VALIDATION OF A NEW 30 METER GROUND SAMPLED GLOBAL DEM USING ICESAT LIDARA ELEVATION REFERENCE DATA

Transcription

1 VALIDATION OF A NEW 30 METER GROUND SAMPLED GLOBAL DEM USING ICESAT LIDARA ELEVATION REFERENCE DATA M. Lorraine Tighe Director, Geospatial Solutions Intermap Session: Photogrammetry & Image Processing Date: 27 th April 2012 Timing: hrs Venue: Hall E 107 Time/speaker minutes

2 Overview This presentation evaluates a sample tile of Intermap s 30m World DEM Product in South America A visual comparison of the World DEM product to ASTER and SRTM is provided to highlight the improvements that Intermap s 30m World DEM product offers A comparison using ICESat reference points is provided to demonstrate statistical improvements of the 30m World DEM compared to other global elevation datasets Intermap Technologies. All rights reserved.

3 Agenda DEM Technologies and Products Data Sets Methods: DEM Fusion Technique Analysis Results Conclusions SAR Imagery Intermap Technologies. All rights reserved. Surface elevation Ground elevation

4 DEM Technologies (e.g. Stereo Images) Multiple airborne and spaceborne technologies are available for the creation of digital elevation models: Microwave Interferometric DEMs - NEXTMap, TerraSAR-X, Tandem-X, COSMO Skymed Laser Point Cloud DEMs: - Airborne LiDAR systems, ICESat Optical Auto-correlated DEMs - Digital Globe, GeoEye, Astrium (SPOT), RapidEye, MS (Clearview30) Intermap Technologies. All rights reserved.

5 DEM Technologies Various airborne and spaceborne technologies are available for DEM generation Intermap Technologies. All rights reserved.

6 NEXTMap Data Products Digital Surface Model 5m post, 1m vertical accuracy Digital Terrain Model 5m post, 1m vertical accuracy Orthorectified Radar Images 1.25m pixel, 2m horizontal accuracy Intermap Technologies. All rights reserved.

7 DEM Data Sets Used A number of SAR and optical air- and space-borne platforms are currently used in aggregation of the NEXTMap World 30 m product. Sensor NEXTMap SRTM ASTER ICESat Country Canada USA Japan USA Wavelength SAR SAR Optical Optical DEM Method IFSAR IFSAR Photogrammetry Reference Vertical Accuracy (LE90%) 6 m 15 m.30 m DEM Type DSM/DTM DSM DSM GCP DEM GSD (m) Void Filled Yes Yes No N/A Hydro-enforced Yes Yes No N/A 7

8 Reference Data Set Validation of the vertical accuracy of the NEXTMap 30m data is of critical importance to any mapping program to ensure that the elevation data meet the required specifications mapping standards. The primary challenge in validating a data collection at the global scale and rugged and remote terrain is obtaining survey-grade reference data that is accurate enough and has suitable coverage to encompass the entire collection area. NASA's Ice, Cloud and land Elevation Satellite (ICESat) employing the Geoscience Laser Altimeter System (GLAS) has collected a unique set of full-waveform LiDAR data with global coverage during campaigns that began in 2003 and ended in This system provides a consistently referenced elevation data set with unprecedented accuracy and quantified measurement errors that can be used to generate GCPs with sub-decimeter vertical accuracy.

9 Overview: Amarillo Test Area (all valid ICESat pts intercepting LiDAR tie lines) T1 LiDAR Tie lines: (T1, T2, T3) LiDAR Check lines: (C1, C2) C1 91-day orbits T2 C2 T3 ~180 km ~400 km LiDAR acquired for Intermap by Airborne 1 in All data tied to Intermap survey points. Systematic errors reduced to ~10 cm. Spec was 20cm due to long baselines. These data averaged to 25 m grid. All vertical data transformed to WGS84(G1150) 9

10 Methods: Data Fusion Technique Data Aggregation or Fusion: Combines multiple Digital Elevation Models (DEM) from varying sources at varying accuracies and resolutions into one homogeneous product Alternate data can be blended or fused to more consistent foundational elevation data Data Fusion can improve vertical accuracy of DEM inputs and bring multiple external data sets to the same reference Data Mining Data Specification Validation Data Alignment Fusion Editing Finishing Intermap Technologies. All rights reserved.

11 Data Specification Determination of what the final output data specification will be will drive the following data fusion processing steps: LiDAR Ground sampling distance Hydro-enforced edit rules (e.g. water edits) Seamless Complete Coverage NEXTMap Intermap Technologies. All rights reserved.

12 DEM Alignment Example DEM Alignment is made to the reference DEM with better absolute accuracies (not necessarily better resolution or precision) to correct for these systematic errors. Before Alignment Before Alignment After Alignment Reference DEM LiDAR Data Set #2 After Alignment Intermap Technologies. All rights reserved.

13 DEM Fusion Example After the DEMs are aligned, the boundary between the two is blended (fused) together to produce a seamless single DEM Intermap Technologies. All rights reserved.

