Printer Friendly

Impact of DEM resolution and accuracy on remote sensing and topographic DEM derived topographic attributes and drainage pattern.

Byline: Misbah Amjad, Muhammad Zeeshan Ali and Muhammad Shafique

Abstract

Digital Elevation Model (DEM) presents the digital representation of surface topography. DEMs can be derived from satellite remote sensing or conventional techniques of surveying. However, inherent errors in a remote sensing based DEMs can lead to uncertainty in the elevation data and computed topographic attributes. This study evaluates accuracy of remote sensing derived SRTM and ASTER DEMs and physical survey derived Topographic maps. For accuracy assessment of the ASTER and SRTM ASTER DEMs, spot elevation were extracted from the toposheet as a reference data. Acccuracy assessment results reveals that the SRTM DEM shows RMSE of +-27 m and ASTER DEM shows RMSE of +-20 m in the study area. The impact of random errors in the ASTER and SRTM DEMs on the derived topographic attributes (aspect and slope) is evaluated through the Monte Carlo Simulation (MCS).

Derived results show higher uncertainty in topographical attributes (aspect and slope) derived from the ASTER DEM than from the SRTM DEM. Results derived from the ASTER DEM shows that mean of slope can deviate 1.2deg whereas, SRTM DEM shows it can deviate 0.8deg. Aspect derived from the ASTER DEM can deviate up to 31.2deg whereas, aspect computed from the SRTM DEM can deviate up to 29.8deg. The drainage networks computed from the ASTER, SRTM DEMs and Toposheet are also compared. The analysis revealed that the drainage network extracted from the SRTM DEM, ASTER DEM and Toposheet are matching.

Keywords: DEM; Monte Carlo simulation; Hydrological modeling; Accuracy assessment.

1. Introduction

Digital Elevation Models (DEMs) have been frequently and effectively utilized as one of the crucial databases in many GIS applications (Zhou and Liu, 2004). DEMs provide continuous representation of the earth's topography with varying resolutions and accuracies (Wechsler, 2007). DEMs are frequently used to derive topographic attributes, including aspect, slope, curvature etc. Often, DEMs are perceived as a true representation of the earth's surface, however, like many other spatial data, they are also prone to inherent errors. The inherent errors in a DEM are subsequently propagated in the computed topographic attributes. The influence of DEM errors on elevation and computed topographic attributes are often ignored and therefore might cause uncertainty in the decision making process (Wechsler, 2003).

Errors in a DEM are classified as general and specific errors. General errors are related to the technical problems during the DEM production. Specific errors are inherent errors in a DEM that cannot be removed and is publicized before systems launching and data provision (Wechsler, 2007). General errors include: systematic errors, blunders and random errors. Systematic errors are associated with the procedures followed for the DEM generation and follow fixed patterns that can cause uncertainty in final result. Blunders are vertical errors, associated with data collection process and can be generally identified and removed before the data is released. Random errors are spatially uncorrelated and remain in the data after blunders and systematic errors are rectified. Random errors result from accidental or unknown combination of inaccuracies and identification of these errors are beyond the control of the users.

These errors leads to uncertainty and are mentioned as root mean square error (RMSE) of the DEM. RMSE defines the vertical accuracy of the DEM (Rodriguez et al., 2013) and is influenced by the nature of terrain, with higher RMSE in rugged terrain and lower in relatively flat terrain (Raaflaub and Collins, 2006). Therefore, it is crucial to evaluate the accuracy of remote sensing derived DEMs, specifically in a rugged topography.

Accuracy of a DEM can be evaluated by comparing DEM derived elevation values with actual elevation values, called as refrence data that can be acquired with high precision field survey or a secondary source (Castrignano et al., 2006). The aim of this study is to evaluate the accuracy in the remote sensing derived SRTM, ASTER and Topographic DEMs and to evaluate the impact of these errors on the computed topographic attributes, including aspect and slope. Moreover, the influence of DEM resolution on the derived drainage network is also evaluated in the study.

2. Study area

The study area selected for this study is comprised of districts Karak, Kohat and Hangu (Fig. 1). The selected area covers 698 km2 and has a moderate topography with elevation values ranges from 400 m to 1520 m above sea level (ASL). It is located in a semi-arid climatic region having extreme temperature in both summer and winter. Temperatures in summer reaches up to 48degC whereas, in winter the temperature drops to 10degC.

