Two campaigns to compare three turbulence profilingtechniques at Las Campanas Observatory

Joanna Thomas-Osipa, Edison Bustosb, Michael Goodwinc, Charles Jenkinsc, AndrewLambertd, Gabriel Prietoa, and Andrei Tokovininb

aGMT/Las Campanas Observatory, Colina El Pino, La Serena, ChilebCerro Tololo Inter-American Observatory, CTIO/AURA Inc., Casilla 603, La Serena, ChilecResearch School of Astronomy and Astrophysics, The Australian National University, Mt.

Stromlo Rd, Weston, AustraliadSchool of Information Technology & Electrical Engineering, UNSW@ADFA, Canberra,

Australia

ABSTRACT

In preparation to characterize the Giant Magellan Telescope site and guide the development of its adaptiveoptics system, two campaigns to systematically compare the turbulence profiles obtained independently withthree different instruments were conducted at Las Campanas Observatory in September, 2007 and January 2008.Slope detection and ranging (SLODAR) was used on the 2.5-m duPont telescope. SLODAR measures the C2

n

profile as a function of altitude through observations of double stars. The separation of the observed doublestar sets the maximum altitude and height resolution. Ground layer (altitudes < 1 km) and free atmosphereturbulence profiles are compared with those obtained with a lunar scintillometer (LuSci) and a multi-aperturescintillation sensor (MASS), respectively. In addition, the total atmospheric seeing was measured by bothSLODAR and a differential image motion monitor (DIMM).

Keywords: turbulence profile, seeing, scintillation, slope detection and ranging, lunar scintillation

1. INTRODUCTION

Two campaigns to systematically compare the turbulence profiles obtained independently with three differenttypes of instruments were conducted at Las Campanas Observatory (LCO) in September, 2007 and January 2008.This is part of an ongoing site testing effort to evaluate potential Giant Magellan Telescope (GMT) sites at LCO1

and motivated by the need to choose instruments with which to characterize the Giant Magellan Telescope siteand guide the development of its adaptive optics system.

A similar campaign was made previously at Cerro Tololo.2 This ESO-CTIO joint campaign compared severalturbulence profiling methods and met with some sucess. It was motiviated by a desire to compare methods ofmeasuring the ground layer thickness for GLAO feasibility studies. They found a good agreement in the integralseeing in the ground layer and free atmosphere as measured by a portable SLODAR3 and a MASS-DIMM(described further in Section 2.3).

Section 2 describes the instruments that were used in this study. A comparison of the results from all of ourinstruments in presented in Section 3 where we show general agreement of total, free atmosphere, and groundlayer integral seeing. Furthermore, qualitative agreement is shown for some turbulence profile examples in Section3.5. Finally in Section 3.6, we present statistics of the turbulence properties from all our instruments during thetwo campaigns.

Further author information: (Send correspondence to J.E.T.) E-mail: jet@lco.cl Telephone: +56-51-207-316

Figure 1. Left: SLODAR attached to the 2.5-m duPont telescope. Right: LuSci attached to the Manquis Ridge DIMMtower. The Magellan telescopes, location of the MASS-DIMM, are seen in the upper left corner.

2. INSTRUMENTATION AND OBSERVATIONS

The observations reported here come from four separate instruments: the Australian National University (ANU)24x24 SLODAR mounted on the duPont telescope located at Manquis Ridge, a DIMM and LuSci mounted ina tower nearby (100 m distant) and a MASS-DIMM located next to the Magellan telescopes at the Manqui site(cf. Figure 1 in Ref 1). Figure 1 shows the SLODAR and LuSci instruments as well as the distance betweenthe duPont and Magellan sites. In two week-long campaigns, in September 2007 and January 2008, data werecollected from all four instruments on a total of 12 nights.

2.1 SLODAR

The ANU 24x24 SLODAR (Slope Detection and Ranging) instrument4 is comparable to a Shack HartmannWavefront Sensor (SHWFS) as used in astronomical adaptive optics. The ANU 24x24 SLODAR instrument isdesigned for the LCO 2.5-m duPont telescope to measure the ground-layer turbulence structure with resolutionsof a few tens of meters.

