Davis, CA.

 

Summary of a continuous raster layer by watershed; Temperature

 

river

numPixels

avg_Band1

min_Band1

max_Band1

stdev_Band1

Alabama

385

16.75

13.24

19.22

1.17

Alazeya

622

-13.29

-14.89

-12.53

0.43

Altamaha

129

17.47

15.44

19.44

0.97

Amazon

16811

24.98

0.97

27.86

3.55

 

 

Summary of a discrete raster layer (ie categorical) by watershed; Vegetation Class

 

FID

class

count

0

0

20075

0

1

1393

0

3

418

0

5

180

0

6

18

0

7

846332

     0

8

56641

0

9

973

0

10

17001

0

15

102

0

16

43453

1

0

5394

1

1

3649

1

2

26

1

3

1671

1

5

60

1

6

41

1

7

449868

1

8

113827

1

9

1362

1

10

79181

1

16

16858

 

By taking the sum total count for each feature(FID), you can determine the percentage of each veg class covering the watershed.


Setup

The raster_stats.py script on cabrillo (in /home/perry/scripts/) does all work flow while starspan and postgis do the heavy lifting.


raster_stats.py requires a very specific data structure in order to work properly:

  • Create and define in the script an "outputs" directory to hold temp files and an "outshp" directory to store the output shapefiles.
  • A raster directory that stores each of your rasters in a data structure like "./rastername/continent_rastername.tif"
  • A vector basins directory with 2 shpfiles: continent_basins.shp (polygon basins) and continent_pours.shp (point pours)
  • A postgres database in which to store and join all the data, must have postgis installed as well as discrete_pivot.sql
  • All of the rasters must be configured as either a discrete or continuous raster in the script
  • All of the config options are hardcoded in the script.


Once all that is in place, simply run

 ./raster_stats.py continent database





Built with Plone