3. Methodology

3.1. Data used

3.1.1.Advanced space borne thermal emission and reflection radiometer (ASTER)

Advanced Space borne Thermal Emission and Reflection Radiometer of the study area was acquired from the USGS website(http://gdem.ersdac.jspacesystems.or.jp/) (Fig. 2a). ASTER DEM was jointly released by the Ministry of Economy, Trade, and Industry (METI) of Japan and the NASA, USA in June 2009. The VNIR bands of the ASTER Imagery are consist of two telescopes, one nadir viewing with a three-band detector (3N) and the other backward viewing (3B) (27.7deg off-nadir) with a single band detector are used to generate DEM (Nikolakopoulos et al., 2006).

3.1.2. Shuttle radar topography mission (SRTM)

Shuttle Radar Topography Mission (SRTM) DEM of the study area was acquired from the Consortium for Spatial Information (CSI) of the Consultative Group for International Agricultural Research (CGIAR) GeoPortal website http://srtm.csi.cgiar.org/ (Fig. 2). SRTM has created an unparalleled data set of global elevations and is freely available for modeling and environmental applications with spatial resolution of 90 m. SRTM provides the first single-pass, space-borne Interferometric Synthetic Aperture Radar (InSAR) and records data over 80% of Earth's landmass from 60deg N to 56deg S latitude (Nikolakopoulos et al., 2006). The collected data is being processed by the Jet Propulsion Laboratory (JPL) having horizontal and vertical accuracy about +-16 m (Smith and Sandwell, 2003).

3.1.3.Toposheet

The topographic sheet of the study area at a scale of 1:50000 is acquired, which is developed by the Survey of Pakistan in 1998-99. Toposheet was subsequently used to digitize contours and streams as shown in Figure 3. Digitized contours were then interpolated to generate a DEM and was resampled to spatial resolution of 30 m.

3.2. Accuracy assessment of SRS DEM

The accuracy of a DEM is a mainly determined by precision of the source data (such as GPS, digitization of contour maps, automated or manual photo grammetric sampling), terrain nature, and the techqnique used for generation of the DEM surface (Gong et al., 2000). In this study, spot elevation extracted from the toposheet were used as a reference elevation data to evaluate accuracy of the SRTM and ASTER DEMs. The elevation for the same locations was also extracted from the ASTER and SRTM DEMs and analysed to calculated RMSE. RMSE of the DEMs were calculated using Equation 1.

(Equation)

Where N is the number of 35 sampled points that were distributed throughout the area and assumed to be the representation of terrain, xi is the spot height elevation and xi is the extracted point elevation data from ASTER and SRTM DEM. The RMSE derived from the accuracy assessment of the DEM sis subsequently used in the Monte Carlo Simulation to evaluate their impact on the computed topographic attributes.

3.3. Monte carlo simulation

Monte Carlo Simulation (MCS) is a stochastic technique and frequnetly used to evaluate the impact of errors in DEMs on computed topographic attributes (Shafique and van der Meijde, 2015). To apply MCS, elevation values of the DEM is assumed as one of the possible values of the accurate elevation. Many simulations were generated to evalute uncertainties in the DEM - computed topographic attributes through statistical evaluation of the distribution of realizations. MCS assumes that errors in a DEM are normally distributed with mean equal "0" and standard deviation equal to "RMSE" of DEM. Elevation errors are assumed to be spatially auto-correlated (Wechsler and Kroll, 2006). Spatially auto-correlated field is created by passing low pass 3x3 filters over random error map. Each pixel in the random error map is replaced by the mean of the surrounding eight pixels (Shafique, 2008).

By applying filter, spatial autocorrelation increases and standard deviation decreases. Error DEM is created by adding the random error map to thhe low pass filtered error map. This process is repeated 40 times to generate error DEMs. A new original DEM is generated from previous step. This DEM was used to generate topographic attributes such as slope and aspect from the SRTM and ASTER DEMs. Thus, from each simulation, a new DEM derived topographic attribute map is created. In order to quantify the uncertainity of each DEM, the random error topographic attribute map is subtracted from original topographic attribute map to evaluate the impact of DEM inherent errors on computed topographic attributes of slope and aspect.

4. Drainage pattern

Digital Elevation Models are often used to derive drainage network. Arc Hydro Tool, an extension of the ArcGIS software is used to extract drainage and watershed from the ASTER, SRTM and Toposheet derived DEMs. Sinks are the naturally occurring artifacts in a DEM and must be removed before the drainage extraction (Borough and MacDonnell, 1998). Subsequently, flow direction is computed which indicate the direction of the steepest descent from the processing cell. It has 8 distinct values; each value indicates the steepest downs lope surrounding eight neighbors also known as D8 algorithm or D[?]. It shows overall drainage pattern of study area. Subsequent to flow direction, the flow accumulation map is computed showing the number of pixels contributing flow in each pixel.