The SLODAR5,6 method retrieves the turbulence layer strength and height from the spatial covariance dataof wavefront gradients (spot motions) taken from observations of a bright double star. The height resolutionis given by h = w/θ, where w is the sub-aperture width and θ is the double star separation. The SLODARmethod can also retrieve information about the outer scale, L0, and power law slope index, β, of the turbulencespectrum. The layer wind speeds can be determined by analysis of the temporal-spatial covariance data.

The ANU 24x24 SLODAR instrument’s key feature is the implementation of Generalized SLODAR.7 Thisimproves the effective altitude resolution to hG = h/3 by interpolation of the turbulence height structure. Thisis achieved by optically moving the (SHWFS) analysis plane from the pupil to fractions of h and combiningsuccessive measurements to produce a higher resolution turbulence profile.

2.2 LuSci

LuSci,8 provided by CTIO, is a lunar scintillometer similar in concept to the lunar SHABAR (SHAdow BAndRanging) built by Hickson and Lanzetta.9 The spatial correlations of the scintillation signal (on order ∆I/I ∼10−4) from the Moon in four linearly mounted detectors are related via weighting functions to the turbulenceprofile in the ground layer (below 1 km). The weighting functions are dependent on the detector configuration(size, baseline separation, and orientation) as well as lunar phase and the turbulence outer scale. The instrumentalresponse to scintillation arising in specific layers can be tuned through a judicious choice of detector baselines(and thus weighting functions).

The LuSci prototype used in the campaigns discussed here had detectors spaced linearly from one end at 10,13.5, and 38 cm respectively. This resulted in the turbulence profiles determined at the ranges of 4, 16, 55, 200mwhere range is the distance from the detector (not vertical altitude). As the range grid is logrithmic so is thelayer thickness grid. In the version of the profile restoration software used for this study, no air mass correctionis performed∗.

2.3 DIMM and MASS-DIMM

Two separate differential image motion monitors (DIMM) were were used in this campaign to measure theseeing. The MASS-DIMM, described below and located at the Manqui site, and another DIMM located at theManquis Ridge site near the duPont telescope. The DIMM, first implemented in a modern fashion by Sarazinand Roddier,11 functions by relating the full-width half-maximum of a stellar image profile from a long exposurein a large telescope to variances in the difference in the motion of two images of the same star through the useof Kolmogorov turbulence theory. The Carnegie DIMM instruments†, using commercially available equipmentlike Meade telescopes and SBIG CCD cameras, are based on the CTIO RoboDIMM‡.

The MASS-DIMM,12 located at the Manqui site, is a combination multi-aperture scintillation sensor (MASS)13

and DIMM. It was fabricated and provided by CTIO§. The spatial scale of the stellar scintillation variation de-pends on the distance to the layer in which the turbulence giving rise to the wave front phase disturbance exists.Thus, the turbulence profile at a small number of discrete layers can be restored by fitting a model to the dif-ferences between the scintillation indices within four concentric apertures.14 Since the MASS senses turbulenceonly above 500 m, the profiles are assumed to be valid for all nearby sites including the duPont telescope site,Manquis Ridge. The difference between the total DIMM and the free atmosphere MASS turbulence integrals isa measure of the portion of the total seeing contributed by a ground layer (below 500 m).

3. RESULTS AND COMPARISONS

Analysis of the full data set is ongoing and what follows is a preliminary overview. A discussion of the SLODARdata reduction and results can be found in Ref. 4. No data taken in the Generalized SLODAR mode is includedin this analysis. The MASS data have been reprocessed to remove overshoots due to strong scintillation15 andfiltered for good quality data according to Ref. 12. The DIMM data have also been filtered for good opticalquality and errors due to poor tracking or clouds. The LuSci data have been filtered for data with nonphysical(negative) profiles. The altitude of the restored layer turbulence integrals is calculated from range via the airmass of the Moon at the time of observation. An air mass correction is then applied to the turbulence integrals.

We have found a qualitative agreement of the profiles produced by SLODAR, MASS, and LuSci. Examplesare shown in Figures 7 through 9. Furthermore there is a general correlation seen in the integral seeing fromthe ground layer, free atmosphere, and total atmosphere as measured by our variety of instruments. Difficultiesin comparing the results from instruments in slightly different locations and altitudes are compounded by thedifferent altitude resolutions and measurement times. In the sections that follow we discuss these issues for eachcomparison as we move from the integral quantities to the turbulence profiles as a function of altitude and time.

