[Paper Review] The Optical Gravitational Lensing Experiment. Real Time Data Analysis Systems in the OGLE-III Survey

A. Udalski|arXiv (Cornell University)|Jan 8, 2004

Geophysics and Gravity Measurements157 citations

TL;DR

This paper presents three real-time data analysis systems—EWS, EEWS, and NOOS—developed for the OGLE-III survey to detect microlensing events and transient objects in near real time. By enabling rapid photometric analysis of 200+ million stars across 35×35 arcminute fields, the systems achieved sub-minute data reduction and detected transients like supernovae and high-magnification microlensing events with a detection threshold of I ≈ 19.8 mag and a typical lag of 6–10 days after brightening.

ABSTRACT

We describe three real time data analysis systems implemented during the third phase of the OGLE survey (OGLE-III). The EWS system is designed to detect on-going microlensing events. The EEWS system monitors the microlensing phenomena for anomalies from the single mass microlensing. The NOOS system detects transient objects in the OGLE-III fields (SNe, microlensing events, variable stars) that normally are below the detection threshold. Information on objects detected by each of these systems is distributed to the astronomical community for follow up observations. Also a short description of the OGLE-III hardware and photometric data pipeline is presented.

Motivation & Objective

• To enable real-time detection of microlensing events and transient objects in the OGLE-III survey due to the high data volume (~3.5 TB/year) and need for timely follow-up.
• To overcome data saturation by implementing on-line photometric reduction within minutes of image acquisition.
• To detect faint, transient objects that briefly exceed the detection threshold of reference images, including SNe, microlensing events, and variable stars.
• To provide rapid alerts to the astronomical community for follow-up observations using automated systems.
• To ensure high-cadence sampling (1–3 nights) for well-covered light curves of transient phenomena.

Proposed method

• The EWS system detects on-going microlensing events by monitoring brightness changes in real time using a sliding window approach on photometric light curves.
• The EEWS system identifies anomalies in microlensing light curves by comparing observed deviations from the standard single-lens microlensing model.
• The NOOS system scans nightly database updates for new objects with at least two detections, flagging them for visual inspection and photometric re-analysis.
• Finding charts are generated for candidates, and promising transients are masked to prevent contamination in future scans.
• Photometry is re-computed using the standard OGLE pipeline, and light curves, images, and web/ftp archives are generated for each detected transient.
• Alerts are distributed via email to subscribers through the OGLE-NOOS mailing list, with public access to data via web and ftp repositories.

Experimental results

Research questions

• RQ1 How can real-time photometric analysis be implemented at scale to handle terabyte-level data from a large-aperture sky survey?
• RQ2 What is the optimal system architecture for detecting microlensing events and transient objects before they reach peak brightness?
• RQ3 How effective are real-time systems in detecting faint transients below the standard detection threshold of reference images?
• RQ4 What is the typical time lag between transient brightening and detection in a high-cadence survey?
• RQ5 Can real-time systems reliably distinguish true transients from artifacts caused by saturated stars or detector defects?

Key findings

• The OGLE-III photometric pipeline reduced all images within several minutes of acquisition, enabling real-time data processing for all survey fields.
• The EWS system detected on-going microlensing events with a success rate of 40–80 detections per year during OGLE-II and was successfully reinstalled for OGLE-III.
• The NOOS system detected more than ten transient objects in the Magellanic Cloud fields within its first month of operation, with ten additional detections in earlier data.
• The typical detection threshold for transients was I ≈ 19.8 mag, with a lag of 6–10 days between brightening and detection.
• The system successfully identified supernovae such as OGLE 2003-NOOS-005 and OGLE 2003-NOOS-011, often detecting them before maximum light.
• The NOOS system was expected to cover Galactic bulge fields starting in 2004, where most transients were predicted to be high-magnification microlensing events of faint bulge stars.