Hide Media Panel
Your Browser does not support WebGL. Try the newest version of Firefox or Google Chrome.
You should see an interactive map here. However, your browser seems to not support JavaScript or you have JavaScript disabled for this web site. Please enable JavaScript or get a modern web browser like Firefox or Google Chrome.

Multitemporal Fine Registration and Volume Balance Series


A multitemporal analysis was performed in order to identify the responsible processes leading to a change of the gravel bar in the given time period. The analysis is based on the available datasets acquired during the measurement campaigns performed in November 2011, June 2012 and July 2013. The data gathered in July 2013 was registered and finegeoreferenced with RiSCAN Pro v1.7.7 and a dedicated Python Script. All datasets are processed into DTM by the same Python script using OPALS modules in order to ensure comparability. Within OPALS difference models and calculations of volume balances are performed in order to identify the vertical difference between the DTM of the different years. Fig. 5.1 illustrates the workflow with its steps and the required software.

Earlier multitemporal analyses of the time period of 2011 and 2012 have already identified ongoing accumulation and erosion processes affecting the gravel bar. The longer time period between the measurements in 2012 and 2013 leads to the expectation that great changes can be analyzed than in the period between 2011 and 2012.

Workflow of data processing and the multitemporal analysis. (Source: own illustration)


Due to the different positions of the laser scanner during the measurements, the gathered data has to be realigned, adjusted to each other and located within a global coordination system afterwards as shown in Fig. 5.2. In order to do so, the scan positions are registered and linked spatially to each other. The registration of the raw data was carried out using RiSCAN Pro v1.7.7.

Registration and georeference. (Source: Riegl 2009)

During the field campaign GPS data of each scan position has been acquired using a total station. The global coordinate of scan position 6 has not been calibrated during the field campaign. Therefore, the gathered data has to be registered by a method called ‘course registration’.

Scan positions and reflector positions of the campaign 2013

The registration of terrestrial laser data is a very important step and essential for further successful analysis (Müller).

Coarse Registration via Corresponding Points

As scan 6 lacks the fine scan of the reflectors the registration for scan position 6 cannot be carried out using tie points. Corresponding points are defined manually between the registered scan position 1 and the dataset of the scan position 6. Five point pairs were selected on top of the reflectors. A standard deviation of 0.019109 could be achieved.


Orientation via Backsight Orientation (GLCS)

To combine various datasets from different sources in a global coordinate system, georeferencing is the essential process. In order to achieve a common structure of the raw data, a geodetic surveying of all scan positions is necessary.

Backsight Orientation is a frequently used method for georeferencing different datasets. Scan position 1 was used to be a so-called master position. Every additional dataset has been adapted to this position via tie points.

The realignment of all single scan coordinates into the global coordinate system using scan position 1 as master position is achieved via registration using tie points.

Multi Station Adjustment (MSA)

A Multi Station Adjustment is being conducted in order to increase the accuracy of the registration of the scans.

Scan position 1 was locked because the backsight orientation was processed on this scan. The MSA was first processed with scan positions 1 to 5 in order to achieve a higher accuracy when searching for tie points within the coarse registration. After the coarse registration of Scan 6 another MSA was carried out. Different MSA parameters were tested and resulted in satisfying values regarding standard deviation, polydata and a histogram. However, Scan 6 did not match with the other scans. For that reason a MSA for Scan 6 was not processed because the position accuracy after the coarse registration without MSA was better than after the MSA (Riegl, 2012).

For the MSA of Scans 1 to 5 different MSA parameters were tested with regard to changes in the standard deviation, the number of used polydata and the distribution in the histogram. For the final MSA the parameters shown in table 5.1 have been used. Table 5.2 shows the statistics that are the used polydata and the achieved standard deviation.

Used MSA parameters
Nearest Point Search
Modeall nearest points
Search Radius [m]3
Max. tilt angle [deg]3
Min. change of error 1 [m]0.005
Min. change of error 2 [m]0.0005
Outlier threshold [1]1.3
Calulation modeleast square fitting
Update displayseldom
Error (StdDev)[m]0.0125
Number of observations used for calculation
Scan pos.s:0

Fine Georeferencing: 3D Correction

To perfom further analysis steps in OPALS or python, the dataset has to be exported from RiSCAN and converted into ASCII files. The data is exported as txt-format.

