Terrestrial Laser Scanner ( TLS ) Measurement in A Volcanic Area : Detection of Error Source and Scanned Object Intensity

Abstract The purpose of the study of Terrestrial Laser Scanner (TLS) application in an active volcanic crater is to detect the influence of gas emission and local atmospheric change into the accuracy of measurement.The measurement was conducted at Papandayan Volcano crater, along with the local temperature, humidity, and air pressure (thp). The measured target located near a gas emission hole gives a significant fluctuative range due to possible false return by gas particles. The refraction index was calculated using thp data. Two correction models were applied: velocitygeometry and atmospheric delay correction. The atmospheric delay correction gives a more reliable result, however their refraction index calculation does not accomodate specific volcanic gas yet. An intensity map obtained from TLS can also be used for rock segmentation. An experiment from some types of volcanic rocks shows that the intensity value is influenced by a weathering degree of rock outcrops. Rock segmentation using TLS intensity data for fresh outcrops is relatively easier, while for weathered rocks it still needs a field check for validation. The temperature of volcanic rocks also contributes to the intensity value. It is found that the intensity increases along with the temperature of rock.


Introduction
Terrestrial laser scanning is a method of sur face sam p ling using laser technology. Scanning with laser has the objective to collect visual data of an object that includes the shape and colour. Data collected can then be used for 3D reconstruction with the help of various types of software. Scanning with laser has advantages compared with conventional methods (e.g. tacheometry, EDM, photogrametry), in terms of higher accuracy and less time consume during a field survey (Pesci et al., 2008). In addition, laser scanning results can record a huge number of points named point clouds that contain data of x, y, z (RGB, I), with xyz is the position of an object, I is the intensity and RGB is the colour of an object. TLS can reconstruct digital model of an object precisely with the accuracy up to submillimeter (Colombo and Marana, 2010). TLS method is able to describe the condition of outcrop/surface closer to the actual condition, both in terms of geometry and spatial with excellent accuracy. Therefore, it can be applied for deformation monitoring (Waggot et al., 2005 andTsakiri et al., 2006).
In addition to these advantages, the TLS has error sources that can affect the measurement results and can reduce the accuracy. These errors are related to four sources, namely: equipment, methodology of data collection, environmental, and scanned objects (Quintero et al., 2008;Reshetyuk, 2009).
A research to determine the error source from instrument has been done by Litchi et al. (2006) and Reshetyuk (2009). The two researchers have conducted modeling and calibration using Faro 880, Callidus CP 3200, Leica HDS 3000, and the Leica Scan Station. In addition, to examine the error originating from the instrument, Reshetyuk (2009) also examined the error source derived from the methodology, particularly in geo-referencing process. Hunter et al. (2003), Jones (2007), Pesci et al. (2008), Bonforte et al. (2011), Vidyan et al. (2013), and Jones et al. (2015) had apllied TLS measurement in a volcanic area, but only limited in deformation and slope stability monitoring, and 3D topography modeling. The researchers did not discuss the performance, nor the accuracy of measurement data in detail. They also did not discuss the error sources of volcanic environments, especially the ones associated with the activity of volcanic gases and local atmospheric conditions. Rueger (1990) had written speed and geometry correction formula for measurement by Electrooptic Distance Meter (EDM) which used visible waves (optical). Meanwhile, Wijaya (2010) had formulated a unified zenith delay correction models for satellite-based distance measurement using micro (GNSS/VLBI) and optical (SLR) waves. Therefore, a correction model of these two researchers was tried to be applied for TLS measurement results, in other words to use or test the old concept with new technology.
A research on the response of scanned object, in this case expressed in terms of intensity, had been conducted by Litchi and Harvey (2002), Pfeifer et al. (2007), and Voegtle et al. (2008).The three researchers scanned artificial objects, metal, and some rocks which were not of volcanic eruption product. Their study emphasizes on intensity values and their relation to the distance and angle of the scan. In this research, the focus is on a volcanic rock, especially the relation between volcanic rock surface properties (including colour, surface texture, crystal content, pores, temperature, and impurities) with intensity values. This study of rock intensity is a preliminary quantification. Implications of the intensity value to distance accuracy are not described in detail, since they require another parameter that should be tested.

