In an ideal world, we aim for having a collection of Hierarchical surveys that would represent the whole landscape of images in astronomy: the largest available sky coverage (whether observations are scattered pointed fields or global sky surveys), images at all wavelengths, images at all resolutions up to the finest ones, images of all depth up to the deepest ones, ... In practice, while we keep that aim, we have to make choices according to the availability of the data as well as resources (human, storage, computing time,...) and this document aims at defining our strategy, laying down our guidelines and criteria when ingesting images into HiPS.

First, we can only ingest data sets we are aware of. We have various ways to check for image data sets of interests, yet it is difficult to be complete. We hope that with the variety of information sources and people at CDS that perform the watch (either naturally on their daily routine or purposefully), we are not missing datasets that would be of high priority to ingest. Users are welcome to suggest datasets they feel would be important to ingest by sending an email to cds-question@astro.unistra.fr

One fundamental criteria for ingestion we have chosen refers to the quality: we create HiPS out of data sets that have been published, i.e. there is a paper associated to each dataset we transform. In practice, this means for example that images taken by amateur astronomers that do not have an associated publication are not ingested. Example: The Mellinger survey is available as a HiPS, while images from Ciel Austral group were not ingested.

Another fundamental guideline is that we always favor situations where experts of the datasets are making the HiPS. We will ingest images for which the data providers are not making a HiPS themselves, but we always prefer to encourage that the people who know their data best make the image selection, documentation and HiPS creation. We therefore are happy to help teams that would need advice or feedback and will favor this kind of interaction. Our goal and missions are clearly not to create at CDS all the possible HiPS from all existing datasets, even if we had the resources to do so. This is also a way to improve the quality of the HiPS (ideally made by the people who know best) as well as a criteria to select datasets to ingest, and a way to foster the usage of HiPS by a wider community.

In practice, there is however a large need that CDS creates a given amount of HiPS directly. Even if we presently manage to do most of the HiPS we would like to ingest, even if sometimes with a long delay and different priorities, we are already not able to create all the HiPS we would like to and this implies to put priorities.

For any new identified image dataset that can potentially become a HiPS, an evaluation is done of whether a HiPS should be made and how high a priority it is. In defining this selection and priorities, several aspects are taken into account and weighted before making a decision:

• How does it completement existing HiPS data sets? This includes arguments like the sky cov- erage, the availability of other HiPS in the same wavelength range, whether it brings a higher resolution or a deeper intensity view, ...
• What is the added value of doing this HiPS given the already existing image data sets?The added value can include cases like gaining the effect of a global view that is not available with the collection of small individual images. It can also be the fact to add a new data release of a dataset for which we already have a HiPS, or to redo a HiPS to be able to have access to FITS images and hence pixel values and dynamics with respect to an existing HiPS with only PNG files.
• Is there a specific opportunity? This includes cases like someone willing to invest a certain amount of effort for a specific survey (e.g. D. Durand collaboration for HST and JWST HiPS),a special moment to highlight our services (e.g. JWST early release data sets) or an internal interest (e.g. redoing the GALEX surveys as a testbed for filtering tools).
• How easy/difficult is it to do? E.g. a survey distributed as a HEALPix file is a straightforward product to add and is usually an easy addition to the collection for a small amount of computation and storage resources. On the other hand, very large optical surveys may require months of computation and some image data sets may require specific knowledge about the data (e.g. which files, pixel units, filters, keywords for blank pixels or instrumental effects, ... ). Again, we reinforce the idea that we always favor the case where experts of the datasets or providers of the datasets make the HiPS. Hence the last question we ask ourselves:
• Can we encourage and help others to do this HiPS?

With all these elements to weight on, we assign a priority degree to the considered data set and compare it with other surveys that are currently in our waiting list. In practice, this means that different data sets will take different times to get ingested and become a HiPS. As the data amounts continues to rise, we will eventually set a limit at which some datasets will not be considered. At the moment, this has mainly happened when no publication was associated or if the HiPS was already available elsewhere (even in a different format, e.g. PNG only).