3.1 SLODAR and DIMM integral seeing

The seeing can be estimated from the SLODAR data in two ways. Fits of theoretical impulse functions toauto-covariance profiles for data from all the sub-apertures provides both the average power law slope, β, andtotal seeing over the altitude range sensed. Alternatively, the seeing can be calculated using the DIMM methodfor an average of approximately 20 pairs of lenslet images across the pupil. This method assumes Kolmogorovturbulence, β = 11/3. Figure 2 shows the SLODAR seeing calculated in both ways in comparison to thatmeasured by the DIMM located nearby. The second method is included explicitly for the purpose of comparing

∗An air mass correction has been implemented along with the ability to produce continuous profiles in the latest versionof the reduction software available as this article goes to press10

†http://www.ociw.edu/˜birk/CDIMM/‡http://www.ctio.noao.edu/telescopes/dimm/dimm.html§http:/www.ctio.noao.edu/˜atokovin/profiler/

Figure 2. Two methods of calculating the total seeing from the SLODAR data compared with data from the nearbyDIMM during the January 2008 campaign. The left panel shows the results from the non-Kolmogorov impulse functionfitting method. The right panel shows the results using the DIMM method.

to the DIMM instrument since Kolmogorov turbulence is implicit in its method. The first method, however, isthe preferred method to calculate seeing from the SLODAR experiment given that the power law slope of theturbulence spectrum is a free parameter. The standard formulas relating the Fried parameter, image motion, andimage quality have to be modified when the slope of the power-spectrum of the turbulence is non-Kolmogorov.There is general agreement seen between both methods and thus it appears that most of the scatter is due tothe different lines of sight utilized by each instrument as may be expected.

3.2 SLODAR and MASS free atmosphere seeing

In order to compare the seeing in the free atmosphere (defined here as above 642 m because the MASS is locatedat the Magellan telescopes which are 142m higher than the duPont telescope) as measured via SLODAR andMASS, it is necessary to convolve the SLODAR profiles with the MASS response function14 interpolated to agrid of SLODAR altitudes and then integrate over altitude. Only SLODAR profiles with maximum altitudesabove 8 km were used for this comparison. The SLODAR profiles from the widest double stars had maximumaltitudes of less than 1 km and clearly are not suited to studying the free atmosphere.

Figure 3 shows an example of the free atmosphere seeing comparison on one night. The data are binned into2 minute intervals. Given the individual time resolutions for each instrument of approximately one minute, a 2minute bin allows the most data to be matched. There is a period of almost two hours during which the MASSand SLODAR appear to agree well and another period of about one hour during which there is a discrepancy. Theperiod of agreement is typical throughout the campaigns but there are several time periods when a discrepancyof this nature (where SLODAR shows more seeing than the MASS) occurred. The total points during thesediscrepant periods is approximately 50-100 points.

This particular discrepant period appears because at this point in time SLODAR reports that more than50% of the turbulence is below 500 m suggesting that the convolution of the SLODAR profile with the MASSresponse in the lowest altitude bin can be problematic. It would appear that the convolution of the SLODARaltitude resolution and MASS response used here overestimates the contribution actually sensed in the lowestaltitude bin. Further work is needed to confirm this.

The correlation between the SLODAR and MASS free atmosphere seeing throughout both campaigns is alsoshown in Figure 3. The correlation is reasonable considering that the two instruments do not always look in thesame direction and are located some distance apart. It does appear however, that there is a small systematiceffect in the sense that in good seeing conditions either the MASS is underestimating the turbulence or the

Figure 3. Left: Example time series of MASS (black +) and SLODAR (red �) free atmosphere seeing. Right: Correlationbetween SLODAR and MASS free atmosphere seeing for both campaigns binned to 2 minutes excluding slodar profileswith maximum heights below 8 km. Discrepant periods such as that near 1 UT have not been removed. The Pearsoncorrelation coefficient is labeled PC.

SLODAR convolution is overestimating the turbulence and vice-versa in bad seeing conditions (although thereare very few points with a free atmosphere seeing larger than 1.5′′so this may be spurious). It is quite likely thatthis effect, at least in the good seeing conditions when the SLODAR convolution is overestimating the seeing, istied to the problematic MASS reponse function convolution in the lowest altitude bin.

