Accuracy Assessment of Digital Elevation Models Produced From Different Geomatics Data

), ASTER V2 (3.815m), and PHR (0.433m). The results showed that the ALOS V1.1 model is the most accurate of the open source models followed by the SRTM V3 model and then followed by ASTER V2. The results obtained from a pair by Pleiades high-resolution (PHR) 1B satellites show a higher accuracy than the results obtained from the open source models.


INTRODUCTION
Digital elevation models (DEMs) derived from remote sensing data provide a valuable and consistent data source for mapping, terrain visualization, telecommunication, navigation, disaster management, planning of civil engineering infrastructures, and orthorectification of airborne and satellite imagery.The DEM could be obtained utilizing technologies such as aerial stereo photogrammetry, airborne light detection and range detection (LiDAR), interferometric synthetic aperture radar (InSAR), and land surveying.Because of the high cost of producing digital elevation models by conventional ground surveys and the inaccessibility of some places due to the roughness of the terrain and the seriousness of the areas (the presence of military waste), it has become necessary to research the evaluation of the results of the less expensive and safer digital elevation models.Four digital elevation models from different Geomatics sources were evaluated in this paper.
The Global Elevation Data Set Shuttle Radar Topography Mission (SRTM V3) was a joint mission by the National Imagery and Mapping Agency (NIMA) and the National Aeronautics and Space Administration (NASA) [1].STRM V3 provides the Earth's highest open DEM resolution.It is based on the Interferometry Synthetic Aperture Radar (SAR) or Interferometric Synthetic Aperture Radar (InSAR) standard, which uses phase-difference estimates obtained from two radar images.The Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER V2) is an international collaboration venture between Japan's Ministry of Economy Trade and Industry (METI), and the National Aeronautics and Space Administration (NASA) of the United States.Nearinfrared stereo imagery is obtained simultaneously with a long-track synchronization at both nadir and off nadir angles.Ninety Nine percent (99 %) of Earth's land mass is covered by the ASTER DEM.The spatial resolution of stereo image is 15 m and the created DEM is 30 m. Advanced Land Observing Satellite (ALOS) has a 30 m resolution.DEM was created after using the archived information about the Panchromatic Remote Sensing Instrument for Stereo Mapping DEM.The spatial resolution of stereo image is 15 m and the created DEM is 30 m. Advanced Land Observing Satellite (ALOS) has a 30 m resolution, DEM was created after using the archived information about the Panchromatic Remote Sensing Instrument for Stereo Mapping PRISM), portable on satellite ALOS by Japan Aerospace Exploration Agency's (JAXA) in year 2006.PRISM consisted of three panchromatic radiometers that provided stereo images along a track with spatial resolution of 2.5 m in the nadir-looking [1].By searching, we will try to indicate the accuracy of each method and the possibility of increasing the accuracy depending on the mathematical methods and algorithms.To achieve the main objective, it is an evaluation of the results of these methods after comparing them with the most probable value and according to specifications.Many researchers gave their advice in this field and in different regions of the world where we can pass these experiments because of their importance in giving the research reliability.
In the work of Nasir et al. [2], 15 cm accuracy of LiDAR points were adopted as a level to compare the different sources of DEM, the open source model, ASTER 30m DEM, SRTM 90m DEM, were used with DEM generated from Pleiades Tri Stereo-pair imagery possess 0.5m spatial resolution.The comparison between Pleiades-10 m DEM and LiDAR point elevation output were RMSE 5.2m, the comparison of ASTER-30m & SRTM-90m DEMs with the Level of comparison RMSE 6.65 as and 7.5 m respectively [2].
Twenty five checkpoints that were collected from 30 cm resolution HGK orthophotos, were used by Alganci et al. [3], to compare the different sources of DEM.DEM evaluated were; ASTER 30m DEM, (SRTM) 30m DEM, (ALOS) 30m resolution DEM, 3 m and 1 m resolution DSMs were produced from tri-stereo images from the SPOT 6 and Pleiades high-resolution (PHR) 1A satellites, respectively.The results of RMSE was PHR DSM(1.57),ALOS(2.14),SPOT DSM(2.26),SRTM(3.53) and ASTER(5.72) of the comparison showed that the DEM produced from PHR is closer to the real value than the rest of the results [3].

THE STUDY AREA
The study area is located East of the Amarah city (Maysan governorate) in the south of Iraq which is outlying the capital (Baghdad), 300-kilometer south.The study area is about 30 Km2 located between (47.40°,47.46°)East and (32.31°,32.38°)North close to the border line between Iraq and Iran.The current study area was chosen due to the availability of data (PHR) and the presence of variation in the terrain, which gives sample room for comparison in contrast to flat terrain.The region consists of a group of hills abandoned by some of the valleys.There are pools of water, especially in the rainy and flood season, and a fertile environment for the growth of seasonal natural plants in it.It contains military wastes from the Iran-Iraq War era in the 1980s and it include oil fields of the Maysan Oil Company.The area can be classified as Primarily Non-Vegetated because of an absence of Woody or Herbaceous life forms and with less than 25% cover of Lichens/Mosses [6].