14 Hydro-enforced Terrain Terrain datasets may need to be hydro enforced (or monotonic - all watercourse features flow downstream, and all standing water bodies are flat through our editing process) Intermap Technologies. All rights reserved.

15 Sample Product Location State of Amazonas, Brazil 1 degree x 1 degree tile: 63 o W, 1 o N 15

16 ASTER 30m v2.0 Global DEM Product Noisy texture throughout )low absolute accuracy) Elevation anomalies present in some areas 16

17 SRTM 90m v4.1 Product Coarse Resolution and Systematic Errors (0-6m range) Contains void areas, particularly in high terrain relief 17

18 NEXTMap 30m World DEM Product No spikes or noisy texture Moderate resolution and no voids 18

19 Example of ASTER Elevation Spikes Large spikes occur frequently in the ASTER 30m v2.0 dataset ASTER 30m v2.0 DEM 19

20 Example of ASTER Elevation Spikes There are voids in the 90m SRTM v2.1 DEM in the same location as the spikes from the ASTER DEM SRTM 90m v2.1 DEM 20

21 Example of ASTER Elevation Spikes Intermap s 30m World DEM contains no spikes and no voids Intermap 30m World DEM 21

22 Cumulative Proportion of Data (%) ICESat Reference Point Comparison 793 ICESat Points o o o Accurate to 25cm RMSE Slope < 10 o Unobstructed terrain Intermap s 30m World DEM Elevation Difference From ICESAT ASTER 30m v2.0 SRTM 90m v2.1 Intermap 30m World DEM min (m) max (m) mean (m) std (m) RMS (m) LE90 (m) Distribution of Elevation Errors ASTER SRTM World DEM Elevation Difference from ICESat (m)

23 Conclusions We reviewed how a variety of elevation data sets from different sources could be fused to create a seamless, consistent elevation foundation for use in a host of applications at global scales. Data fusion process is critical in a world where we work with multiple data sets, comprised of different technologies, resolutions, accuracies, over areas of interest. A global 30 m DEM of this kind of accuracy is needed! Intermap Technologies. All rights reserved.

24 For various parts of the world, digital maps of Earth's topography are limited, inaccurate, or nonexistent. For example, many mountain chains, inhospitable deserts, and dense tropical rain forests have topographic coverage that is totally inadequate mainly because of the difficulty in getting to these locations. Even where topographic maps exist, they may have been created in such a way as to limit their usefulness. Neighboring countries may generate topographic data using entirely different methods. This lack of standardization effectively limits the scope of regional or global studies where precise topography is important. Creation of a high-resolution global elevation model from the best-of-breed datasets among the existing NEXTMap, LiDAR, SRTM, ASTER, and GTOPO30 has been validated using ICESat GLAS LiDAR elevation data as reference data is presented. The vertical accuracies of ICESat reference data is within 25cm RMSE on (relatively) flat and open areas. The positional accuracy is published to be about 4m. Because the footprint is quite large (~65mx65m), the ICESat points were filtered for use as reference data. The reference data for the globe were manually assessed for suitability and categorized based on land cover (barren and five vegetation classes) by cross comparison with optical imagery. The impact of vegetation cover and terrain slope on the vertical accuracy of the 30 m elevation data were assessed. Results indicate that the high resolution 30 m DEM performs better in barren low sloped terrain, than in vegetated and slopes greater than 10 degrees. The vertical accuracy results offer an improvement over the SRTM V4 and ASTER V2 near global elevation data sets. This data set has great implications for national governments, who now have a high resolution seamless DEM for their entire country Intermap Technologies. All rights reserved.

25 25 Digital Elevation Models (DEMs) provide fundamental information that is required for many 3D geo-spatial applications. Recent technological advances in airborne and spaceborne sensors have led to a new era of DEM generation. Foremost among these technologies are the use of stereo photogrammetry, light detection and ranging (LiDAR), stereo radargrammetry based on synthetic aperture radar (SAR) images, and interferometric SAR methods. These technologies are generating high resolution surface and terrain elevation models which play a critical role in applications such as climate modelling, forest inventory, flood analysis, topographic mapping, land and fire applications, and biomass studies. Obtaining high-resolution elevation data covering a specific area can be expensive, and often results in the collection and purchasing of multiple datasets over time. Ultimately, it may also lead to data collections that differ in resolution, dates, formats, and more. DEM fusion techniques combine data from different sources into a single, consistent, homogeneous elevation dataset that is often more accurate and useful than any of the individual input elevation datasets. This paper presents sample data fusion methods, an overview of the unique characteristics of DEMs derived from these different technologies and discusses the benefits of the fusion process for a variety of applications such as water management, urban planning, telecommunications and flood plain mapping. Additionally, an approach for the production of a fused Global DEM using the best available datasets among the existing NEXTMap, LiDAR, SRTM, ASTER, GTOPO30 and ICESAT GLAS elevation data sources is proposed. Several sample data fusion results are presented and discussed Intermap Technologies. All rights reserved.