Turbulence and wind speed profiles for simulating TMT

Turbulence and wind speed profiles for simulating TMT AO performance Tony Travouillon M. Schoeck, S. Els, R. Riddle, W. Skidmore, B. Ellerbroek, G. Herriot TMT. AOS. PRE. 09. 030. DRF 01 S. Els 1

On the menu today What and how are we measuring. The data available. How to read this data. How we are so far using it for AO simulations. S. Els TMT. AOS. PRE. 09. 030. DRF 01 2

Measuring turbulence with a MASSDIMM A. Tokovinin TMT. AOS. PRE. 09. 030. DRF 01 3

Where did we make those measurements Tolar La Silla Las Campanas AURA ALMA Santiago 100 km Paranal Armazones Tolonchar

Tolar

Armazones from the air The little white speck on the summit is the site testing telescope.

Tolonchar

San Pedro Mártir MASS/DIMM telescope

A special care given to crosscalibration and error management TMT. AOS. PRE. 09. 030. DRF 01 10

Wind profiles We use the NCEP/NCAR reanalysis data Publicly available Global grid (~250 km resolution) Data available every 6 hours 16 layers from 761 m to 25 km Data verified against balloon measurements We also have ground data taken with SODARs and weather stations TMT. AOS. PRE. 09. 030. DRF 01 11

Turbulence data available Huge database Between 150, 000 and 285, 000 profiles per site. Will be made public in the future. TMT. AOS. PRE. 09. 030. DRF 01 12

How can that much data be useful to AO simulations? AO simulations that are CPU intensive may not run every individual profiles The difficulty we are dealing with here is the following: There is no such a thing as an average or typical profile This difficulty comes from more the statistical nature of turbulence and is an issue on all time scales “An average profile does not have an average seeing” TMT has looked at several options TMT. AOS. PRE. 09. 030. DRF 01 13

Short and long term variations TMT. AOS. PRE. 09. 030. DRF 01 14

Median profiles How do you select a median profile? Median of each layer? Selected around Median seeing? Selected around median isoplanatic angle? See Els et al. (2009) PASP for full details TMT. AOS. PRE. 09. 030. DRF 01 15

Median profiles A solution considered for TMT simulation: Averaging profiles around the median open loop wavefront variance due to the combined effects of fitting and servo-lag error: TMT. AOS. PRE. 09. 030. DRF 01 16

A new approach…creating a standard night Auto-regressive model Generate time series that reproduce the 1 st and 2 nd order temporal statistics of the log(seeing) Uses all data to recreate a time series of seeing of manageable size for simulations Driven by temporal autocorrelation vector and Gaussian Random number generator TMT. AOS. PRE. 09. 030. DRF 01 17

A new approach…creating a standard night Keeps statistical characteristics of the site while reducing the number of profiles Method to be presented at next OSA conference in September by Herriot et al. TMT. AOS. PRE. 09. 030. DRF 01 18

Conclusion TMT has collected a high quality and statistically representative sample of turbulence parameters at 5 sites. Database to be made public. High potential for AO simulations. Lessons learned: When possible, run the simulation on all data and then calculate the statistics on the results. For time intensive simulations, it is important to realize that there is no such a thing as an average profile. Simulations give noticeably different results depending on selection criteria. TMT. AOS. PRE. 09. 030. DRF 01 19