II. Methodology
In order to evaluate the vertical accuracy, an independent, high-precision source must be selected in accordance with ASPRS standards.The independent source of higher accuracy for QA/QC check points should be at least three times more accurate than the required accuracy of the geospatial dataset being tested.(RTK/GNSS) technique to achieve these criteria was used in this study.The steps are summarized by the following: A. More than thirty check points were observed by using dual frequency GNSS using Real Time Kinematic (RTK) observation technique.The horizontal accuracy was 10mm±ppm and a vertical accuracy was 15 mm± 1.0 ppm.The base point for the RTK measurement is a boundary pillar obtained from the State Commission of Survey.Its coordinates were adopted as is, without static survey, see Figure 6.The observed points (RTK check point) were collected from the field surveys of the same researcher.The collected points were divided in to two groups.The first group were not clear objects on the ground and used as a checkpoint for vertical accuracy assessment because of check point normally not well-defined [7].It was 30 checkpoints (sample size(, see Figure 7.The second group were clear features (well defined) on the ground.The second group were used to carry out the geometrical correction for the images of Pleiades during the process of processing and extraction of DEM.The horizontal and vertical coordinates of each point were referenced to UTM, WGS84, and EGM96.

I. The evaluation criteria according to the standards
The National Standard for Spatial Data Accuracy (NSSDA) uses root mean square error (RMSE) to estimate positional accuracy and an indicator of evaluation.RMSE is the square root of the average set of square variations between coordinate values and coordinate values for similar points from an independent source with higher accuracy [7] (RTK) observation.It is an independent, highresource used in this study.
n=The number of samples (points). Indicates accuracy as in equation ( 2)

II. Bias Correction
To reduce the variation between the observed Z ( Z GNSS ) and the calculated Z( Z DEM ), a linear transformation function will be used to derive bias corrected of the DEMs elevation as follows [8]: Z is observation data, w is estimation data, u, v are UTM coordinate of estimation, (, , c, 0) are linear transformation coefficients.Apply the equation ( 3) on the first point until n Points 1=.1+.1+.1+ 0…… (3a) 2=.2+.2+.2+ 0…… (3b) n=.n+.n+.n+ 0…… (3c) Summarize the equation (3a) to (3b) as follows: The sum of the squares of deviation is given by: The function S (a, b, c, Z0) is a minimum when Dividing equation ( 6) by 2 and rearranging yields the normal equation then it solved as matrix system =……… ……………….(… .(7) The above system has been mathematically resolved and found the value of the matrix [X], the direct linear transformation coefficients for each type of DEM were calculated as in Table 3.

I. Absolute Error value
As shown in Figure 10, the maximum errors recorded for PHR and ASTER DEM were 1.41 m and 20.88 m respectively; while the minimum errors recorded for PHR and ALOS DEM were 0.02 m and 0.07 m respectively.

II. Analysis of the results under the standards
According to the generic ASPRS, 2014 vertical accuracy standards for digital elevation data,enabling an unlimited number of vertical accuracy classes for non-vegetated vertical accuracy(NVA) and vegetated vertical accuracy(VVA) , as shown in table4.

III. Statistical Analysis
There are many statistical tests that give an indication of the normal sample distribution, Kolmogorov-Smirnova and Shapiro-Wilk test are used, under confidence level of 95%, the null hypothesis (H0) (samples with normal distribution).The null hypothesis was rejected in the test Shapiro-Wilk to the ASTERV2 DEM error (Data are not subject to normal distribution ( because the level of significance below 5%; see Figure12 and Table 5.The null hypothesis (H0) for DEM error Pleiades, SRTM and ALOS were accepted.Therefore, data are subject to normal distribution, see Figure 11, 13, 14, and Table 5.

CONCLUSIONS
1) The results of the preliminary data analysis showed that the digital elevation model ALOS V1.1 gave the highest accuracy among the free models accurately 1.65m, followed by SRTM V3 model 4.37m and then ASTER model 29.36 m.
2) The results of the corrections using direct linear transformation method was mixed.The model most responsive to corrections was ASTER V2 DEM (the value of RMSE is 15.12 m was decrease to 3.82 m).Next comes the model SRTM V3 DEM (the value of RMSE is 2.23 m was decrease to 1.32m).Then the PHR DEM (the value of RMSE is 0.64 m was decrease to 0.43 m), ALOS V1.1 DEM gives a poor response to the correction process using the direct linear transformation method from (RMSE 0.84 m to RMSE 0.83 m). 3) DEM extract from Pleiades high-resolution (PHR) 1B satellite gave high accuracy compared to open source models, the results were logical because of the high resolution of the images used as well as the adoption of ground control points during image processing.4) As mentioned at the end of the introduction, most literature review showed that the digital elevation models produced by the Pleiades satellite are the most accurate compared to other free models.The ALOS model tops the free models with the highest accuracy, then the SRTM model follows.The ASTER model is less accurate than the other models.The obtained results agree with previous studies, which gives them reliability in obtaining data and processing processes.

Figure 1
Figure 1: Case Study Location

Figure 6 :Figure 7 :
Figure 6: Boundary pillar used as base point for the RTK measurement

C
. Extract Z-Value from the different datasets of DEMs.Four basic steps were used to extract the value of elevation (Z DEM) using GIS environment, Enter the different layer of the DEMs simultaneously with creation of shapefile(shp*) that contains thirty Ground control points information (Enter and data management stage).Matching layers, DEM layers and feature point layer, use Spatial Analyst Tools to extract the Z value as in Figure (9) and Table2.

Figure 10 :
Figure 10: Minimum and maximum value of the error (Absolute value)