Error Source of TLS in Volcano Environment
Error sources within volcanic environment is very complex due to several factors that occur simultanously and they can affect the measurement results, as illustrated in Figures 1 and 2. Sources of error associated with wave dispersion are due to separation of spectrum frequency of laser beam on inhomogeneous medium/atmosphere, in this case the medium containing particles of volcanic gases.
Local weather in an active volcanic crater tends to be fluctuative quickly, and will contribute in reducing the accuracy of measurement. Hunter et al. (2002) has applied TLS in Mount Etna and found that TLS has limitation to penetrate volcanic fog. Radiometric contribution has to be considered. Ingensand (2006) offended correlation between scanned object factors and measurement results. He states that in theory, the error of the coordinate system (S xyz ) from a point cloud is a function of: • distance (D); which correlates with atmospheric parameters • reflectivity of the object; affects the signal to noise ratio (SNR) • ambient light; affect SNR • angle of incidence ray (θ) • wavelength; correlated with surface roughness of rock and produce spots/speckles • (white noise) • colour of the object surface • object surface roughness • geometry of the object; could lead to multipath effects  Bonforte et al. (2011) stated that volcanic rocks of black porous lava had poor reflectivity characteristics. It means that the nature and types of objects have different responses to TLS. Boehler and Marbs (2002) said that TLS measurements in areas of dust or water vapour/ gas could cause an error resembles the edge effect. This is because laser propagation velocity changes due to variations in temperature and pressure. Edge effects occur if part of the object edge is reflected, while the other part was behind or around the scanned object. Even if there are no other objects in the vicinity, the laser will not be reflected at all. Grantham et al. (1997) who conducted a study on the atmospheric effects into TLS measurement results discovered a phenomenon of dropout pixels, a condition in which the laser pulse is too weak to trigger a receiver, so the record is zero or maximum value of the pixel. This phenomenon is associated with laser attenuation. Grantham et al. (1997) also concluded the presence of false return phenomenon, which were laser dispersed by rain water droplets, aerosols, water vapour, or dust particles suspended in the air, that was strong enough to trigger the receiver. Most of transmitted laser energy is reflected by the particles and partly by the object surface. These conditions build up laser pulses (multiple return pulses) and contaminate generated data, thus complicate object scanned identification process.

TLS Acquisition in A Volcanic Crater Area
TLS range measurement was conducted at Papandayan Volcano, West Java, using Leica Scan Station C10. Three planar targets were placed near a gas emission hole at some distances: ± 50 m (target A), ±100 m (target B), and ±150 m (target C). During the measurement, volcanic gas  (Haerani, 2016  emission often crossed in front or near the target, mostly to target A. Each target was measured fifty times, but on target C~150 m the measurements can only be done twenty-one times due to the instrument problem. The specifications of the instrument could actually reach optimal distances up to 300 m, but the volcanic crater conditions with a lot of gas emission obstructed the view between TLS and the target. During the distance measurement, weather data recording thp (air temperature, relative humidity, and air pressure) also performed automaticsensor (TNDD TR-73U) every 5 seconds. Two thp sensors were placed on TLS (base) and in the target (rover). At measurement of target B~100 m, a gap occured ± 60 minutes at rover sensor due to data downloading process.