3.3 SLODAR and LuSci ground layer seeing

Again to compare two instruments with severely different altitude resolutions, it was necessary to convolve theSLODAR profiles with the LuSci response function8 interpolated to the grid of SLODAR altitudes and integratethe ground layer as sensed by LuSci (essentially up to 1 km). Now that a method to produce a continuousprofile10 from the LuSci data exists, this will not be necessary in future analyses. For consistency, the data fromboth instruments were binned to two minute intervals. Figure 4 shows that there is good general agreement inthe temporal structure of the seeing in the ground layer. Furthermore, a reasonable correlation exists betweenthe ground layer seeing sensed by the two instruments given that while LuSci is only approximately 100 m distantfrom the duPont telescope, is almost never pointed in the same direction as the SLODAR. This and the fact ofthe high temporal variability in the turbulence activity suggests that we should not expect an exact correlation.A floor on the turbulence measured by LuSci is possible given the dearth of points below 0.2′′ but more data inreally good conditions are necessary to confirm this.

3.4 LuSci and DIMM-MASS inferred ground layer seeing

The DIMM (at Manquis Ridge near the SLODAR mounted on the duPont telescope) measurements of totalseeing can be combined with the MASS (at Manqui near the Magellan telescopes) free atmosphere seeing to inferthe ground layer seeing. In this case the ground layer has a thickness of 642 m given the 142 m difference inaltitude between Manquis Ridge and Cerro Manqui, the site of the MASS. Because the LuSci response functionsare sensitive to altitudes up to and higher than 1 km, we also calculate the seeing in what we will call theboundary layer, here defined as altitudes below 1142 m. To do this we only remove the seeing contribution fromthe top 5 MASS layers or those above 1142 m. The seeing from the MASS and DIMM are binned to one minuteintervals to retrieve a ground layer or boundary layer seeing value for each minute of data. These measurementsalong with the LuSci data are then binned to two minute intervals for consistency with the other comparisons.

Figure 4. Left: Example time series of LuSci (black +) and SLODAR (red �) ground layer seeing. Right: Correlationbetween LuSci and SLODAR ground layer seeing for both campaigns binned to 2 minutes excluding slodar profiles withless than 3 resolution elements below 1 km. The Pearson correlation coefficient is labeled PC.

Figure 5. Left: Correlation between LuSci and MASS-DIMM inferred ground layer seeing. Right: Correlation betweenLuSci and MASS-DIMM inferred boundary layer (below 1 km) seeing. Data from both campaigns binned to 2 minutes .The Pearson correlation coefficient is labeled PC.

Figure 5 shows the correlation between the LuSci ground layer seeing and both the MASS-DIMM groundlayer and boundary layer seeing. As one might expect, given the range of altitudes sensed by each combination,the MASS-DIMM boundary layer correlation appears to be better than the ground layer correlation.

Since there is also a DIMM located with the MASS instrument we can also compare the ground layer seeingat Cerro Manqui, the location of the Magellan telescopes, for these nights. An example can be found in Figure 6.The difference between the ground layer seeing at these two sites is not always this stark but it has been shownstatistically to persist most of the time.1

3.5 ProfilesSystematically examining how the profiles from two different instruments compare is significantly more difficultthan looking at the integrated quantities and this work is ongoing. The examples found in Figures 7, 8, 9, and

Figure 6. The ground layer turbulence as seen by LuSci and inferred from the MASS and DIMMs at two different locations.Manqui is the site of the Magellan telescopes, 142m higher and approximately 1 km distant from Manquis Ridge wherethe duPont telescope is located.

10 show cases of obvious turbulence structure that agree qualitatively both temporally and spatially.

Given the precision of the MASS profiles as measured by two side by side MASS instruments pointingin the same direction16 and the temporal variability of the turbulence strength, it would be foolish to expectinstruments at different locations and pointing in different directions (often opposite in the case of LuSci) to agreeany better. Scatter in the side-by-side MASS measurements were attributed to both the temporal variabilityand the miss-atribution of turbulence between layer altitudes.

