• No se han encontrado resultados

meãma i taaike, ñṵãta naane arü gṵ na tiã daugüü̃, ngḛta na ũgü niĩ tüü̃ na kàgü,

Anexo 5: VOCABULARIO TIKUNA

1. La situación del ritual en Arara, la voz de la curaca.

An instantaneous classification is an extremely important function to catch bright and rare transients. It is however anticipated that sources may change classification once more information has been collected. In the flow diagram shown in Figure 4.1 I have included two steps prior to visualisation - ‘Update Source Monitoring’ and ‘Re-fit Archival Images’. Once the positions of candidate transients have been identified, elliptical Gaussians should be fitted to that position in all further images that are processed and potentially all previous images, if available. Furthermore, if a transient falls – in flux – below the RMS, the RMS should be submitted to the database and subsequently used in the variability search. In the early testing I conducted, the database was not yet capable of correctly handling the RMS mea- surements. Elliptical Gaussians had to be fitted manually to the region of interest to derive the complete lightcurve.

I therefore experienced a few examples of interesting transient behaviour slip- ping through the net of the transient detection algorithms, with regards to RMS measurements. In one case a test dataset was created using a VLA image of the source 3C48. The field contains a bright point source at the centre; this source was modelled and removed from the image, leaving a relatively blank field. A sequence of images were then created (by updating the time stamps in the FITS headers) us- ing the image with 3C48 present as well as the image with 3C48 removed. Three images of 3C48 removed were followed (in time) by one image of 3C48 present (unremoved) and continued for a number of cycles. The transient pipeline identi- fied the source when it was ‘on’, however, it did not submit the RMS to the database when it was ‘off’. As the source had the same flux (when it was on) the variability query incorrectly classified it as having Vv= 0. After the data had been refitted i.e. utilising the RMS when the source was off, the true variability could be quantified. In the top left panel of Figure 4.4, I show a lightcurve depicting this behaviour for clarity. Both the processes of refitting and monitoring are represented in the flow di- agram in Figure 4.1 with a loop that cycles back to the transient search and database. In future versions of the database this loop, including the re-fitting/monitoring will be fully implemented.

A further example is shown in Figure 4.5 (note this figure shows the actual lightcurves extracted from the database). An early 12 hour LOFAR commissioning observation (L2010 06928, see Chapter 6) was calibrated and imaged by Dr. Jess Broderick. The dataset was then divided into 12× 1 hour chunks and re-imaged. The final FITS images were then processed through the LTraP. The full 12 hour im- age was used as the first image (in the sequence of images being searched), as it had the largest number of sources and best image fidelity. In this particular example, the consequence of using an image with higher dynamic range as the first image, is that many sources in the following images sit below the source extraction thresh- old, due to the increase in RMS (see the top right panel of Figure 4.4 for a further example). In accordance with the flow diagram the source was refitted/monitored. See the bottom panel of Figure 4.5 for the fully fitted lightcurve. We can see that the source is not particularly variable; there was a slight variability attributed to un- certainty in absolute flux calibration. This was a commonly observed false transient trigger which is largely solved by monitoring sources. It is obviously anticipated that LOFAR will on occasion perform a very deep observation of a given field. This will again potentially trigger a number of single epoch transient detections, if not handled correctly. In parallel to monitoring and re-fitting, the single epoch transient search described above should be improved by using the RMS values, to assess if

0 5 10 0 0.5 1 1.5 Time Flux Real Lightcurve Detection level (5σ) Current Lightcurve 0 5 10 0 0.5 1 1.5 Time Flux Real Lightcurve Detection level (5σ) Current Lightcurve 0 5 10 0 0.5 1 1.5 Time Flux Real Lightcurve Detection level (5σ) Current Lightcurve 0 5 10 0 0.5 1 1.5 Time Flux Real Lightcurve Detection level (5σ) Current Lightcurve

Figure 4.4: Example (simulated) lightcurves of transient behaviour that LTraP currently struggles to correctly identify. The flux and time are plotted in arbitrary units. The red lines show the current lightcurves that are extracted using the LTraP. The blue line shows the correct lightcurve, which utilises the RMS measurements in the transient query.

particular sources would be expected to be detected in different images.

Two further examples are shown in Figure 4.4. On the bottom left panel the red line shows the source lightcurve, without the RMS measurements. The blue line shows the true lightcurve of the source. By using the RMS measurements we can only set an upper limit on the variability of the (simulated) source, with respect to the detection threshold. The bottom righthand panel shows a mix of variability and RMS measurements. In this case, from the time 1 to 3, by using the RMS measurements we may define this source as highly variable. From the time 4 to 7 the detection threshold is above the transient flux. It is difficult to use the RMS measurement in this case, because we cannot be sure what changes in flux have occurred. The pipeline will need to understand the difference in these two cases

Figure 4.5: Top panel: A lightcurve retrieved from the database of a single epoch transient automatically identified by the pipeline in LOFAR commissioning data. Bottom panel: The full lightcurve of the same source when refitted/monitored with elliptical Gaussians. In this example the sequence of images consisted of one 12 hour LOFAR observation, followed by re-imaged one hour consecutive chunks of the original full 12 hour dataset. The first 12 hour image yields one unique source, because the source sits below the detection threshold in the subsequent images. This error is rectified by re-fitting/monitoring unique source positions and then reassessing the variability.

and respond accordingly.