Intensity Value Experiment
A simple experiment to determine the intensity of a volcanic rock was performed by scanning some examples of volcanics (volcanic bombs, lava, and pumice) with various surface properties. Volcanic rocks selected in this experiment are a common type of volcanic rock in Indonesia (andesitic-basaltic type).
The intensity measurements were carried out inside a room with a distance of 3.6 m, with different lighting (dark and bright light conditions). Reshetyuk (2006) concluded that for TLS type Leica HDS 3000, there was no systematic differences in ambient lighting. This experiments also did not perform a scan from several different distances, because the addition of the distance would reduce the intensity values (Pfeiffer et al., 2008). Volcanic rocks scanned are in dry conditions to avoid the water particles that can serve as a dominant reflector (water film). In this experiment, pure intensity values derived from the rock surface were expected to be obtained. Point cloud used Cyclone v.7 for object filtering process, converted to ASCII format. Plotting and normalization of intensity values were done by using Matlab software. Scanning process used a high resolution setting with air temperature at the time of scanning was 25.9 -26°C; 59 -61% of relative humidity, and 933.3 m bar of air pressure.
In order to determine the relationship between an intensity value and object temperature, a scanning process was performed at some areas of Papandayan volcanic crater, to obtain an intensity map. Some parts of point clouds on the intensitymap was sampled and converted to ptx format for viewing the intensity value, then correlated with the estimated temperature value obtained with a thermal camera. The type of thermal camera used is FLIR T-440.
Calculation from Ciddor (1996) produces a refractive index that has encompassed all known factors (except oil contamination of the atmosphere as vapour/liquid water and the effects of absorption). In addition, these calculations have included all of physical parameters and units, as well as valid for all atmospheric conditions and a wave length between 350 to 1,300 nm. (Rueger, 1990) As mentioned before, in this study, a TLS Leica ScanStation C10 was used, which had Time of Flight (TOF) measurement principle. The distance measurement can be formulated as follow: Where: (dn/dh) is vertical gradient of refractive index of air, and z is the angle between the direction of gradient of refractive index and the tangent to wave path.

Velocity and Geometry Correction
Geometric correction (K3) obtained used equation as follow:

Unified Zenith Delay Model (Wijaya, 2010)
The phenomenon of atmospheric delay has been modeled by Marini and Murray (1973) and Saastamoinen (1973). In 2004, Mendes and Pavlis created a new zenith delay model that was more accurate and could be applied to SLR wavelength (Satellite Laser Ranging). The formulation of Mendes and Pavlis (M-P) model uses the equation refractive index and density of water vapour from Ciddor (1996) as well as Ciddor and Hill (1999). This equation is commonly used in SLR with a wavelength of 0532 µm. The final formulation of hydrostatic zenith delay model (M-P) is: Wijaya (2010) created a unified zenith hydrostatic delay model (ZHD) from (M-P) models and dispersion refractivity group equation of Ciddor (1996) and resulted in the formulation: ZHD formulation of equation (14) and (15) is slightly different. Mendes and Pavlis (2004) created ZHD models by first derived density of dry atmosphere ρd as a function of pressure (P), temperature (T), and the vapour pressure (e). Then based on the ideal gas equation, they derived the total atmospheric density ρt and group refractivity based on modified dispersion I J O G Terrestrial Laser Scanner (TLS) Measurement in A Volcanic Area: Detection of Error Source and Scanned Object Intensity (N. Haerani et al.) formulation for wavelength 0532 µm. ZHD formulation by Wijaya (2010) indicates that equation by Mendes and Pavlis (2004) can be calculated in a more simple way. Besides, his formulation was prepared without simplifications and approximations. The values of dry air are not defined in the formula, so any dispersion formula can be used to determine ZHD, but this case is recommended to use the dispersion formulation of Ciddor (1996). Zenith wet delay was integrated from Saastamoinen model (1973) and the definition of slant wet delay by Davis et al. (1985). The final formulation is obtained as follows: To estimate atmospheric delay at any elevation angle, mapping function (MF) is applied to map relationship between elevation and ray propagation. Marini (1972)  with ∈ is the elevation angle of the rays, and the coefficients a i (i= 1, 2, 3) were determined by least square fitting on beam curve depending on variation of meteorological parameters at the measurement location. MF is based on the assumption of hydrostatic equilibrium and symmetry azimuth and stating ratio of atmospheric delay at multiple elevation geometry, for example, delay to zenith direction.