Pixels with zero accumulation are considered mountain, ridges or steep slope and pixels with higher accumulation are regarded as stream(Shafique e tal., 2014). Flow accumulation map is used as an input for extracting stream network. Denser stream network can be obtained by lowering the threshold value (Gopinath et al., 2014). The extracted streams are further classified according to the Strahler classification scheme. These drainage network is computed from SRTM, ASTER, and Topographic DEM and the impact of DEM resolution and accuracy on the derived drainage network is evaluated.

5. Results and discussion

5.1. Accuracy assessment

Accuracy assessment of the ASTER DEM shows RMSE of +-20 m and the SRTM DEM shows RMSE of +- 27 m. This shows that RMSE of the ASTER DEM falls within the predefined vertical accuracy, whereas, the accuracy of SRTM is higher than the global assessment that is +- 16 m.

5.2. Uncertainty from DEM - derived topographic attributes

Uncertainty in the slope and aspect computed from the ASTER and SRTM DEMs is evaluated using MCS (Table 1) . Terrain slope computed from the ASTER DEM can deviate 1.2deg and aspect can deviate up to 31.2deg (Fig. 4). Results derived from the SRTM DEM shows that slope can deviate 0.8deg and aspect can deviate up to 29.8deg (Fig. 5). Therefore, the SRTM DEM was found to be more consistent in computaion of topographic attributes.

Table 1. MCS derived uncertainty in slope and aspect computed from the ASTER and SRTM DEMs.

###DEM###Topographic###Minimum###Mean###Maximum###Standard

###attribute###Deviation

###ASTER###Slope###0.09###1.2###2.33###0.2

###Aspect###0.3###31.2###179###45.7

###SRTM###Slope###0.30###0.8###2.45###0.10

###Aspect###0.6###29.8###179.1###42.8

6. Drainage pattern analysis

After fill sinks, minimum elevation of the ASTER DEM is increased from 508 m to 509 m, for the SRTM DEM the minimum elevation is raised from 513 m to 514 m. Minimum elevation of both the SRTM and ASTER DEMs were raised because of smoothening and averaging effect. Flow direction computed from the ASTER and Toposheet DEMs shows that water drains from North towards South due to maximum accumulation of pixels. Whereas, SRTM shows that water enters from North and East and drains towards South. Number of pixels according to direction encoding and area regarding each direction is given in Table 2 and Figure 6.

In the flow accumulation maps, maximum number of pixels is recorded in interpolated topo DEM i.e. 772740 whereas, from the ASTER DEM and SRTM DEM shows 756896 and 86178 respectively as shown in Table 3. The toposheet is showing higher accumulation values than the ASTER and the SRTM DEM and hence showing greater number of streams. Subsequently, the ASTER DEM is showing dense stream network than SRTM DEM due to higher accumulation value. Flow accumulation map computed from the SRTM DEM is shown in Figure 7.

Table 2: Descriptive statistics for flow direction from the ASTER and SRTM DEM and Toposheet.

###Direction###Number of pixels###% of Area

###Encoding

###Toposheet SRTM###ASTER###Toposheet SRTM###ASTER

###1 East###94563###14878###105589###12###17.3###14

###2 South-East###101986###9420###80762###13###11.1###11

###4 South###170472###19712###157379###22###23###21

###8 South-West###76933###6497###72141###10###7.5###9

###16 West###53089###8903###82314###7###10.3###11

###32 North-West###54698###4436###53301###7.0###5.1###7

###64 North###128685###14986###130736###17###17.3###17

###128 North-East###92314###7346###74674###12###8.5###10

The pixels with flow accumulation of >1 Km2 were extracted as stream network (Fig. 9) and reclassified according to Strahler stream order (Table 4 ). The variation in the streams extracted from the ASTER, SRTM and Topographic DEM is shown in Figure 8 and Figure 9.

Table 3. Number of pixels with corresponding area from the ASTER, SRTM DEM, and Toposheet.

###DEMs###Number of###Area in

###Pixels###km2

###ASTER###756896###681

###SRTM###86178###698

###Toposheet###772740###695

Table 4. Stream length and number of streams according to Strahler stream order.

###Number of Streams###Length of Streams

###Strahler###ASTER SRTM Toposheet###ASTER###SRTM###Toposheet

###Stream

###Order

###1st Order###297###271###433###276###247###241

###2nd Order###124###132###247###100###109###100

###3rd Order###131###125###151###94.3###95###94.3

7. Impact of DEM resolution on terrain representation

The impact of DEM resolution on the terrain representation is assessed graphically through horizontal elevation profile of 7 km from the ASTER, SRT and Topo DEMs (Fig. 10). The impact of DEM resolution on smoothening, shape and height is significant on terrain features with a base width smaller than the respective DEM resolution. Topographic features larger than the DEM resolution, but still smaller than the neighborhood size, will be suppressed and smoothened during the topographic attribute computation.

