THE FEATURES OF SPORTS COMPLEX 'SUNKAR' MONITORING BY TERRESTRIAL LASER SCANNING

: The paper is aimed at the study of monitoring workflow in adverse observation conditions for sports structures. The application of terrestrial laser scanning for monitoring various objects is well-studied. However, the deployment of TLS for large sports structures is challenging. Such structures have huge sizes, complex geometry and need the control of particular elements. The monitoring gets complicated in the case when the structure is placed on a very steep slope, e.g., ski jumps or bobsleigh tracks. The subject of the presented study is the complex of ski jumps 'Sunkar' emplaced in Almaty, Republic of Kazakhstan. The monitoring network for the complex was created using a robotic total station. Due to the extreme sizes of the object (height 70 m, width 8 m, length over 100 m), TLS was accomplished from all points of the network to ensure the necessary density of point clouds and reliable georeferencing. The main observation parameters were the displacements of the ski jumps regarding the longitudinal axis. To retrieve the axis coordinates, the point cloud was sliced along the ski jump surface. The geometry of the longitudinal axis was simulated using spline functions. As an output, the suggested monitoring workflow is provided that ensures the quality of the necessary results.


INTRODUCTION
Terrestrial laser scanning (TLS) plays an essential role in the tasks of geospatial monitoring of engineering structures. The main advantage of TLS is high data redundancy. Such redundancy allows the detailed modeling of different structure elements (Vezočnik et al. 2009). By comparing simulated point clouds, one may detect the deformed regions and the total displacement of the structure from its initial position (Holst et al. 2017). TLS has been successfully studied and applied for various monitoring problems in recent years. Numerous scholars have stressed the importance of TLS for monitoring bridges (Cha et al. 2017, Rashidi et al. 2021, cooling towers (Ioannidis et al. 2006), hydro-technical objects (Schäfer et al. 2004, Schneider 2006, Koska et al. 2008, tunnels (Lindenbergh et al. 2005, Nuttens et al. 2010, Xu et al. 2019, structural elements testing (Nguyen and Weinand, 2020), pipelines (Qiu andWu, 2008, Shults et al. 2022), historical buildings (Sternberg 2006, Shults et al. 2017, open pit mines (Tong et al. 2015), etc. The development of the total stations with scanning options has extended the opportunities of TLS. The surveyors have obtained an opportunity to create a monitoring network and scan the object using the same device simultaneously. On the other hand, the progress in software development has allowed the processing of vast amounts of data and output of any necessary results, e.g., models (solid, wireframes, points), meshes, drawings, crosssections, point clouds, and much more. These achievements have created premises for extended applications of TLS monitoring almost to any engineering structure. Among those structures, sports structures take an essential role. These structures have a pretty sophisticated geometry and large sizes and often are emplaced in regions with complex geological conditions. These circumstances are pertinent for such structures as ski jumps. Ski jumps always build up in mountainous regions that are characterized by very severe geological conditions and are subjected to landslide processes and adverse weather conditions. The paper presents the case study of ski jump monitoring using the scanning total station, monitoring results, and the technical workflow of the monitoring.

MONITORING OBJECT
The study object is the Sunkar International Ski Jumping Complex located in the southern part of Almaty, Republic of Kazakhstan, at a height of 900 m above sea level. The complex consists of five ski jumps. The longest one is K125, length 125 m. The neighboring ski jump is K95 and has a length of 95 m. Those two ski jumps ( Figure 1) have been selected for detailed monitoring.  The geometry of ski jumps is pretty complex (Figure 2). The curved ramp is made of steel trusses overlaid by a concrete runway and covered hill for landing. In general, the ramp presents a geometrical curve resembling a parabola. However, in reality, the geometry is more complex. One may see that the complex is located in regions prone to landslide activity. During the year, the sports complex undergoes serious loads of snowfalls and wind gusts. Another factor is the seismicity of the Almaty city. For years this region has been subjected to permanent earthquakes, some of which had disastrous effects, e.g., in 1911, 1936, 1967, and 1971. Those earthquakes had brought about many landslides. Summing up, the sports complex is affected by various environmental factors. These conditions dictate the necessity of geospatial monitoring of the sports complex.

Monitoring Workflow
The starting point of any monitoring project is the development of a monitoring workflow that is presented in the form of a flowchart. For this specific object, the monitoring flowchart has been developed ( Figure 3). The flowchart contains three main blocks presenting the design step, fieldwork and processing/modeling step, and the results delivered. The design step supposes the various accuracy calculations, establishing scanning stations, choosing the equipment, point foundation, target preparation, and emplacement. The scanner check/calibration before the fieldwork is essential to ensure that it provides the necessary accuracy, being that any monitoring task requires high measurement accuracy. The fieldworks begin with the network creation and coordinating the targets for further cloud georeferencing. The network creation follows the scanning step. This step actually takes up to 10% of the working time for the whole project. Thanks to the total station with the scanning function, network creation, and scanning are possible with the same equipment. The data after point cloud acquisition go to processing. The processing is the most timeconsuming stage. The separate point clouds are referenced during the processing into one point cloud model defined in the common coordinate system. Upon the finishing of georeferencing, the point cloud is triangulated, meshed, and, if necessary, modeled. The comparison step finishes the processing stage. The comparison is possible in different ways. The simplest case is the comparison of reference target coordinates. However, in this case, the approach has no difference from ordinary total station surveying. Much more informative is the comparison between TIN models from different observation epochs. This comparison provides a whole picture of the structure deformation. Except for comparing different TIN models, for the relatively new structures, it is possible to equalize the scanning results with the existing design model. The final step is the results reporting. The comparison output may present spatial displacements of specific points (targets), the structure displacement in total, the changes of the structure deformation (geometry changes), possible crack detection, and prediction models for displacement evolution. The forecasting models are built up based on several observation epochs and may include different observation parameters, such as volume of precipitation, temperature variation, groundwater level, etc., for the case of the given object, of interest are vertical displacements along the longitude axis of the ramp. Thus, the appropriate processing method has been suggested and considered in Section 4.

