Quality Assessment of Shuttle Radar Topography Mission Digital Elevation Data. Thanks to. SRTM Data Collection SRTM. SRTM Galapagos.

Similar documents
Utility of National Spatial Data for Conservation Design Projects

Terms GIS GPS Vector Data Model Raster Data Model Feature Attribute Table Point Line Polygon Pixel RGB Overlay Function

Validation of the ASTER Global Digital Elevation Model (GDEM) Version 2 over the Conterminous United States

GIS and Remote Sensing

Digital Elevation Models (DEM)

Pierce Cedar Creek Institute GIS Development Final Report. Grand Valley State University

Estimating Probability of Success Rate

Introduction to GIS - 2

GeoSUR SRTM 30-m / TPS

CHAPTER 1 THE UNITED STATES 2001 NATIONAL LAND COVER DATABASE

ENGRG Introduction to GIS

Harrison 1. Identifying Wetlands by GIS Software Submitted July 30, ,470 words By Catherine Harrison University of Virginia

Lab 1: Importing Data, Rectification, Datums, Projections, and Coordinate Systems

The Looming Threat of Rising Sea Levels to the Florida Keys

ENGRG Introduction to GIS

Summary Description Municipality of Anchorage. Anchorage Coastal Resource Atlas Project

A Detailed Examination of DTM Creation Methods and Sources. Study Area Overview

Modeling the Rural Urban Interface in the South Carolina Piedmont: T. Stephen Eddins Lawrence Gering Jeff Hazelton Molly Espey

Southern California Earthquake Center

Targeted LiDAR use in Support of In-Office Address Canvassing (IOAC) March 13, 2017 MAPPS, Silver Spring MD

Map Projections. Displaying the earth on 2 dimensional maps

AN EVALUATION ON THE DATA QUALITY OF SRTM DEM AT THE ALPINE AND PLATEAU AREA, NORTH-WESTERN OF CHINA

Kimberly J. Mueller Risk Management Solutions, Newark, CA. Dr. Auguste Boissonade Risk Management Solutions, Newark, CA

Classification of Erosion Susceptibility

South Florida Coastal Storm Surge and Mapping Study

Aerial Photography and Imagery Resources Guide

Welcome to NR502 GIS Applications in Natural Resources. You can take this course for 1 or 2 credits. There is also an option for 3 credits.

Aerial Photography and Imagery Resources Guide

Lab 1: Importing Data, Rectification, Datums, Projections, and Output (Mapping)

Positional Accuracy of the Google Earth Imagery In The Gaza Strip

Chapter 6. Fundamentals of GIS-Based Data Analysis for Decision Support. Table 6.1. Spatial Data Transformations by Geospatial Data Types

Preliminary Calculation of Landscape Integrity in West Virginia Based on Distance from Weighted Disturbances

Semester Project Final Report. Logan River Flood Plain Analysis Using ArcGIS, HEC-GeoRAS, and HEC-RAS

Investigation of the Effect of Transportation Network on Urban Growth by Using Satellite Images and Geographic Information Systems

ACCURACY ANALYSIS OF SRTM HEIGHT MODELS INTRODUCTION

2011 Land Use/Land Cover Delineation. Meghan Jenkins, GIS Analyst, GISP Jennifer Kinzer, GIS Coordinator, GISP

Ground Truth Annual Conference. Optimized Design of Low Distortion Projections. Michael L. Dennis, RLS, PE

Projections and Coordinate Systems

GRAPEVINE LAKE MODELING & WATERSHED CHARACTERISTICS

Urban Erosion Potential Risk Mapping with GIS

Miami-Dade County Technical Update Meeting South Florida Coastal Study. May 11, 2016

Huron Creek Watershed 2005 Land Use Map

Development and Land Use Change in the Central Potomac River Watershed. Rebecca Posa. GIS for Water Resources, Fall 2014 University of Texas

Existing NWS Flash Flood Guidance

Appendix J Vegetation Change Analysis Methodology

OBJECT-BASED CLASSIFICATION USING HIGH RESOLUTION SATELLITE DATA AS A TOOL FOR MANAGING TRADITIONAL JAPANESE RURAL LANDSCAPES

LANDSLIDE RISK ASSESSMENT IN YOSEMITE NATIONAL PARK. Edna Rodriguez December 1 st, 2016 Final Project