8. Conclusion

This study evaluate the impact of DEM resolution and vertical errors on computed topographic attributes and drainage pattern. Accuracy of remote sensing based ASTER and SRTM DEMs is assessed by comparing with the spot heights from Toposheet. In the study area the SRTM DEM has an RMSE of +-27.3 m and the ASTER DEM shows RMSE of +- 20 m. These errors in the DEMs are also due to the random errors that also propagate in the DEM- computed topographic attributes. The impact of random errors in ASTER and SRTM DEM is computed through MCS. Analysis revealed that aspect derived from the ASTER DEM is prone to an error of 31.2deg and from the SRTM DEM is susceptible to an error of 29.8deg. Slope derived from the ASTER DEM has an uncertainty of 1.2deg and from the SRTM DEM has an uncertainty of 0.8deg.

So, the SRTM DEM was found to be more consistent in deriving topographic parameters than the ASTER DEM. Toposheet, ASTER and SRTM DEM were utilized to compute stream network. The automatic extraction of watershed delineation depends on the accuracy and resolution of DEM. The resolution of the DEM has sigificant impacts on the terrain reperesentation mainly due to smootherning and averging effect.

References

Castrignano, A., Buttafuoco, G., Comolli, R., Ballabio, C., 2006. Accuracy assessment of digital elevation model using stochastic simulation. 7th International Symposium on Spatial Accuracy Assessment in Natural Resources and Environmental Sciences, Citeseer.

Gong, J., Li, Z., Zhu, Q. Sui, H., Zhou, Y., 2000. Effects of various factors on the accuracy of DEMs: an intensive experimental investigation. Photogrammetric Engineering and Remote Sensing, 66(9), 1113-1117.

Gopinath, G., Swetha, T., Ashitha, M., 2014. Automated extraction of watershed boundary and drainage network from SRTM and comparison with Survey of India toposheet. Arabian Journal of Geosciences, 7(7), 2625-2632.

Nikolakopoulos, K. G., Kamaratakis, E. K., Chrysoulakis, N., 2006. SRTM vs ASTER elevation products: Comparison for two regions in Crete, Greece. International Journal of Remote Sensing, 27(21), 4819-4838.

Raaflaub, L. D., Collins, M. J., 2006. The effect of error in gridded digital elevation models on the estimation of topographic parameters. Environmental Modelling and Software, 21(5), 710-732.

Rodriguez, E., Morris, C. S. J., Belz, E., 2013. A Global Assessment Of SRTM Performance. American Society for Photogrammetry and Remote Sensing, 72(3), 249-260.

Shafique, M. (2008). Predicting topographic aggravation of seismic ground shaking using geospatial tools (A case study of Kashmir Earthquake, Pakistan). Msc, International Institute of Geo-Information Science and Earth Observation Enschede.

Shafique, M., Israr, S., Shah, M. T., Khan, M. A., 2014. Remote sensing based strategy of stream sediment sampling for mineral exploration in Peshawar Basin, Khyber Pakhtunkhwa, Pakistan. Journal of Himalayan Earth Sciences, 47(2), 141-148.

Shafique, M., Van der Meijde, M., 2015. Impact of uncertainty in remote sensing DEMs on topographic amplification of seismic response and Vs 30. Arabian Journal of Geosciences, 8(4), 2237-2245.

Smith, B., Sandwell, D., 2003. Accuracy and resolution of shuttle radar topography mission data. Geophysical Research Letters, 30(9), 1467.

Wechsler, S., 2007. Uncertainties associated with digital elevation models for hydrologic applications: are view. Hydrology and Earth System Sciences, 11(4), 1481-1500.

Wechsler, S. P., 2003. Perceptions of Digital Elevation Model Uncertainty by DEM Users. URISA Journal, 15, 57-64.

Wechsler, S. P., Kroll, C. N., 2006. Quantifying DEM Uncertainty and its Effect on Topographic Parameters. Photogrammetric Engineering and Remote Sensing, 72(9), 1081-1090.

Zhou, Q., X. Liu., 2004. Analysis of errors of derived slope and aspect related to DEM data properties. Computers and Geosciences, 30(4), 369-378.
COPYRIGHT 2016 Asianet-Pakistan
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

 
Article Details
Printer friendly Cite/link Email Feedback
Publication:Journal of Himalayan Earth Sciences
Date:Dec 31, 2016
Words:3225
Previous Article:Landslides induced vulnerability and risk assessment in Muzaffarabad and Balakot, Pakistan.
Next Article:Seismic Hazard Assessment by Preparing S1 and Ss maps for Baluchistan Province, Pakistan.
Topics:

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters