- Project report
In the last decades, topographic records from laser scanning have become an indispensable tool in field studies and provide a detailed shape of the topography by a data density of several points per square meters as well as an accuracy of levels better than 150 mm (Mandlburger et al., 2009). The survey data allows the generation of a precise Digital Terrain Model (DTM), the basic topographic data to provide the initial terrain input to depict the area of interest. The dataset is being used in detection of hydrological processes, changes in riverine topography and for precisely modeling of flood inundation extents. More detailed topographic data can be achieved by fixed-position terrestrial laser scanning (TLS). TLS is a reliable survey technique for the detection of 3D objects and its geometric structures with a large amount of detailed surface elevation points. Several types of detection methods are used in fluvial studies. In the study of Wang et al. (2012), different DTMs (ASTER GDEM, SRTM) were utilized for hydraulic modeling of a glacial lake outburst flood in Southeast Tibet. They show that for this purpose ASTER GDEM and SRTM can provide valuable DTM sources when data of higher resolution is not available.
In this paper, we will evaluate the applicability of different DTMs for a one-dimensional hydraulic modeling of a floodplain area. First, we give a short overview of the project. Second, we describe the preparatory work steps needed to create a hydraulic model. Third, we present and compare the different floodplains and inundation depths. Finally, we discuss the DTMs’ suitability regarding floodplain modeling.
In June 2013, an enormous flood event was recorded on Danube, Elbe and Saale. The Neckar in Baden-Württemberg was also affected but with lower degree. In figure 6.1 one can see pictures of the study site during the flood event and afterwards. Nevertheless, the occurrence of high water level in the region enables potentially the construction of a hydraulic model. Such models have a great necessity in fluvial studies, particularly in the mapping and estimation of flood inundation extents (Hohenthal et al., 2011). Furthermore, it is useful in flood protection and prevention.
In this study, we compare different Digital Terrain Models (DTM) and their effect respectively suitability on hydraulic modeling. Therefore we use SRTM-, ASTER-, Airborne Laser scanning (ALS)- and Terrestrial Laser scanning (TLS)-data.
Figure 6.2 illustrates the general workflow of this study and is described in more detail in the following. First of all, it is necessary to document the flood event, which we want to model afterwards. Therefore, on 3rd June 2013 a photo documentation and determination of reference points were performed.
Discharge data of the barrage Ziegelhausen and a hydrograph of Ladenburg were obtained from Wasser- und Schifffahrtsamt Heidelberg.
Since the aim of the study is to compare the suitability of DTMs with different resolution for hydraulic modeling, SRTM (90x90 m), ASTER GDEM (30x30 m), ALS (1x1 m) and TLS data were acquired. However, the ALS data does not fit on aerial images and the other DTMs. Therefore it was shifted about 110 m to the south and 70 m to the east. Afterwards, the ALS-DTM was resampled to 10x10 m to get another DTM at an intermediate level. All DTMs then were cut to the size of our study site to accelerate the processing time. As a last preparation step before modeling, a landscape model was generated by digitizing the aerial image provided by ESRI's ArcGIS.
To process the gathered data, computations were run with ArcGIS. As a first step, all terrain lower or equal to 98 m was detected (Raster calculator: DTM≤98). We chose 98 m as the water level for the simulated flood event, because it is the same height like the reference points we had detected previously. Now all the terrain below the water level is detected, even the parts which are not connected with the river and therefore are not flooded. For this purpose polygons were created from the raster and afterwards selected. Therefore we used a river-vector layer which we extracted from the landscape model. As a result all the area which is below 98 m and in contact with the river is detected as flooded area. This step was performed with the SRTM, ASTER, ALS and ALS10m data. An intermediate step was necessary for TLS data, because there are parts in the DTM which have no data like the river itself. First we detected all noData-values and filled them with 91,5018, the height of the DTM’s lowest point. After that we could proceed like with the other DTMs.
To compare all the different floodplains, clipped the flood extent layers were clipped to get an identical maximum extent. Subsequently we calculated the area of floodplain (all layers are in the same projection). As a last step the inundation depths were calculated (Raster Calculator: 98-DTM). Thereafter we reclassified the data to 1m intervals. Again, area which is below 98 m but actually not in contact with the Neckar and consequently not flooded, is displayed. Therefore the raster files were converted to polygons, clipped with the flood extent-layer and converted back to raster data. Hence we get maps of the inundation depths based on the different DTMs as a final result.
In the following we present the results of the analysis described above. Based on SRTM data the floodplain extends over 54.4% of the survey area. As you can see in figure 6.3 a, the river course is not well displayed in the flooded area in the southwestern part. The maximum inundation depth is not more than 7 m. The ASTER result shows a quite different extent. The northwestern part is flooded and higher depths are reached. However, a part of the Neckar course is also missing. Overall 87.1% of the area are flooded. Both results do not correspond to the extend we found in June, especially in the southeastern part, which was not flooded at all.
Regarding the ALS data, both floodplains are nearly identical (fig. 6.4). They cover 47.3% of the survey area and trace the course of the river properly. The inundation area is comparable to the real event.
The last analysis based on TLS data is the only one on which one can detect the gravel bar from the lower inundations depths in contrast to the river. Due to missing data in the original DTM and the further processing described above, the southeastern part is shown as a flooded area. Because of the smaller expansion of the DTM, we did not calculate the flooded area.
An overview of the different DTMs and the floodplains is given in table 6.1.
|Source||USGS||METI & NASA||LGL BW, processed||LGL BW||Hydro-Change³|
As seen in the chapter Results, the accuracy of the computed floodplain is increasing with higher resolution of the DTMs. However, the ASTER DTM produces the worst results although it has a higher resolution than SRTM (30x30 m vs. 90x90 m). But both of them do not represent the inundation area of the flood event from June 2013 correctly. So even though they are freely available, they are not appropriate for hydraulic modeling in the scale of our study. Regarding a greater extent, ALS-data is the most proper solution. Both resolutions used in the scale of our study (10x10 m and 1x1 m) create barely different result. Thus ALS data – even with a spatial resolution of 10x10 m- should be used in studies regarding floodings. If one performs a study in greater detail, the TLS-data, which is the most elaborated to gather, will bring you the most detailed results. Because of its high resolution of 25x25 cm even little levees and barriers or depressions are represented in the DTM. In case of an inundation event these little differences could determine whether an area will be flooded or not. But because of the high costs and time consuming work, ALS data could be used to get an overview of the whole area and only important parts could then be scanned with TLS in more detail.
In this study we computed the inundation area and depths based on a known water level. For this purpose simple GIS software like ArcGIS is adequate. But as soon as the water level is not and only the discharge is known, modeling software is necessary.
To detect fluvial geomorphology high-resolution light detection and ranging altimetry data gathered by ALS and TLS has shown great potential for further studies. The field of application is extensive and will enhance in future years.