The PAU Survey

Enrique Fernández

The Physics of the Accelerating Universe Survey (PAUS) is an ongoing extragalactic survey carried out with the William Herschel Telescope in La Palma, Canary Islands, equipped with the PAUCam Camera. The distinctive feature of PAUS is the ability to measure redshift of galaxies by photometric methods, with roughly an order of magnitude higher accuracy than that provided by other past and existing photometric surveys. This is accomplished by equipping PAUCam with 40 narrow-band filters, each 130A wide (FWHM) and equally spaced by 100A, spanning the region from 4500A to 8500A [1]. In addition the camera also has a set of six standard broad-band filters, in the six bands u, g, r, i, z, Y, and an external large size filters (full field of view), also in those bands.


PAUS originated in the context of the PAU Project, funded, in 2007, by the Consolider Ingenio 2010 Program of the (at the time) Spanish Ministry of Research and Innovation.

The PAU project was approved in 2007 and ended in 2014. Its main deliverable was the PAUCam camera, built by 5 of the 7 groups that were originally in the Consolider Project, namely from CIEMAT and IFT (in Madrid), and from IEEC, PIC and IFAE (in Barcelona). These groups also developed the large amount of software needed for the control of PAUCam and for the data processing from their production at the Telescope to their analysis at the labs [3,4]. The same groups also collaborate closely in other projects, notably in DES, DESI and EUCLID, described elsewhere in this report.

PAUCam operates as a Visitor’s Instrument at the prime focus of the William Herschel Telescope (WHT) in the Canary Island of La Palma. Starting in 2016 other groups have joined the PAUS Collaboration, namely from Durham University, Plymouth University and University College of London in the UK, from Leiden Observatory in the Netherlands, from ETH in Switzerland, and from Bonn University in Germany. A group from Tsinghua University in Beijing, China, joined in 2021. The observing nights are granted from the Isaac Newton Group of Telescopes (ING), a Consortium of United Kingdom, the Netherlands and Spain, that operates several telescopes at the La Palma site, the WHT among them. The proposals for observation periods are submitted twice per year to the TACs (Time Allocation Committees) on those three countries, that advise the ING Management.

PAUS Operation

Since first light in 2005 until the end of 2019, PAUS has observed for about 215 nights with high efficiency (only 8.9 effective nights lost), but unfortunately with very bad weather conditions, particularly in the fall and winter. For that reason the effective number of good observing nights has only been half of the above, namely 101 nights of good data. PAUS has chosen to observe in fields where redshift data is available from other observations, either photometric or spectroscopic. These include the COSMOS field [5], containing over one million galaxies, collected from several telescopes (in satellites and ground-based), with a coverage of 2 square degrees in the equatorial region, and the W1, W2, W3, W4 fields of the CFHTLS [6]. The COSMOS field has been completely covered by PAUS while the CFHTLS fields are not yet completed. In terms of square degrees we do not quote detailed numbers, as they depend on the particular analysis being pursued and also vary depending on the observation strategies. A rough number is 0.7 square degrees per good night of observation. During the two years 2020 and 2021 no data has been taken, due to ongoing work in the WHT telescope needed to accommodate the future WEAVE spectrometer, and also to the occurrence of two major disruptions in 2021, the Covid-19 pandemic and the eruption of the Cumbre Vieja volcano in the south of the La Palma island. Future running of PAUCam will be possible after the installation of WEAVE, but detailed plans are not yet developed.