TLS Range Measurement
Distance measurement results (initial data) are shown in Figures 3 to 5.These figures noted some data fluctuation. The measurement of Target A~50 m fluctuations occur at the beginning to 12 th measurement, after that the measurement results are relatively stable. The results of measurements at distances of B~100 m and C~150 m fluctuations occur during the measurement time. Fluctuations of Target A~50 has the range from 1 mm (smallest) to 5 mm (largest), at B~100 m and C~150 m, the smallest is 1 mm and the largest is 2 mm. Fluctuations in A~50 m is relatively larger due to target position that is closer to the gas emission holes, while the positions of target B~100 m and C~150 m are further from the gas emission holes. The volcanic gas emission activity was continuing along the measurement process, some of gas particles passed in front of the target while laser beam hit the target. The average of measurement data can be seen in Figures 3 to 5 and Table 1.
The record of temperature, relative humidity, and air pressure (thp) data at base (TLS) and target is shown in Table 2. The temperature range at TLS base station is 15.3 -23.6 °C, relative humidity 83 -53%, and pressure of 785.1 -783.2 hPa. The temperature range at target is 14.7 -24.2 °C, relative humidity 87 -43%, and pressure of 783.1 -782.1 hPa. The range thp between TLS base station and target did not show any significant difference although thp at both places experienced fluctuation, but they give the same pattern.

Corrected Distance and Comparison of Two Correction Models
The corrected distance using velocity and geometry correction model are calculated using Equation (5) until (13). The correction result is summarized in Tables 2 and 3 and Figures 6 and 7.
The comparison of initial measurement with velocity and geometry correction results ( Figures  6 and 7) shows that for a distance of ~50 m the corrected distance is shorter.While for B ~100 m and C ~150 m, it shows longer distances. K1shows a significant value in each range, even reaching a fraction of cm at A~50m. However, K2 and K3 show a very small value, as summarized in Tables 3 and 4. These small values of K2 and K3 are interpreted due to the involvement of spheroid curvature variables (R, radius of the earth ), while the distance measured in this study I J O G Terrestrial Laser Scanner (TLS) Measurement in A Volcanic Area: Detection of Error Source and Scanned Object Intensity (N. Haerani et al.) 187 is relatively short. These small values cause K2 and K3 overly into K1 (ignored). In the further discussion, only K1 with ZHD model results which will be discussed.
ZHD correction model for TLS measurement distance applied uses Equation (15) to (19). The result of ZHD correction model can be seen in Table 5. Corrected distance for Target A shows a shorter distance for both correction model (K1 and ZHD), whileTargets B and C indicate the opposite pattern. In this case, ZHD model is considered to represent more actual size. Based on the difference value obtained from both correction model (Table 5), ZHD models show a consistent value than K1. In addition, the consistency of the difference is related to the distance between targets which are relatively the same, about 50 m.
Distribution of observation data is expressed by box plots on the right column of Figures 6  b, d, f, and 7 b, d, f. Targets A, B, and C show asymmetric distribution patterns, marked by elongated whisker on one side. Targets A and B show a distribution tendency of negative skewness, while target C shows a positive skewness. All distribution pattern of d, K1, and k shows the same pattern. This asymmetry indicates that there are outliers on data out of central tendency.  on local coordinates is presented in Figure 8, while the 3D position in Figure 9. Both figures show that ZHD correction gives a smaller shift position than K1. For target A, the data tend to be more fluctuative at the beginning of measurement, possibly related to the phenomenon of false return (laser beam reflected by volcanic gas particles).

