Assessing the Accuracy of DEMs Derived from Contour Line Maps and Comparing with ASTER DEMs

Introduction
DEM accuracy properties including grid size could however vary from source to source for an area of interest.The sensitivity of principal topographic derivatives used in hydrologic modeling to DEM resolution, however has been systematically explored in few studies.A medium and large scale topographic map is one source of digital elevation models in many countries. In Iran GIS users digitize the1:50000scale maps to constructing of DEMs with10×10meters resolution.However these data are used for scientific analysis but the accuracy and quality of them is not completely known.Therefore in this paper we first examined the accuracy of a DEM derived from topographic maps and then investigated its quality by comparison the results with results derived from an ASTER DEM.

Methodology
The materials of this paper included the contour line map with1:50000 scale, IRSpan satellite image with5.8×5.8 m resolution and an ASTER DEM from Terra satellite in2003. We first transformed the scanned map to UTM coordinate system and then digitized the contour lines by ArcGIS Editor Tools.A TIN map was constructed from digitized contours,then transformed to a DEM with 10m meters cell size as we called it TOPO DEM. The river network of the study area was firstdrawn from IRS pan satellite Image by ArcGIS sketch tools after geometric transformation of image in UTM coordinate system. The criterion used to define first-order stream channels was based on channel morphology and a length of over50m.The extraction of the drainage network of the study area carried out from both DEMs,in raster format with a 10m*10m grid cell size for theTOPO DEM and30x30m for ASTER DEM.Hydrology toolsets in ArcGIS software, version9.3 (ESRI2008)was used for extraction of drainage system.
For determination of TOPODEM accuracy we took the extracted networks from satellite images as a real stream channels. This is partly because more detailed scale of the satellite images guarantees a good reference map to compare the network obtained from the DEM. The comparison process has been done in both raster and vector formats. These comparisons included morphometric characteristics as river frequency, stream length, stream density and drainage ratio as well as the spatial pattern of the drainage lines, which was evaluated by visual analysis and calculating the differences rate between two networks. We used the drainage network characteristics extracted from ASTER DEM as a criterion to assessing the quality of the TOPO DEM.
Study Area
The study area with 443 Km2 is located in the east part of Iran(Figure2)which consists of4geomorphologic units including mountain, pediment, alluvial fans and flood plain. Highest elevation in the area is about2200 m in the northeast corner and the lowest elevation at the southwest part is about 1000m. Mountainous area with199 km2 is the main geomorphologic unit which is characterized with dendretic drainage pattern.The Elevation is ranging between1000m in the flood plain area to 2200 m in the mountainous sections.

Results and Discussion
A)DEM accuracy in comparison with real ground
Visual study and numerical results showed only a network Map which is derived in 50 threshold value is near to the network extracted from satellite image. River order and cell frequencies are being decreasing in higher threshold values. The comparison of the cell and vector data presented good agreement for all stream orders but there are large differences for 4, 5 and sixth-order streams.
The results showed better agreement in the mountainous and hilly parts only for the first and second-order streams. The river length measurements show satisfied agreement degree for pediment and mountain units especially for second-order rivers. We found the lower agreement degrees for alluvial fans and plain sectors.
Drainage density calculated for whole basin area shows a good agreement between rivers, which extracted by different methods. Obviously, differences appear in separate Geomorphologic units as we can see completely incorrect results for alluvial fan and plain sectors for rivers which are delineated from DEM. The best agreement of drainage density is recognized for the pediment where the surface slop is moderate. In the mountain area drainage density for the layer extracted from satellite image is higher than the layer extracted from TOPODEM while the results are inversely for alluvial fans and plain areas. The best agreement of drainage density is recognized for pediment where the surface slop is moderate.
B)DEM quality in comparison with an ASTER DEM
In this case we used an ASTER DEM for assessing criteria;both DEMs must to be in a same cell size, therefore the TOPODEM resampled to30×30 meters dimensions. Based on stream frequency calculations we found out only small differences between two DEMs.Stream frequency for all stream orders in threshold 50 is completely equal for two DEMs.In other thresholds also the amount of the stream frequency explained high quality of the DEM. The differences between two DEMs were arising when we derived the river length in different thresholds. The lowest differences are associated to threshold of50while it increases for higher or lower thresholds. There are recognizable differences at the some river orders, for instances we can refer to the rivers with 1 order.

Conclusion
The DEM based on1:50000 topographic maps have enough accuracy for extracting of the first and second-order rivers in mountain areas with50 threshold values. The morphmetric analysis based on extracted networks from these DEMs give weak accuracy through alluvial fan and plain areas but there is a good agreement for river length in the mountain,hill and pediment landforms.10*10 in resolution of TOPO DEMs is not enough to derive river networks in the alluvial fan and plain sectors. It is possible to increase the threshold values for network extraction on the alluvial fan and plain sectors. However the comparison of TOPODEM data with real ground illustrates a weakly coincidences but in comparing with a DEM from satellite sources resulted good quality in the data group. We found out large similarities between TOPODEM and ASTER DEM in drainage network analysis. The differences rate in all morphometric parameters and at different geomorphologic units were very low. It means the TOPODEM data has a high quality and GIS users can use these data with high confidence.