All parameters are exported in the following column order: X-value [m], Y-value [m], Z-value [m], Range [m], R [0-255], G [0-255], B [0-255], Amplitude [0-255], Reflectance [dB].

The fine georeferencing is performed by means of a Python script written by Bernhard Höfle. In order to align the data gathered in 2013 with the data gathered in 2012, the script uses pairs of coordinates from both datasets and aligns the content with each other (coordinate pairs of 2013 in the left column, coordinate pairs of 2012 in the right column).

The corresponding point search is realized manually in RiSCANPRO. The aim is to find constant points that should not have changed during the flood events in May and June 2013 and generally over the time period between the campaigns of 2012 and 2013. In total, six 3D-coordinates were chosen. Four are located on the west bank of the river Neckar, close to the street. Each point has to be selected in both point clouds (2012 and 2013) from fixed objects such as an utility pole, a house, a traffic stake near the total station, a street lamp and a road sign. At the side of the sand bank no points with a constant position could be found. The reason for these findings is the growing vegetation that partially blocks the view and also leads to changed positions of branches of trees between 2012 and 2013. Some of the branches that could be found in the data of 2012 could not be found in the point cloud of 2013.

Trees that were not affected by the flood event or by any other changes had to be examined. Corresponding points could be found in a crutch of an old tree.

The 3D correction also requires the dataset that has to be adjusted. In OPALS the scans are processed into an .odm-file first and converted into a xyz-file after. The order of the command line parameters for the transformation is “coordinate pairs of both point clouds”, “point cloud to be adjusted” “output (adjusted point cloud)”. Corresponding command lines can be found in 5.7 Appendix 1.

Impact of different Cell Sizes on the DTM and the calculated Volume Balance

The DTM can be generated with different cell sizes. To be able to compare the DTM of 2013 to the DTMs of 2012 and 2011, the DTMs of the current study were generated with a cell size of 25 cm. However, the impact of DTM cell size on volume balance needs to be assessed. Based on the datasets of 2011 and 2012 an analysis of the effects of different cell sizes on the resulting volume balance was realized.

For this process the DTMs, as well as the pond which was identified in the dataset of 2012, were clipped. The gravel bar without surrounding vegetation was chosen as study area and the water boundary was erased to prevent errors from laser points in the low water. The DTM as well as the pond which was identified in the dataset of were clipped in ArcGIS with the tool “extract by mask”.

For the datasets of 2011 and 2012 points from the point cloud were selected within a distance of 5 cm, 10 cm and 15 cm to the particular DTM. This way DTMs with different cell sizes could be created without having to go through the whole process of DTM creation. The processing was conducted in OPALS, the Command Lines can be found in the 5.7 Appendix 2. First a new column was created (NormalizedZ) and the relative height of all data points with respect to the terrain model was derived. Then, all points within the particular distances (5 cm, 10 cm, 15 cm) were filtered.

Thus DTMs with the cell sizes 10 cm, 15 cm, 25 cm and 50 cm were created using OpalsCell using the minimum of the points, followed by the creation of particular difference models and eventually a calculation of volume balances. For each cellsize and for each distance difference models were calculated between 2011 and 2012 to then identify the volume balance changes dependent to these parameters. OpalsDiff was used for the processing using bilinear interpolation.

The volume balance calculation is based on the following formula:

Number of pixels * Pixel size * Mean of height difference model

More detailed information about the use of opalsDiff can be found in chapter 5.5.2. All essential values for this calculation are taken from ArcGIS. Table 5.3 shows the resulting volume balances in relation to the cell size (row) and distance to the base DTM (column).

Volume balances of the difference model of 2011/2012 in cm³ in relation to cell size and distance to the base DTM.
Cell Size/Distance DTM 5 cm10 cm15 cm
1 cm -5,20 2,90 1,79
2 cm -0,70 -11,66 -17,04
3 cm -26,14 -47,68 -57,84
4 cm -56,01 -85,83 -99,51
5 cm -82,03-117,37-132,52
10 cm -159,47-197,17-211,91
15 cm -193,92-226,48-238,49
25 cm -236,23-256,05-260,57
50 cm -260,73-265,93-266,39
500 cm-367,59-415,65-431,54

Table 5.3 points out that the difference volume increases with increasing cell sizes. It can be assumed that relocation processes are captured more coarse-grained with larger cell sizes and thus are roughly over- or underestimated. It has to be considered that the point density decreases with smaller cell sizes and the calculated DTMs and thus difference models are incomplete which leads to greater uncertainties. On the other hand large cell sizes cause strongly generalised DTMs.