Comparison of 2D position for each targets based
The difference between initial data (d) and corrected data (K1 and k) is presented in Table 6. The table shows that ZHD correction gives values close to the goal accuracy (based on Leica Scan Station C10 specification), which is 6 mm for position and 4 mm for distance. While K1 gives a higher value as well as positive and negative values. Thus, from an experiment using velocity and ZHD model, correction method using ZHD model is better than the fisrt one. Figures 8 and 9 also note that ZHD correction (k) gives a shorter corrected distance which meets Fermat principle stating that the actual distances between two points measured by electromagnetic waves are those that have shorter travel times. In  Figure 7. Comparison between initial slope distance (d), corrected distance using velocity correction model (K1), and ZHD model (k).   addition, in terms of volcanic environment, the shorter corrected distance is in accordance with the phenomenon in Figure 2, where the distance measurement is influenced by the parameters of surface volcanic activity and the local atmospheric conditions. RMS error of atmospheric delay correction calculation method can be estimated with pressure parameters (P), temperature (T), and the water vapour pressure (e o ) which can be formulated as follows (Abshire and Gardner, 1985):

I J O G
.... (20) with σP, σT, and σe o declare RMS on P, T, and e o . Figure 10 shows the simulation results of a decrease in P, T, and e o at elevation angle 5° up to 90°. From the figure, it is seen that P gives the greatest RMS. Simulation value of P, T, and e o at 10° elevation and equation (9) gives the RMS value of 4.3 mm (Hulley, 2007).  Haerani et al.) In this study, the elevation between TLS stands and Targets A, B, and C is very small (about 0.54°). If this value is simulated into Figure 10, the RMS value is greater than 4.3 mm. Table 7 shows that RMS value for all three targets on parameters of temperature, pressure, and water vapour pressure is in the range of 8.89 to 10.69 mm. Moreover, because of this small elevation value, it is assumed that there is no horizontal gradient effect. The differences in topographic height between TLS stand with targets A, B, and C are 0.4, 5.7, and 20.5 m, respectively.
Atmospheric delay model assumes that atmospheric conditions is in hydrostatic equilibrium. Irregularities related to turbulence, temperature changes, and vertical acceleration generated an error against the meteorological measurements (thp) itself. Saastamoinen (1973) estimated the maximum error of 1.5 cm at the elevation of 10°, while Hauser (1991) stated that for mountainous environment the error was smaller than 1 cm for elevation of up to 20°. In this study, the meteorological parameters are limited to thp variations in an active crater environment, whereas the effects of turbulence and atmosphere stratification layers of active crater surface is not discussed in detail.     Tables 8 and 9 show the results of scanned volcanic rocks. For PBCB, there is a separation for dark (black-gray) and light (red-brown) parts to detect any differences on the intensity values.