Scientific results

The PAUS objective is to survey an area of about 100 deg2 down to magnitude iAB ≈ 22.5 with a redshift error σ68=0.0035 (1+z). The redshift precision has already been achieved for the best 50% of all sources, selected on the basis of a quality cut, as published in [2]. The code to accomplish this task was the template-based code BCNz2 [2]. But as already mentioned in the 2020 report, a new code, named DEEPZ, was developed inside the PAUS Collaboration for redshift determination based in Deep-Learning techniques and has also been published [7]. This results in an improved determination of redshifts, particularly at high magnitudes. Deep Learning techniques have also been applied to improve the photometry of galaxies, and important topic for future surveys. A new code, named Lumos, has been developed inside PAUS for that task. It improves the Signal to Noise Ratio by a factor of 2 with respect to the aperture photometry algorithm, and it is more robust when correcting some imperfections we had in the images, such as scattered light affecting the periphery of some ccd’s. This work is now published [8]. Fig. 1 shows the distribution of the error-normalized differences between two determinations of the flux of the same galaxy in the same narrow-band filter, for a large sample of galaxies. The red curve is estimated with MEMBA (the standard PAUS algorithm) and the black is that obtained by Lumos. Clearly the Lumos distribution is much closer than MEMBA’s to the Normal distribution, expected from such a statistic.
Figure 1: Distribution of the error-normalized differences between two determinations of the flux of the same galaxy, obtained by two different methods (see text).
The PAUS collaboration has also published five more papers covering several topics: on forecasting Lyman_Alpha intensity mapping combining the PAUS data with that of eBOSS and DESI [9]; on obtaining an improved photo-z sample in the COSMOS field [10]; on projecting the intrinsic alignments and clustering measurements using the PAUS narrow-band filters [11]; on photo_z measurements with Gaussian processes [12]; and on measurements of galaxy properties with approximate Bayesian computations [13].


[1] The Physics of the Accelerating Universe Camera Cristobal Padilla et al. Astron.J. 157 (2019) 6, 246 e-Print: arXiv:1902.03623v2 [astro-ph.IM] 28 Mar 2019
[2] The PAU Survey: early demonstration of photometric redshift performance in the COSMOS field M. Eriksen et al., Mon.Not.Roy.Astron.Soc. 484 (2019) no.3, 4200-4215 e-Print: arXiv:1809.04375 [astro-ph.GA]
[3] The PAU Survey: Operation and orchestration of multi-band survey data Nadia Tonello et al. Astron.Comput. 27 (2019) 171-188 e-Print: arXiv:1811.02368 [astro-ph.IM]
[4] The PAUS Survey: Data Reduction of Narrow Band Images. PAUS Collaboration (to be published, now under collaboration review).
[5] COSMOS survey page:
[6] The Canadian French Telescope page:
[7] The PAU Survey: photometric redshift using transfer learning from simulations M. Eriksen et al., Mon.Not.Roy.Astron.Soc. 497 (2020) 4, 4565-4579. e-print:
[8] The PAU survey: Estimating galaxy photometry with deep learning Laura Cabayol et al. Mon.Not.Roy.Astron.Soc. 506 (2021) 3, 4048-4069 e-print: [astro-ph.CO]
[9] The PAU survey: Lyα\alphaα intensity mapping forecast Pablo Renard et al. Mon.Not.Roy.Astron.Soc. 501 (2021) 3, 3883-3899 e-print: [astro-ph.CO]
[10] The PAU Survey: An improved photo-𝑧 sample in the COSMOS field Alex Alarcon et al. Mon.Not.Roy.Astron.Soc. 501 (2021) 4, 6103-6122 arXiv:2007.11132v2 [astro-ph.GA]
[11] The PAU Survey: Intrinsic alignments and clustering of narrow-band photometric galaxies Harry Johnston et al. Astron.Astrophys. 646 (2021) A147 arXiv:2010.09696v2 [astro-ph.GA]
[12] The PAU Survey: narrow-band photometric redshifts using Gaussian processes John Y.H. Soo et al. Mon.Not.Roy.Astron.Soc. 503 (2021) 3, 4118-4135 e-print:
[13] The PAU survey: measurement of narrow-band galaxy properties with approximate bayesian computation Luca Tortorelli et al. JCAP 12 (2021) 12, 013 e-print: [astro-ph.GA]