Similar effects can be observed with different distances to the base DTM: The greater the distances of the selected points to the DTM, the greater the vertical variation of the resulting DTM and calculated volume balances. It has to be considered that there might be small vegetation as well as bigger stones that may fall within a 5 cm, 10 cm or 15 cm distance or might be cut out by the same. No matter which distance is chosen, needed stones will fall out or redundant vegetation will fall out in the calculation of the DTM. Obviously greater search radii or distances lead to higher volume balances.

Multitemporal Analysis

Erosion stakes

In November 2011 three Erosion stakes were put into the ground at three different positions of the gravel bar. The exact positions are illustrated in Fig 5.4. In 2011 the top of the stakes were at the same level than the ground. The visibility of the stakes would now indicate erosion and the disappearance of the stakes into the ground would indicate accumulation. The erosion stakes give a first impression of processes at the gravel bar, which can then be analysed in detail by the calculation of difference models and volume balances.

Digital terrain model 2011 including the positions of erosion stakes north, center and south. (Source: own image)

The erosion stakes were sought using the total station with following results:

Picture of the erosion stakes north (A) and south (B). (Photographer: Eva Zimmermann, 07.07.2013)

Multitemporal Analysis 2012-2013

Analyses of changes on the gravel bar are carried out by calculating a difference model and the volume balance of the data acquired in the two years.

Calculation of DTM

The DTM calculations are processed.by the DTM Group by means of a Python Script, where the point cloud is passed through filters and interpolations in order to extract vegetation or to refill areas of missing data (for more information see chapter 3).

Calculation of difference model

The calculation of a difference model is performed via the OPALS module opalsDiff. The difference model describes the extent of erosion and accumulation within one year (between June 2012 and July 2013). For this calculation interpolated DTMs were used in order to achieve a maximum completeness of the difference model with less “nodata pixels”. OpalsDiff offers different interpolation methods, which adapt the cell size of the input file to the cell size of the difference model. In this case the interpolation “bilinear” has been used, as in the data processing of the 2012 survey. The bicubic method would amplify the difference in height whereas bilinear interpolation maintains them (Backendorf et al. 2012). The command line of the calculation of the difference model via opalsDiff can be found in 5.7 Appendix 3. Coloring of the difference model with opalsZColor

For the coloring of the difference models the colour palette differencePal.xml of the OPALS module zColor was used. This palette offers five colors for negative and positive differences and two colors for values out of the range covered by the scale. After testing several scale options “Scale”=0.06 was chosen for the coloring. The command line of opalsZColor module can be found in 5.7 Appendix 4.

Calculation of the volume balance

The calculations need to be carried out via Excel and ArcGIS (Fig. 5.1)

The volume balance is calculated on basis of the formula:

“Number of pixels * Pixel size * Mean of height difference model”

Number of pixels: 80449
Pixel size: 0.0625 m2
Mean of height difference model: -0.38 m
Volume balance: -1697.96 m3
Size of gravel bar: 5044.4375 m2
Mean increase: 299.51 m3
Mean decrease: -1961.89 m3

Between 01.06.2012 and 07.07.2013 the average erosion was about 34 cm.

Analysis of changes 2011/12 – 2012/13

In order to maintain the comparability of the changes between 2011/12 and 2012/13 all DTMs and difference models were processed by the same script.

The difference model of 2012/2013 illustrates the strong erosion that occurred on most parts of the gravel bank. In 2011/2012 the volume balance was about 652 m3, which is 1000 m3 less than in the latest time period. There are two main reasons for this huge difference between the erosion rates. First, the time periods between the campaigns varied from six months (18.11.2011 to 1.6.2012) to 13 months (campaign on 7.7.2013). Another explanation for the difference in volume balance is the strong flood in May and June 2013. This assumption can be approved by the bent erosion stake south, probably caused by heavy material movements on the gravel bar. The erosion stake north stood out 56 cm at the time of the campaign in 2013, the calculations at the same point result in 32 cm. Even though both measurements indicate strong erosion there is a great difference in the degree of erosion. This could be an indication for an underestimation due to interpolations in the DTM and difference model.