Intensity of Volcanic Rocks
The histograms in Figure 11 show that a darker colour has narrower ribbons of intensity values with peak values in the range of 0.15. While lighter colour has a wider ribbon and gives two patterns, one is in the range of <0.15 and the other at 0.15 to 0.2. Observing this histograms, it can be concluded that the darker colour has a more focused intensity value than the lighter one.
The intensity value histogram for the other rock samples can be seen in Figure 12. Figures  12 a, b, c, and d represent samples of fresh rocks, while images on the right columns (Figures 12 e, f, g, and h) indicate a group of weathered/altered rock samples (PCDB-1, GLBA-2, TPF, PCDB-2). From both groups of rock samples, the fresh rock provides intensity value histogram with clear dominant values and narrow band, while the weathered/altered rock has a wider intensity value range with several peak values.
To detect the correlation of mean intensity to scanned object properties, some graphs ( Figures  13 to 15) are created based on megascopic ob- . Atmospheric delay variation related to pressure, temperature, and water vapour pressure for elevation 5° to 90° (Hulley, 2007).   servations of these rocks including colour (dark to light), surface texture (coarse to fine), and mineral/crystal content (abundant to few). All plots at Figures 13 to 15 show nonlinear patterns, but these graphs indicate that rocks with a lighter colour, smoother surface, and less crystal content, will give a higher intensity value. The range of mean intensity value for volcanic rocks in this study is from 0.16 to 0.24. This value range is lower than the other scanned objects measured from previous authors (Table 10).
Volcanic rocks in Indonesia generally consist of andesitic-basaltic type. Thus, it is assumed that the value of its intensity is in the range of 0.16 to 0.19 (represented by rock sample codes from PBCB up to GLBA-2). While TPF is more acidic (dacite) and very rarely produced by the recent volcanic eruption associated with caldera formation.
The preliminary result of volcanic rock intensity shows that the value is correlated with properties of rock surface, indicated by very clear colour parameters, surface roughness, and crystal *) obtained from histogram content. This implies that volcanic rock with a lighter colour (lava dacite-andesite), smooth surface (non-vesicular), and less crystals content (nonporphyritic structure) will reflect a greater Table10. Mean Intensity Range of Volcanic Rocks Compared to Some Types of Scanned Objects energy (intensity value) against TLS censor.The weathering/alteration degree of volcanic rocks affects the intensity value. Volcanic rocks that have been altered contain more minerals on its surface that serve as impurities. These preliminary assessments of intensity value of volcanic rocks show that segmentation/extraction of fresh volcanic rock outcrop can well be conducted, but for weathered/altered rocks it will be more complex. Therefore, it requires a detailed field survey to get more description.
The following discussion is about the effect of scanned object temperature with intensity value. Point cloud sampling on the intensity map was applied to thirteen areas that are considered to represent the range of temperature values. Sampling location and the range of temperature values can be seen in Figure 16. The mean value of intensity for each area is presented in Table 11. Figure 17 Note: Scanning distance was reatively the same (~3 m); All scanned objects are on dry condition)  shows a graph of relationship between the mean intensity with the temperature of the rocks. Furthermore, Figure 16 shows that the highest temperature recorded by a thermal camera is 73.8°C and the lowest one is 20.5°C. For the calculation purpose, the limit of intensity value is taken at 70°<x<20° C. #PC indicates the number of point clouds (which contains the value of intensity) involved in the calculation. Plotting results of mean intensity value against temperature of rock show a positive linear pattern, where intensity values increase along with temperatures of rocks ( Figure 17). However, on the rocks with temperature below 20°C, the intensity values show a higher level. It can be interpreted that within the temperature of 30°C on the object, scanning will start to affect the intensity value.
The maximum value of the measured temperature is ± 70°C. In practice, the temperature of emission gas holes in the surface can reach a higher temperature (200 -600°C). Maata et al.  (1993) conducted a scan on a converter with temperatures of 1050° -1400° C. The measurement results show that the received signal is increased by 28% in line with the increase in temperature. Thus, the pattern of increase in value of intensity against object temperature is linear.

Conclusions
Based on the data obtained, it can be concluded that atmospheric condition and surface activity of an active volcano crater (emission of volcanic gases) influence significantly into TLS range measurement. The application of first velocity correction K1 does not provide optimal results due to sensitivity of model to temperature changes, while the second velocity correction (K2) and geometry correction (K3) are not significant for a short distance. Correction using atmospheric delay model gives acceptable results, both in terms of correction magnitude value (consistent at 3.9 mm) or by dispersion phenomena. An atmospheric delay correction model gives a better accuracy in terms of 2D and 3D position.
Index refraction calculation using Ciddor formulation has already accommodated all parameters, but for applications in volcanic environments, it is necessary to find other alternative models or a new compiled model that involves dominant volcanic gas contents (N 2 , CO 2 , SO 2 , H 2 S, HCl, NH 3 , and H 2 O).
The brightness colour of the rock, crystal content, and roughness of rock surface affect the intensity value, in this case colour parameters are more dominant. The measurement of intensity values indicates that the volcanic rocks of andesitic-basaltic type derived from some volcanoes in Indonesia have intensity values ranging from 0.16 to 0.19.The temperature of the scanned object affects the mean value of the intensity recorded by TLS, in this case the mean value of the intensity starts to show the pattern of increase in temperature of 30° C. An increase in the value of intensity then has a linear pattern along with the object temperature.

I J O G
Terrestrial Laser Scanner (TLS) Measurement in A Volcanic Area: Detection of Error Source and Scanned Object Intensity (N. Haerani et al.)