If this is your first experience with LiDAR data you might want to check out the MGISAC s whitepaper on LiDAR here:

Metadata for 2005 Orthophotography Products

Preliminary Data Release for the Humboldt Bay Sea Level Rise Vulnerability Assessment: Humboldt Bay Sea Level Rise Inundation Mapping

An Internet-based Agricultural Land Use Trends Visualization System (AgLuT)

Geographic Information Systems class # 1 February 19, Coordinate reference systems in GIS: geodetic coordinates

ANCILLARY DATA VARIABLES

Geog 469 GIS Workshop. Data Analysis

ELEVATION. The Base Map

Effects of input DEM data spatial resolution on Upstream Flood modeling result A case study in Willamette river downtown Portland

EXPERIENCE WITH HIGH ACCURACY COUNTY DTM MAPPING USING SURVEYING, PHOTOGRAMMETRIC, AND LIDAR TECHNOLOGIES - GOVERNMENT AND CONTRACTOR PERSPECTIVES

Delineation of high landslide risk areas as a result of land cover, slope, and geology in San Mateo County, California

Abstract: Contents. Literature review. 2 Methodology.. 2 Applications, results and discussion.. 2 Conclusions 12. Introduction

Digital Elevation Models (DEM) / DTM

Bathymetry Data and Models: Best Practices

Comparison in terms of accuracy of DEMs extracted from SRTM/X-SAR data and SRTM/C-SAR data: A Case Study of the Thi-Qar Governorate /AL-Shtra City

New Land Cover & Land Use Data for the Chesapeake Bay Watershed

Potential Restorable Wetlands (PRWs):

GIS Semester Project Working With Water Well Data in Irion County, Texas

Digitization in a Census

General Overview and Facts about the Irobland

GIS for the Non-Expert

Hydrology and Floodplain Analysis, Chapter 10

VIC Hydrology Model Training Workshop Part II: Building a model

EMERGENCY PLANNING IN NORTHERN ALGERIA BASED ON REMOTE SENSING DATA IN RESPECT TO TSUNAMI HAZARD PREPAREDNESS

Spatial Process VS. Non-spatial Process. Landscape Process

Manitoba s Elevation (LiDAR) & Imagery Datasets. Acquisition Plans & Opportunities for Collaboration

GeoWEPP Tutorial Appendix

FR Exam 2 Substitute Project!!!!!!! 1

Spatial Effects on Current and Future Climate of Ipomopsis aggregata Populations in Colorado Patterns of Precipitation and Maximum Temperature

GPS- vs. DEM-Derived Elevation Estimates from a Hardwood Dominated Forest Watershed

Using NFHL Data for Hazus Flood Hazard Analysis: An Exploratory Study

Erosion Susceptibility in the area Around the Okanogan Fire Complex, Washington, US

Viewshed Creation: From Digital Terrain Model to Digital Surface Model. Edward Ashton Pennsylvania State University

ACCURACY ASSESSMENT OF ASTER GLOBAL DEM OVER TURKEY

Object Based Imagery Exploration with. Outline

Display data in a map-like format so that geographic patterns and interrelationships are visible

Land Use and Land Cover Semantics - Principles, Best Practices and Prospects. Ola Ahlqvist

Raster Spatial Analysis Specific Theory

Management and Use of LiDAR-derived Information. Elizabeth Cook, GIS Specialist

Geo 327G Semester Project. Landslide Suitability Assessment of Olympic National Park, WA. Fall Shane Lewis

THE USE OF GEOMATICS IN CULTURAL HERITAGE AND ARCHAEOLOGY FOR VARIOUS PURPOSES

Alaska, USA. Sam Robbins

SECTION 4 PARCEL IDENTIFIERS 4.1 LONGITUDE AND LATITUDE

Handling Raster Data for Hydrologic Applications

NR402 GIS Applications in Natural Resources

THE ROLE OF GEOCOMPUTATION IN THE HYDROLOGICAL SCIENCES

Boreal Fen probability - metadata

Town of Chino Valley. Survey Control Network Report. mgfneerhg mc N. Willow Creek Road Prescott AZ

How to Create Stream Networks using DEM and TauDEM

AGOG 485/585 /APLN 533 Spring Lecture 5: MODIS land cover product (MCD12Q1). Additional sources of MODIS data