In addition to the mentioned reasons, the results of volume balance depend on several parameters that can only be estimated for the calculations and therefore lead to a high probability of errors. In 2012 there was a pond on the gravel bar, which did not exist any longer in 2013and the erosion stake centre could not be found any more in 2013, so in conclusion there must have occurred accumulation on this part of the study area. In the difference model the area of the former pond could not be included in volume balance calculations because there was no data of 2012.

Accumulations in the difference model 2012-2013 of more than 30 cm in the southeast of the gravel bar can be caused by the flooding event in May and June 2013 as well. During the decrease of the water level the vegetation in the southeast of the gravel bar might have detained material on the edge.

When interpreting the results it is important to consider the influence of different input parameters and parameter choices throughout the progress. As chapter 3 reveals, the choice of the DTM cell size has a strong impact on volume balances: the bigger the cell size the higher the volume balance. Also interpolation methods in the OPALS module opalsDiff affect the difference model and thus the volume balances.

Visualisation in the form of a colored difference model is helpful to locate erosion and accumulation processes on the gravel bar. However the main conclusions given by the visulisation highly depend on the subjective choice of the color scale. Both difference models are illustrated with the same scale (0.06). Depending on the color scale different aspects of processes on the gravel bar are emphasised.

Difference model 2011-2012, used DTMs generated in 2013 with Python script of the DTM group. (Source: own calculation)
Difference model 2012-2013. (Source: own calculation)

In addition it has to be considered that the results of the registration of the data as well as the fine georeferencing vary according to the adjustment decisions of the respective person in charge.


After all the multitemporal analyses allow drawing several certain conclusions. On the gravel bar both erosion and accumulation can be observed, with higher values during the larger time period between 2012 and 2013 than in the shorter period between 2011 and 2012. The values of volume balance depend on quality of registration, georeferencing and also on choices of parameters in the processing via OPALS and python. Therefore the results such as volume balance and difference models lack of significance. However terrestrial laser scanning and its processing methods offer considerably great datasets for calculations in a wide area. It can be assumed, that the gravel bar is under strong influence of the water and will constantly change in its extent and form.


1) Fine Georeferencing: 3D Correction
Coordinates of the point pairs:
Processing of scans into odm file:
-inFile ilvesheim_2013-07-07.RiSCAN\Export\new\Scan_1.txt
ilvesheim_2013-07-07.RiSCAN\Export\ new\Scan_5.txt
-outFile ilvesheim_2013-07-07.RiSCAN\Export\new\2013.odm
-iformat ilvesheim_2013-07-07.RiSCAN\FormatDef_2.xml
Conversion of odm file into xyz file:
-infile ilvesheim_2013-07-07.RiSCAN\Export\new\2013.odm
-outfile ilvesheim_2013-07-07.RiSCAN\Export\new\2013.xyz
-oFormat ilvesheim_2013-07-07.RiSCAN\FormatDefexp.xml
2) Impact of different cell sizes on the DTM and the calculated volume balance
Creation of new column (NormalizedZ) and derivation of relative height of all data points with respect to the terrain model:
-inFile 2012_Daten\Basis\RiScan\2012\2013\2011.odm
-gridFile 2012_Daten\Analyse\2012\MASK\
-attribute NormalizedZ=z-r[0] -nbThreads 1
Filtering of all points within the particular distances (5, 10, 15 cm):
-infile 2012_Daten\Basis\RiScan\2012\2013\2012.odm
-outfile 2012_Daten\Basis\Riscan\2012\2013\Punktabstand_15cm\
-filter "Generic [NormalizedZ <=0.15 and NormalizedZ >= -0.15]"
Calculation of DTM:
-inf 2012_Daten\Basis\RiScan\2012\2013\Punktabstand_15cm\
-outf 2012_Daten\Basis\RiScan\2012\2013\Punktabstand_15cm\
-feature mean -feature max -feature min -feature median -cellSize 0.50 -nbThreads 1
3) Calculation of difference model
--inf DTMs\2012_sandbank.tif DTMs\2013_water_removed.tif
--resampling bilinear
--outf Differenzmodelle\diff_bilinear_2012_2013.tif
4) Coloring of the difference model with opalsZColor
--inf Differenzmodelle\diff_bilinear_2012_2013.tif
--outf Differenzmodelle\Zcolor006_diff_bilinear_2012_2013.tif
--scale=0.06 --pal differencePal.xml