Analysis of Resolution and Resampling on gis data Values E. Lynn Usery

Analysis of Resolution and Resampling on GIS Data Values

• E. Lynn Usery

• U.S. Geological Survey

• University of Georgia

• Michael P. Finn

• U.S. Geological Survey

The People Who Did the Work

• Michael P. Finn, Computer Specialist

• Douglas Scheidt, Student Programmer

• Gregory Jaromack, Student Programmer

• Thomas Beard, Cartographic Technician

• Sheila Ruhl, Cartographic Technician

• Morgan Bearden, Cartographic Technician

• John D. Cox, Cartographic Technician

Outline

• Introduction and Objectives

• Study Areas

• GIS Databases for Parameter Extraction

• AGNPS Parameter Generation

• Resolution Effects

• Resampling Effects

• Conclusions

Objectives

• Develop GIS databases as input to Agricultural Non-Point Source (AGNPS) Pollution Model

• Create a tool for generating input, executing the model, and analyzing output

• Determine effects of resolution and resampling

Introduction -- AGNPS

• Operates on a cell basis and is a distributed parameter, event-based model

• Requires 22 input parameters

• Elevation, land cover, and soils data are the base for extraction of input parameters

Study Areas

• Four Watersheds

• Little River, GA
• Piscola Creek, GA
• Sugar Creek, IN
• EL68D Wasteway, WA

Georgia Watersheds

Indiana Watershed

Washington Watershed

Watershed Boundaries

• NAWQA Boundary

• Defined by USGS WRD personnel from contour maps

• GIS Weasel

• Automatically computed from DEM data

Comparison of Watershed Areas (hectares)

GIS Databases for Parameter Extraction

• National Elevation Dataset (30-m)

• National Land Characteristics Data (30 m)

• Augmented with recent Landsat TM data

• Soils databases from USDA soil surveys

• Scanned separates, rectified, vectorized, tagged

• Resampled the 30-m data to 60, 120, 210, 240, 480, 960, and 1920 meters

• 210-m roughly matches 10 acre grid size

AGNPS Parameter Generation

• AGNPS Data Generator

• Input parameter generation

• Details on generation of parameters

• Extraction methods

AGNPS Data Generator

• Created to provide interface between GIS software (Imagine) and AGNPS

• Developed interface for Imagine 8.4, running on WinNT/2000

AGNPS Data Generator

Input Parameter Generation

• 22 parameters; varying degrees of computational development

• Simple, straightforward, complex

Creating AGNPS Input

• Input Data File Creation

• Format generated parameters into AGNPS input file
• Use a “stacked” image file to create AGNPS data file (“.dat”) -- ASCII

Input Parameter Generation

Details on Generation of Parameters

• Cell Number

• Receiving Cell Number

• SCS Curve Number

• Uses both soil and land cover to resolve curve number

Details on Generation of Parameters

• Slope Shape Factor

Details on Generation of Parameters

• Slope Length

• A concern; max value should be 300 ft.

• Parameters 10, 11, 12, 14, 15, 16, and 17

• Uses Spatial Modeler to lookup attributes from soils or land cover

• Parameters 13, 18, 19, 20, and 21

• Hard coded on advice from experts

Details on Generation of Parameters

• Type of Channel

• Uses TARDEM program
• Creates a Strahler steam order

Extraction Methods

• Used object-oriented programming and macro languages

• C/ C++ and EML

• Manipulated the raster GIS databases with Imagine

• Extracted parameters for each resolution for both boundaries using AGNPS Data Generator

Creating AGNPS Output

• AGNPS creates a nonpoint source (“.nps”) file

• ASCII file like the input; tabular, numerical form

AGNPS Output

AGNPS Output

• AGNPS Output

Creating AGNPS Output Images

• Output Image Creation

• Combined “.nps” file with Parameter 1 to create multidimensional images
• Users can graphically display AGNPS output
• Process: create image with “x” layers, fill layers with AGNPS output data, set projection and stats for image
• Multi-layered (bands) images per model event

Creating AGNPS Output Images

Creating AGNPS Images

Results

• Resolution effects

• Tested with two independent collections
• Elevation at 3 m and 30 m resolution
• Land cover at 3 m and 30 m resolution
• Comparison of values

Regression Results

• 3 m to 30 m comparison

• Elevations -- R2 of 0.81

• Land cover – McFadden’s pseudo R2 of 0.139, meaning little correlation

• Derived parameters, e.g., slope, problematic because of degraded data source

Results

• Resampling effects

Experimental Approach

• Analysis requires DEM, slope, and land cover at 30, 60, 120, 210, 240, 480, 960, 1920 m cells

• Starting point is 30 m DEM and land cover

• Calculate slope at 30 m cell size from DEM

• Resample land cover

• How to generate slope at 60 m and larger cell sizes? How to aggregate land cover?

Method of Calculation

• Slope calculated from DEM

• 30, 60, 120, 210, 240, 480, 960, 1920 m cells

• Compute slope from 30 DEM

• Aggregate DEM from 30 m to each lower resolution

• Compute slope from aggregated elevation data

Sample of Slope Generation Approaches

Results - DEM

Results - DEM

Image Results -- DEM

Results -- Slope

Results -- Slope

• Slope

• Method of calculation affects results
• Higher resolution aggregation directly to large pixel sizes yields better results than multistage aggregation (e.g., 30 m to 960 m is better than 30 m to 60 m to 120 m to 240 m to 480 m to 960 m)
• Even multiples of pixels hold results while odd pixel sizes introduce error

Slope Image Comparison

Sample of Land Cover Aggregation Approaches

Results - Land Cover -- 120 M Pixels

Results - Land Cover -- 210 m Pixels

Results - Land Cover -- 480 m Pixels

Results-Land Cover -- 960 m Pixels

Image Results - Land Cover

Image Results - Land Cover

Statistical Testing

• Selected 500 random points over the watershed

• Compared elevation, slope, and land cover values at the 500 points

• Computed R2 and pseudo R2 between resolutions

• Plotted R2 and pseudo R2 against resampled resolutions from 30 m data

Conclusions

• Automatic generation of AGNPS parameters from elevation, land cover, and soils

• Resolution affects results

• Elevation and derivatives (slope) hold values well because of averaging methods of resampling
• Land cover (categorical data) is inconsistent across resolutions because of nearest neighbor resampling

Conclusions

• Resampling retains values better with even multiples of original pixel sizes

• Aggregation directly from higher resolution to lower retains values better than multiple intermediate resampling

Resolution and Resampling Effects of GIS Databases for Watershed Models

• E. Lynn Usery

• U.S. Geological Survey

• University of Georgia

• Michael P. Finn

• U.S. Geological Survey