Uncertainties in the Shuttle Radar Topography Mission (SRTM) Heights: Insights from the Indian Himalaya and Peninsula

Manas Mukul,Sridevi Jade,Malay Mukul,Vinee Srivastava

doi:10.1038/srep41672

Abstract

The Shuttle Radar Topography Mission (SRTM) Digital Terrain Elevation Data (DTED) are used with the consensus view that it has a minimum vertical accuracy of 16 m absolute error at 90% confidence (Root Mean Square Error (RMSE) of 9.73 m) world-wide. However, vertical accuracy of the data decreases with increase in slope and elevation due to presence of large outliers and voids. Therefore, studies using SRTM data “as is”, especially in regions like the Himalaya, are not statistically meaningful. New data from ~200 high-precision static Global Position System (GPS) Independent Check Points (ICPs) in the Himalaya and Peninsular India indicate that only 1-arc X-Band data are usable “as is” in the Himalaya as it has height accuracy of 9.18 m (RMSE). In contrast, recently released (2014–2015) “as-is” 1-arc and widely used 3-arc C-Band data have a height accuracy of RMSE 23.53 m and 47.24 m and need to be corrected before use. Outlier and void filtering improves the height accuracy to RMSE 8 m, 10.14 m, 14.38 m for 1-arc X and C-Band and 3-arc C-Band data respectively. Our study indicates that the C-Band 90 m and 30 m DEMs are well-aligned and without any significant horizontal offset implying that area and length computations using both the datasets have identical values.

Highlights

The initial validation of global C-Band 90 m (C90) Shuttle Radar Topography Mission (SRTM) heights used dual-frequency Real Time Kinematic (RTK) Global Positioning System (GPS)[6,7]
We address the following important issues related to the use of SRTM data and its vertical uncertainty using ~200 high-precision Global Position System (GPS) points as Independent Check Points (ICPs) (Fig. 1, Supplementary Table S1) in India: (1) The minimum horizontal precision of GPS ICPs required for accurate pixel registration for 90 m, 30 m and 15 m resolution SRTM Digital Terrain Elevation Data (DTED) for defining a quantitative threshold for selection of ICPs. (2) The accuracy of C-Band and X-Band SRTM data in India and the Himalaya
We first address the issue of minimum precision of the position of the ICPs required for accurate pixel registration and analyse the “as is”, void and outlier filtered SRTM data (Table 1) and study the effect of re-sampling on vertical accuracy of C-Band SRTM data by the analysis of outlier-filtered 191 C-Band 90 m SRTM (C90), 193 C30 and 194 C30_15 ICPs (Table 2)

Summary

Introduction

The initial validation of global C-Band 90 m (C90) SRTM heights used dual-frequency Real Time Kinematic (RTK) Global Positioning System (GPS)[6,7]. (3) Large outliers and voids exist in the SRTM dataset[6,7,8] This suggests that assessment of the local accuracy of SRTM heights needs to be carried out before its statistically meaningful use and a single representative value for entire continents is oversimplification. Geomorphometric analysis[22], vegetation cover studies[23], analysis of avulsion threshold of rivers[24], assessment of tsunami[25], modeling of soil particle size distribution[26], measuring the surface elevation changes of glaciers[27], morphology of terraces[28] and fault geometry of active folds[29] have used SRTM data “as is” without outlier analysis or propagation of errors associated with vertical (height) uncertainties in SRTM data.

Methods

Results

Discussion

Conclusion