Monitoring Network
The observation network has been created using scanning total station Leica Nova MS60. The geometry of the sports complex dictates the network geometry. The network is stretched along the ski jumps ( Figure 4)   Figure 4. Scheme of the monitoring network.
In Figure 4, apart from the network scheme, the error ellipses are given. The sizes of the error ellipses are presented in Table 1. The network was adjusted using the standard least squares procedure. The lowest accuracy has been obtained for the points located down the ski jump hill. As far as the monitoring of the landing hill needs lower accuracy, the obtained results are considered acceptable. Meanwhile, the points around the ski jump ramps were determined with accuracy that does not affect monitoring results. In what follows, these points were used for scanning stations. It is worth mentioning that for every observation epoch, the network is remeasured, and the appropriate corrections to the point coordinates are applied.

Scanning Results
The scanning has been accomplished by Leica Nova MS60. The point cloud is not highly-dense and contains 40K -50K points ( Figure 5).

Figure 5. Georeferenced point cloud for two ski jumps.
Two observation epochs were conducted. Since the main interest has the displacements in a vertical direction for the ski jump ramp, it was decided to make cross-sections along the longitude axis. An example of such a cross-section is portrayed in Figure  6. The cross-section clearly demonstrates the two parts of the ski jump, the upper part corresponds to the ramp, and the bottom part corresponds to the landing hill. As far as the primary goal of the monitoring is the study of the ramp displacements, further analysis has been carried out for the upper part of the ski jump.

Spline Functions
The final step of monitoring is displacement modeling. However, for the TLS data, this task becomes a bit tricky. Insofar as the point clouds for the different observation epochs have different distributions and it is impossible to measure the coordinates of the same points, the solution is to apply the interpolation technique. The strict correspondence between the points in different epochs has to be established for the coordinate comparison. Otherwise, the discrepancies will distort the actual values of the determined displacements. In Figure 7, the case of correspondence search between the design model and TLS point cloud is presented. Figure 7. The correspondence between the design model and point cloud.
Therefore, it is necessary to have a definite correspondence between the points in different observation epochs. The straightforward approach is to compare the preliminary marked points, but we only have recognizable points on the ski jump for georeferencing targets. Thus, it has been suggested to apply interpolation techniques to achieve the required results. For the interpolation models, the cubic spline functions were chosen. The splines have become very popular recently thanks to their robustness. There are many applications that one may find, e.g., in (Kermarrec et al. 2017, Kermarrec et al. 2021). Yet no one has pointed out the application of one spline type better than another (Shults et al. 2021). Typical spline expressions for the spatial curve will be ( ) = 0 + 1 + 2 2 + 3 3 ; ( ) = 0 + 1 + 2 2 + 3 3 ; ( ) = 0 + 1 + 2 2 + 3 3 , where 0 , 1 , 2 , … , 0 , 1 , 2 , 3 c = unknown spline coefficients = spline parameter (spatial distance between the points or increments along the coordinate axis) In our case, these functions are especially useful since they provide a pretty smooth interpolation for the cross-sections without outbreaks (Figure 8). The different spline functions ensure different interpolation depending on the end conditions, e.g., "Natural" cubic splines, Not-a-knot cubic splines, Hermite splines, modified Hermite splines, B-splines, t-splines, etc. The given study uses the three types of relatively simple cubic splines with cubic, parabolic, and linear end conditions. Visually the difference between these splines is hardly discernible. The modeling has shown that all of the considered spline functions provide similar outputs. The results of interpolation using cubic spline with parabolic end conditions are presented in what follows.
During the monitoring, we obtained the design model of the ski jump complex. It allowed us to generate the ideal cross-section through the center of the ski jump ramp (Figure 9) and further used it for comparison. Figure 9. Design cross-section.
The cross-section was generated with 1-meter step. Consequently, the monitoring results were interpolated to get the values for the same points of the design cross-section. Spline interpolation of the TLS data is presented in Figure 10. One may see that the points are located unevenly; in turn, we need to interpolate them to find the corresponding points for the design values. Moreover, comparing cross-sections for different observation epochs has the identical drawback. So, the points for the different epochs have to be interpolated too. Having the interpolated values, we may determine the displacements.

Displacements Modeling
Firstly, the interpolated coordinates of the first epoch were compared with the design values. The differences in coordinates indicate the possible deviation of the ski jump surface as it was built ( Figure 11).  These figures suggest that the displacements do not have a place regardless of some values that reach 10 mm. These abnormal values must be treated as blunders or some artifacts on the surface. Contrary to our expectations, the structure is relatively stable and has not undergone any deformations during the observation time.

CONCLUSIONS
The presented paper outlines the results of geospatial monitoring of the sports complex 'Sunkar'. The flowchart of the monitoring workflow was developed to carry out the monitoring. The suggested workflow supposes displacement determination differently, i.e., using interpolated point clouds and a design model. As the primary interpolation method, cubic splines were chosen and studied. The study proved the high efficiency of the TLS data for monitoring tasks. However, our findings are not generalizable beyond the scope of this study. Future research will have to confirm whether the suggested workflow is applicable to other similar structures. Our dataset was limited to two observation epochs. Obviously, it is necessary to have at least ten observation epochs to develop prediction models. Therefore, further studies will be focused on the prediction models. Apart from that, in what follows, we are intended to study the surface interpolation models using different spline models.