PyAnalySeries: an interactive tool for time series exploration and alignment using pyleoclim

[PBrockmann]

Hi Pyleoclim team,

First, thanks a lot for the work you’ve done on pyleoclim — it’s a really solid and valuable toolbox for time series analysis.

I’m currently developing a project called PyAnalySeries:
https://github.com/PaleoIPSL/PyAnalySeries

The idea is somewhat complementary to notebooks. I’m fully convinced that notebooks are the best approach when you want fine control over parameters and full transparency of the methods.

What I’m trying to build instead is a more interactive application (PyQt + Matplotlib) focused on exploration and manipulation of series — especially for:

• quickly placing/editing tie-points (pointers) between series,
• performing interpolations in a very visual and iterative way,
• testing transformations with immediate feedback.

This work is also inspired by the historical AnalySeries software, which has been widely used in the community. However, it’s not really possible to port it as-is, so the idea is more to revisit its concepts in a modern Python application.

In that context, I’ve started integrating Pyleoclim calls in several places, notably for:

• Detrending,
• Frequency-domain filtering,
• PSD estimation.

The goal is really to combine your robust implementations with a more interactive workflow.

I’d be very interested to hear your thoughts on this kind of approach, and whether you see potential overlaps or synergies between the two projects.

Thanks again for your work — it clearly makes things easier on my side.

Best,
Patrick

[CommonClimate]

Salut @PBrockmann, and thank you for initiating this discussion. I think I speak for all of us here in saying that we're delighted to see pyleoclim used for any and all things! I'm sure you are right that many users would feel more comfortable in a standalone, interactive desktop app, and indeed we've thought of doing that for a long time (primarily to import/edit/export/visualize LiPD datasets).

Obviously our code is open source, so you get to use it any way you want. AnalySeries was a game changer for many people at the time (2001?) and my impression is that it could use some refreshing. I think there is an opportunity here for your work to introduce PyAnalySeries users to more options/features in timeseries analysis and visualization, since I believe pyleoclim can do most of what AnalySeries could do (except placing/editing tie-points), and much more (e.g. SSA, wavelets, outlier detection, surrogate testing). Conversely, those applications might motivate us to develop more features if we see clear demand. We'd of course love for you to be involved in pyleoclim development if you have bandwidth to implement those features.

Re: dependencies, @khider has long enshrined best practices into our development. We are very mindful of making API changes at this point so we don't break everyone's code downstream, but PyAnalySeries would be the first third-party package to depend crucially on our API stability. We try to be good about issuing deprecation warnings and the like, but if you have suggestions on how we can do this better, we are happy to hear them.

Happy to hop on a Zoom call anytime (if we can bridge the 9h time difference...).

Amitiés,
Julien

[khider]

Hi Patrick,

I used to use AnalySeries and would be very happy to see the methods we have in Pyleoclim reaching a broader audience though a more GUI approach. Happy to help you anyway we can. As @CommonClimate mentioned, let us know if you want to jump on a call.

Best

Deborah

[PBrockmann]

Thanks for your message — that’s exactly the direction of PyAnalySeries.

The application was initially focused on interpolation, in particular the placement and editing of tie-points. More recently, it has been extended with features leveraging Pyleoclim methods, including detrending, frequency filtering, and power spectral density analysis. These functionalities are already integrated and usable in the current version.

A call would indeed make sense. I’ll discuss it with my (informal) steering group and get back to you shortly.

Frequency Filter

Power Spectrum (PSD)

[PBrockmann]

Hi,

It seems that insolation-related computations may currently be a missing component in Pyleoclim.

I have just refactored the insolation/astronomical computation code from PyAnalySeries to clearly separate the calculation backend from the user interface. The core logic is now implemented in a standalone module, making it easily reusable outside the GUI environment—for instance directly within notebooks.

This modular approach could facilitate integration into Pyleoclim, leveraging the existing inso package for astronomical solutions. Such an addition would likely be valuable for the paleoclimate community, by providing a consistent and accessible way to generate orbital parameters and insolation series within the Pyleoclim ecosystem.
What do you think of this ?

Here is a capture from a notebook.

Best
Patrick

[CommonClimate]

Oh that would be SO, SO useful! In our tutorials we've used climlab to compute insolation, but there are multiple reasons why your solution is preferable. I think it would make sense to have that as a standalone package that could be a dependency of Pyleoclim (PyInsol?), unless it uses some very weird packages.

[PBrockmann]

Insolation calculations rely solely on the Insolation module, and integrating the results into Pyleoclim time series objects should be fairly straightforward.

I recently separated the core use of the Insolation module from the construction of insolation-related quantities, so it can be reused more easily in other contexts. See:
https://github.com/PaleoIPSL/PyAnalySeries/blob/main/resources/insolationAstroSeries.py

Happy to discuss how this could be integrated on the Pyleoclim side if useful.

Note that the Insolation module will be updated very soon (use for now pip install inso==0.7.0)

[CommonClimate]

I was not aware of the inso package, which appears to be what you use. I'll take a look and see if that is something we want to include. Is it documented anywhere?

[PBrockmann]

The inso module is now available on PyPI (v1.2.0):
https://pypi.org/project/inso/

[CommonClimate]

Thanks @PBrockmann. Indeed the code is available, but how to use it is rather mysterious without proper docstrings. How would we do this with inso, for instance? Writing documentation is time-consuming, but it does help!

[PBrockmann]

Yes, I fully agree. Proper docstrings and usage examples make a big difference.

That said, I am not the author of the inso module myself.

For inso, this notebook shows how to compute the standard quantities currently proposed in PyAnalySeries:
https://github.com/PaleoIPSL/PyAnalySeries/blob/main/notebooks/inso_02.ipynb

Reproducing the same calculations in Pyleoclim should be fairly straightforward.

[CommonClimate]

Thanks! That's a useful example. I will give it a try in the next few days.