3.6 Campaign overview and statistical pictures

A primary motivation for making these instrument comparisons is to understand if a combination of low reso-lution, portable instruments such as LuSci and MASS/DIMM are capable of giving a similar statistical pictureof the turbulence characteristics to those of a higher resolution instrument like SLODAR that demands a largertelescope (on the order of 1 m minimum). Table 1 shows the 10, 25, 50, 75, and 90th percentiles for the distri-butions of seeing in three altitude ranges (near the ground, free atmosphere, and total) as measured by our suiteof instruments.

Figure 7. Example MASS and SLODAR profiles as a function of time on Jan 22, 2008. Only a small portion of theSLODAR data from that night are shown. Both instruments show bursting activity between 1 and 2 km near 8 UT.

Figure 8. Concurrent MASS and SLODAR turbulence profiles. The MASS profile has been scaled to form an image usingthe scale in the bar at the right. Contours of the smoothed SLODAR profiles are displayed in units of C2

n(h) on the sametime and altitude scale as the underlying MASS image

Figure 9. Example LuSci and SLODAR profiels as a function of time on Jan 21, 2008. Only a small portion of theSLODAR data from that night are shown. Both instruments show stronger activity very near the ground and above 200m near 2 UT.

Figure 10. Concurrent LuSci and SLODAR turbulence profiles. The LuSci profile has been scaled to form an image usingthe scale in the bar at the right. Contours of the smoothed SLODAR profiles are displayed in units of C2

n(h) on the sametime and altitude scale as the underlying LuSci image

Table 1. Statistical distributions of the seeing as measured by the instruments in our campaign

Instrument (type) Number of Points 10% 25% 50% 75% 90%SLODAR non-Kolmogorov (total) 1666 0.48 0.58 0.70 0.87 1.03SLODAR DIMM (total) 1666 0.50 0.60 0.74 0.90 1.06DIMM (total) 1666 0.52 0.60 0.70 0.85 1.02SLODAR (free atmosphere) 546 0.27 0.34 0.47 0.61 0.77MASS (free atmosphere) 546 0.22 0.28 0.39 0.59 0.80SLODAR (ground layer) 405 0.32 0.40 0.49 0.64 0.76LuSci (ground layer) 405 0.27 0.32 0.42 0.56 0.68MASS-DIMM (ground layer) 788 0.27 0.40 0.51 0.67 0.86MASS-DIMM (boundary layer) 805 0.27 0.41 0.52 0.68 0.89LuSci (ground layer) 788 or 804 0.28 0.36 0.46 0.63 0.82

Each set (SLODAR/DIMM, SLODAR/MASS, SLODAR/LuSci, and LuSci/MASS-DIMM) was binned sep-arately to a 2 minute interval to maximize the number of points in each set as discussed in each of the previoussections. Thus, the distributions for each instrument comparison set (delineated with solid lines in Table 1) aremade up of a distinct number of points in time. The effect of this can be seen in the difference between the twodifferent LuSci ground layer distributions. The LuSci data that matched the SLODAR data had a slightly betteroverall seeing distribution than the LuSci data that matched the MASS-DIMM data. This is not surprising whenwe consider the fact that there are twice as many data points in the LuSci to MASS-DIMM correlation (Figure5) as in the LuSci to SLODAR correlation (Figure 4).

We can see that the distributions of free atmosphere and total seeing integrals match very closely although theydo not have exactly the same shapes. For example, the MASS senses less turbulence in the 10th percentile thanSLODAR does but the opposite is true for the 90th percentile. In the comparison of LuSci and SLODAR data,we see a small systematic shift in the distributions with LuSci sensing less turbulence overall. This systematiceffect continues in the comparison of LuSci to the ground and boundary layer inferred by the MASS-DIMM.Although the effect is quite small, it is probably real given that it shows up in two different comparisons. Furtherwork is needed to understand the cause or at least to come to a conclusion on the proper combination of theLuSci and MASS profiles.

4. CONCLUSIONS

In our preliminary analysis, we have found a qualitative agreement of the profiles produced by SLODAR, MASS,and LuSci. There is a general correlation seen in the integral seeing from the ground layer, free atmosphere, andtotal atmosphere as measured by our variety of instruments. Furthermore, we have seen that the turbulencein the vicinity of Las Campanas Observatory is highly temporally variable and site dependent. It is, therefore,recommended that GMT location be studied in depth. Finally, we find that low resolution profiles would appearsufficient to characterize the statistics of the turbulence characteristics although some extra work is needed todefine the process for this.