Implementation of CLIMAP and GIS for Mapping the Climatic Dataset of Northern Iraq

A GIS View of Hydrology

Flood Hazard Zone Modeling for Regulation Development

Transcription:

Quality Assessment of Shuttle Radar Topography Mission Digital Elevation Data Third International Conference on Geographic Information Science College Park, Maryland, October 20-23 Ashton Shortridge Dept. of Geography Michigan State University ashton@msu.edu Thanks to Scott Oppman, Oakland County (Michigan) Information Technology Dept for data! NASA personnel and many others for flying the & processing all that data! Shuttle Radar Topography Mission Flown in February, 2000 Collected data over 80% Earth's land area All land between 60 degrees N, 56 degrees S Data released at 1 arc second interval for US Released at 3 arc second interval for row Data Collection Radar signals transmitted from Shuttle Received back at two antennas One in shuttle bay One on end of 60m boom Difference between two signals used to reconstruct elevation http://www.jpl.nasa.gov/srtm/missionoverview.html Data Resources Galapagos Information: http://srtm.usgs.gov/ Americas Download: http://seamless.usgs.gov/ Global Download: http://glcfapp.umiacs.umd.edu:8080/esdi/index.jsp http://photojournal.jpl.nasa.gov/catalog/pia02735

GTOPO 30 Galapagos Quality? How to consider data quality when this is the best available data for most everywhere?! ~1 km cells Find a data-rich location Examine error using careful methods Quantify Error and correlate with other characteristics Outline Case Study: part of Oakland County, MI Available data A non-raster based methodology for evaluating raster data accuracy Relationship between error & land cover Ortonville and the Shuttle Mission Study site: northern Oakland County, MI Ortonville (population 1,535) and environs 8.4 km x 6.7 km region Area facing rapid development at Detroit urban fringe Diverse land cover Varied topography (for MI!) Survey Elevations Survey Elevations Oakland County GIS contracted for detailed high accuracy countywide DEM from Woolpert LLC Aerial Photography collected in April 2000 Derived Points and Breaklines Stated accuracy 1 foot vertical 2.5 ft horizontal 46,065 points Range: 277 360 m. mean: 310.6 6.6 km 8.3 km

Oakland Data Characteristics Irregular postings Michigan State Plane, southern zone NAD 83, Units int'l feet Vertical units: International Ft. (NAVD 88) DEM Data Data obtained from http://seamless.usgs.gov/ 1 NED NAD83, vertical units meters, NAVD 88 3 (to match with non-us product) WGS84, vertical units meters As DEMs NED Land Cover Data 1 1992 NLCD Land Cover Modified Anderson Level 2 from Landsat TM NAD83, from Seamless 30 meter 2001 Land Cover Michigan GAP, Multiple Anderson Levels from Landsat TM Michigan GeoRef (oblique Mercator) Relatively Incompatible

1992 NLCD 2001 GAP-IFMAP : 11 (Water) : 21 (LI Resid) : 22 (HI Resid) : 23 (Com/Ind) : 41 (Dec. Forest) : 42 (Evg. Forest) : 43 (Mix Forest) : 81 (Pasture) : 82 (Row Crops) : 91 (Wood. WL) : 92 (Em. H. WL) : 11 (Water) : 21 (LI Resid) : 22 (HI Resid) : 23 (Com/Ind) : 41 (Dec. Forest) : 42 (Evg. Forest) : 43 (Mix Forest) : 81 (Pasture) : 82 (Row Crops) : 91 (Wood. WL) : 92 (Em. H. WL) Methods How to integrate this data? Different datums, coordinate systems, vertical units, spatial resolutions... Identify a method that is gentlest on the original data vs A Raster Methodology Decide upon a common system Datum / Projection / Coordinate System Origin, Dimensions, Cell Size Preprocess data to that system Project, Resample, Clip rasters Project, Convert Oakland Co. points to raster Subtract TRUE from & Intersect with Landcover Alternative, Point-Based Methodology (I) Assume elevations are gridded spot heights Not areal averages Decide upon a common system Datum / Projection / Coordinate System Locations at which to conduct analysis I chose to compare at the DEM locations Preprocess data to that system Convert rasters to points, project the points Project Oakland Co. points Alternative, Point-Based Methodology (II) Interpolate 'True' heights at and NED spot locations IDW, power 2, closest 6 neighbors Interpolate land cover classes at and NED spot locations Nearest - Neighbor Subtract 'True' from 'DEM' & Intersect with Landcover

