In this paper we presented the efforts undertaken at APEX to obtain the best possible absolute intensity calibration for the science data product which is delivered to the end user (the PI). We introduced the various steps taken at (or even before) the observations until the reduction pipeline, and estimated the uncertainties introduced already in the observation process, like pointing and focusing error.
For bolometer observations, the array parameters can be estimated very accurately and their errors are negligible. The main uncertainty for calibration comes from the opacity determination. At APEX, we use a method based on skydips and radiometer measurements to minimize the errors. In addition, regular calibrator observations are used to further reduce the remaining uncertainty, which is estimated to 10% for LABOCA observations and 25% for SABOCA observations.
For heterodyne data we found that the calibration scale at APEX is in very good (APEX-2) to moderate (APEX-1) agreement with other telescopes, and estimated the calibration uncertainty to 8% and 12% for APEX-1 and APEX-2, respectively. For the latter a dicrepancy between LSB and USB tuning has been found, and the USB numbers seem to be more reliable.
Based on the findings presented here, we will continue to improve the calibration methods in the future to obtain an even lower absolute calibration uncertainty for APEX data. Especially for SABOCA data we are working on a flux density scale for secondary calibrators, as well as on an improved opacity determination, in order to reduce the errors. As a courtesy to the user, the APEX staff maintains a set of webpages where the calibration information for LABOCA[17] and the results from the SHFI monitoring program[18] are available.
Acknowledgements. We would like to thank all APEX staff and partner institute astronomers who have carried out calibration observations throughout the years, as well as Giorgio Siringo and Axel Weiß for their valuable input about bolometer calibration. We acknowledge the use of gnuplot[19] and R[20] for some data analysis and plotting purposes.