ACKNOWLEDGMENTS

We are grateful for discussions with M. Johns, M. Phillips, P. McCarthy, S. Shectman, D. Fabricant, M. Lloyd-Hart, J. Codona, and R. Angel. This work would not be possible without the diligent efforts of the LCO technicalstaff and site testing operator, Cesar Muena.

REFERENCES[1] Thomas-Osip, J., Prieto, G., Johns, M., and Phillips, M., “Giant Magellan Telescope site evaluation and

characterization at Las Campanas Observatory,” in [Astronominal Telescopes and Instrumentation ], Proc.SPIE submitted (2008).

[2] Sarazin, M., Butterley, T., Tokovinin, A., Travouillon, T., and Wilson, R., [The TololoSLODAR Campaign, Final Report ], ESO internal document (2005). https://www.eso.org/gen-fac/pubs/astclim/paranal/asm/slodar/The Tololo SLODAR Campaign.htm.

[3] Wilson, R., Bate, J., Guerra, J. C., Hubin, N., Sarazin, M., and Sauntera, C., “Development of a portableSLODAR turbulence profiler,” in [Astronominal Telescopes and Instrumentation ], Proc. SPIE 25490 (2004).

[4] Goodwin, M., Jenkins, C., Conroy, P., and Lambert, A., “Observations of ground-layer turbulence,” in[Astronominal Telescopes and Instrumentation ], Proc. SPIE submitted (2008).

[5] Wilson, R., “SLODAR: Measuring optical turbulence altitude with a Shack-Hartmann wavefront sensor,”Mon. Not. R. Astron. Soc. 337, 103108 (2002).

[6] Butterley, T., Wilson, R. W., and Sarazin, M., “Determination of the profile of atmospheric optical turbu-lence strength from SLODAR data,” Mon. Not. R. Astron. Soc. 369, 835845 (2006).

[7] Goodwin, M., Jenkins, C., and Lambert, A., “Improved detection of atmospheric turbulence with SLODAR,”Opt. Express 15, 14844–14860 (2007).

[8] Tokovinin, A., “Turbulence profiles from the scintillation of stars, planets, and Moon,” in [Workshop onastronomical site evaluation ], Rev. Mex. Astron. Astrof. (Conf. Ser.) 31, 61–70 (2007).

[9] Hickson, P. and Lanzetta, K., “Measuring atmospheric turbulence with a lunar scintillometer array,”PASP 116, 1143 (2004).

[10] Tokovinin, A., [Restoration of continuous turbulence profile from lunar scintillation ], CTIO internal docu-ment (2008). http://www.ctio.noao.edu/˜atokovin/profiler/restor3.pdf.

[11] Sarazin, M. and Roddier, F. A&A 297, 294 (1990).[12] Kornilov, V., Tokovinin, A., Shatsky, N., Voziakova, O., Potatin, S., and Safonov, B., “Combined MASS-

DIMM instrument for atmospheric turbulence studies,” Mon. Not. R. Astron. Soc. 383, 1268–1278 (2007).[13] Kornilov, V., Tokovinin, A., Vozyakova, O., Zaitsev, A., Shatsky, N., Potanin, S., and Sarazin, M. in

[Astronominal Telescopes and Instrumentation ], Proc. SPIE 4839, 837 (2003).[14] Tokovinin, A., Kornilov, V., Shatsky, N., and Voziakova, O., “Restoration of turbulence profile from scin-

tillation indices,” Mon. Not. R. Astron. Soc. 343, 891–899 (2003).[15] Tokovinin, A. and Kornilov, V., “Accurate seeing measurements with MASS and DIMM,” Mon. Not. R.

Astron. Soc. 381, 1179–1189 (2007).[16] Els, S. G., Schock, M., Seguel, J., Tokovinin, A., Kornilov, V., Riddle, R., Skidmore, W., Travouillon, T.,

Vogiatzis, K., Blum, R., Bustos, E., Gregory, B., Vasquez, J., Walker, D., and Gillett, P., “Study on theprecision of the multiaperture scintillation sensor turbulence profiler (MASS) employed in the site testingcampaign for the thirty meter telescope,” Applied Optics 47, 2610–2618 (2008).