Elsevier

Remote Sensing of Environment

Volume 104, Issue 4, 30 October 2006, Pages 409-415
Remote Sensing of Environment

Accuracy assessment of the processed SRTM-based elevation data by CGIAR using field data from USA and Thailand and its relation to the terrain characteristics

https://doi.org/10.1016/j.rse.2006.05.012Get rights and content

Abstract

Shuttle radar topographic mission (SRTM) has created an unparalleled data set of global elevations that is freely available for modeling and environmental applications. The global availability (almost 80% of the Earth surface) of SRTM data provides baseline information for many types of the worldwide research. The processed SRTM 90 m digital elevation model (DEM) for the entire globe was compiled by Consultative Group for International Agriculture Research Consortium for Spatial Information (CGIAR-CSI) and made available to the public via internet mapping interface. This product presents a great value for scientists dealing with terrain analysis, thanks to its easy download procedure and ready-to-use format. However, overall assessment of the accuracy of this product requires additional regional studies involving ground truth control and accuracy verification methods with higher level of precision, such as the global positioning system (GPS).

The study presented in this paper is based on two independent datasets collected with the same GPS system in Catskill Mountains (New York, USA) and Phuket (Thailand). Both datasets were corrected with differential base station data. Statistical analysis included estimation of absolute errors and multiple regression analysis with slope and aspect variables. Data were analyzed for each location separately and in combination. Differences in terrain and geographical location allowed independent interpretation of results.

The results of this study showed that absolute average vertical errors from CGIAR dataset can range from 7.58 ± 0.60 m in Phuket to 4.07 ± 0.47 m in Catskills (mean ± S.E.M.). This is significantly better than a standard SRTM accuracy value indicated in its specification (i.e. 16 m). The error values have strong correlation with slope and certain aspect values. Taking into account slope and aspect considerably improved the accuracy of the CGIAR DEM product for terrain with slope values greater than 10°; however, for the terrain with slope values less than 10°, this improvement was found to be negligible.

Introduction

SRTM data were collected during 11-day mission in February 2000. Since then, they were described in details (Farr and Kobrick, 2000, Rabus et al., 2003, Werner, 2001) and became accessible for free download over the Internet (e.g. at ftp://e0srp01u.ecs.nasa.gov and http://seamless.usgs.gov/). However, between two SRTM products that include raster data with 30 and 90 m spatial resolution, only 90 m data is available globally (80% of the Earth surface) while the 30 m data are available only for the USA territory.

Since SRTM data became widely available, many studies utilized them for applications in topography (Falorni et al., 2005, Koch and Lohmann, 2000), geomorphology (Guth, 2003, Stock et al., 2002), vegetation cover studies (Kellndorfer et al., 2004), tsunami assessment (Blumberg et al., 2005), and urban studies (Gamba et al., 2002). SRTM data verification was performed using various altimetry data (Helm et al., 2002, Sun et al., 2003) and digital elevation models (Jarvis et al., 2004, Muller, 2005, Smith and Sandwell, 2003).

Because SRTM data produced a number of voids due to lack of contrast in the radar image, a methodology based on spatial filtering was developed to correct this phenomenon (Dowding et al., 2004, Jarvis et al., 2004). The final seamless dataset with voids filled in is available at the website of Consultative Group for International Agriculture Research Consortium for Spatial Information (CGIAR-CSI) via http://srtm.csi.cgiar.org/. Dataset compiled by CGIAR-CSI has the following advantages:

  • 1.

    Dataset is seamless.

  • 2.

    There is user-friendly interface for downloading specific DEM areas of interest.

  • 3.

    Data are already pre-processed for the immediate use.

  • 4.

    Data are available in GeoTiff format that is accepted by most GIS applications.

  • 5.

    The website is supported by extensive documentation describing the process of filling voids in SRTM product.

These advantages make CGIAR-CSI SRTM data product a very valuable resource, especially for cases when analysis of terrain had to be done promptly, for example, during Asian Tsunami of 2004 (Blumberg et al., 2005). However, the accuracy of this product is yet to be assessed. Partial assessment of its accuracy was done by the Centro Internacional de Agricultura Tropical (CIAT) in South America to verify performance of the developed DEM (Jarvis et al., 2004), but global user community would gain more benefits from other regional assessments.

This study is designed to assess CGIAR DEM accuracy by comparing elevation values from processed CGIAR-CSI SRTM with elevation values collected by differential GPS system in two different geographical localities. Both datasets are based on the same vertical datum (WGS84).

Section snippets

Study areas characteristics

Elevation data in this study were collected in two geographically independent regions: Catskill Mountains (New York, USA) and Phuket (Thailand). Catskill area was surveyed in November 1999 and Phuket was surveyed in May 2005. Both surveys were performed with the same unit, Trimble ProXR, with real-time differential mode. After collection, data were corrected using information from local base stations.

Study area in Catskill Mountains covers slopes with elevations 577–811 m above mean sea level (

GPS settings and surveying

Among various methods of accuracy assessment, GPS survey provides the best way to map features on terrain with high accuracy. In this study, GPS data were collected along roads and passable trails with Trimble ProXR receiver. It collects data using 12 channels, L1 frequency (1575.42 MHz), CA code (Coarse Acquisition Code for modulating L1 frequency), real-time receiver, and integrated beacon. These settings allow collecting high-accuracy signals into a data logger connected with the receiver (

CGIAR-CSI SRTM data accuracy

Table 1, Table 2 show descriptive statistics for CGIAR-CSI SRTM and GPS data for Phuket and Catskill areas. Results of t-tests confirm lack of significant differences between data obtained by the two methods in either study area. Table 3 summarizes discrepancies between CGIAR-CSI SRTM and GPS measurements for Phuket and Catskill Mountains study areas. Average absolute error of CGIAR-CSI SRTM data was found to be 7.58 ± 0.60 m (Phuket) and 4.07 ± 0.47 m (Catskills).

Linear regression analysis reveals

Conclusions

Analyses presented in this paper indicate that:

  • 1.

    Absolute vertical accuracy of CGIAR-CSI SRTM data for our datasets proved to be two to four times higher than the value of 16 m presented in the original SRTM requirement specification.

  • 2.

    Both slope and aspect characteristics of the terrain have significant impact on accuracy of CGIAR-CSI SRTM data. Accuracy particularly suffers on terrains with slope values higher than 10°. Aspect of the terrain influences both the magnitude and the sign of errors in

Acknowledgements

Many thanks to Ms. Patricia Ahmetaj and Ms. Lori Machung, New York City Department of Environmental Protection, for their assistance in field data collection in Catskill Mountains. Mr. Woody from Phuket (Thailand) provided invaluable assistance in field work during May 2005.

References (27)

  • T.G. Farr et al.

    Shuttle radar topography mission produces a wealth of data

    Transactions of the American Geophysical Union

    (2000)
  • J. Farrell et al.

    Differential carrier phase GPS-aided INS for automotive applications

  • J.A. Farrell et al.

    Carrier phase GPS-aided INS-based vehicle lateral control

    Journal of Dynamic Systems, Measurement, and Control

    (2003)
  • Cited by (0)

    View full text