Platforms Methodology 1 (Raster) implemented in Arc 8.2 (ESRI) Methodology 2 (Point) implemented in R 1.9.1 (Open Source Statistics Software) Descriptive Statistics & LC Correlations in R for both approaches Raster Method Error Statistics Results - Error Mean: 2.92 m.; SD: 3.79 m.; RMSE: 4.78 m NED Error Statistics Mean: 1.06 m.; SD: 1.49 m.; RMSE: 1.83 m Points Method Error Statistics Mean: 2.95 m.; SD: 3.93 m.; RMSE: 4.92 m NED Error Statistics Mean: 1.07 m.; SD: 1.51 m.; RMSE: 1.85 m NED Error (Raster) DEM Error (Point) -9.4 15.1 m. Error 1992 NLCD : 11 (Water) -9 22.7 m. : 21 (LI Resid) : 22 (HI Resid) : 23 (Com/Ind) : 41 (Dec. Forest) : 42 (Evg. Forest) : 43 (Mix Forest) : 81 (Pasture) : 82 (Row Crops) : 91 (Wood. WL) : 92 (Em. H. WL)

Land Cover and Error Error by LC Class (2001) Error split by overlying land cover type Significant difference (p-value < 2.2e-16) in mean error between Land Cover Classes One-way test of means Kruskal-Wallis rank sum test Forest Classes associated with substantial positive error bias too high Upland Oak Forest Mixed Dec. Pines Upland Mixed For. Error by LC Class (1992) NED Error by LC Class (2001) Forest NED Error by LC Class (1992) Wilcoxan Rank-Sum Test () mu=0 mu=3 mu=4 Class p-value p-value p-value Desc 11 0.398 1 1 Open Water 21 *** 1 1 Low Intensity Residential 22 0.158 1 1 High Intensity Residential 23 0.622 1 1 Commercial/Indust/Transport 41 *** *** *** Deciduous Forest 42 *** *** *** Evergreen Forest 43 *** 0.003 0.075 Mixed Forest 81 *** 1 1 Pasture/Hay 82 *** 1 1 Row Crops 91 *** 0.033 1 Woody Wetlands 92 *** 1 1 Emergent Herb. Wetlands *** indicates << 0.0001

Discussion - Methods Point-based method minimized change to elevations Interpolation must occur to evaluate error at each node in projected NED and Differences with raster method were slight Elevation differences reduced RMSE ~10 th meter lower Effect of forested land cover reduced Still highly significantly biased Discussion Error RMSE is well within specifications < 16 meters (4.9 m. for study area) Error significantly higher than zero Average ~ 3 m. is too high Significantly more bias in forested areas Means in the 4-6 meter range Discussion Error (II) RMSE magnitude strongly linked to forested land cover Returns not striking the ground RMSE 6-8 meter range Opportunity for statistical error models Employ landcover characteristics to adjust (co)variance models Identify canopy height? meets basic specs Conclusions But mean error is positive (biased too high) And variation of error is correlated with landcover Forests and Error Forests introduce positive bias Evergreen forests may be more error-prone Expect regions not experiencing leaf-off conditions to have higher error than Michigan Preprocessing choices make slight difference Split into Land Cover Classes (1992 NLCD Land Cover Classification System) Error (m.) Code NPts Mean STD RMSE Description 11 74-0.38 3.16 3.16 Open Water 21 202 1.35 2.39 2.74 Low Intensity Residential 22 24 0.52 2.36 2.37 High Intensity Residential 23 64 0.04 1.84 1.82 Commercl/Indust/Trans 41 2740 4.75 4.17 6.32 Deciduous Forest 42 252 6.08 4.13 7.35 Evergreen Forest 43 28 5.04 4.05 6.42 Mixed Forest 81 1447 1.34 2.72 3.03 Pasture/Hay 82 1289 0.91 2.82 2.96 Row Crops 91 722 3.44 3.40 4.83 Woody Wetlands 92 202 1.08 3.21 3.37 Emergent Herb. Wetlands