Add an option to run OGGM on calendar years instead of hydro years

[fmaussion]

While working on #824, I started thinking again about this problem. I also wrote this email to a few WGMS specialists:

it is really a small detail, but I'd like to get it "right" once for good. In OGGM, we use the "hydrological year" as time unit, with the "hydrological year" starting at month 10 in the northern hemisphere and at month 4 in the southern hemisphere. The main reason we use this convention is because we compare our modeled MB with the WGMS, which documents it this way in the guidelines for submission (https://www.wgms.ch/downloads/WGMS_GuidelinesforDataSubmission.pdf).

But I wonder if the equator really is the best separation line for the definition of the hydrological year. In particular, I wonder if the Peruvian and Central American WGMS correspondents really use the SH definition of an hydrological year: for example, the climate in Peru is equally well suited for a NH hydrological year convention. Do you have any comment on that, Michael and Antoine?

Anyways: it's not a big deal and I'm happy to keep the equator as separation between NH-hydro and SH-hydro years. The only "funny" situation is the RGI region "Low Latitudes" where our tabular files have different dates for different glaciers as a result.

We had a similar discussion at one GlacierMiP meeting - OGGM currently stores its annual output in "hydrological years units", and we should maybe switch back to calendar time for that reason...

As always, the best would always be to give all options to the user: calendar or hydro years.

Funny enough, the problem is more complex than one would expect, in particular because it involves changes in the mass-balance, flowline, and utils.compile* modules...

[fmaussion]

the WGMS answer to this topic:

Dear Fabien,

when looking into the reported data in the FoG database, the answer is (as always) a bit more complicated then we would like it to be...

For mid-latitude glaciers, the concept of the hydrological year (NH: 01 Okt to 30 Sep, SH: 01 Apr to 31 Mar) make a lot of sense and is also implemented in most of the glaciological observation series.

For the tropics, this concept does not make a lot of sense. In fact, the observations are report for quite different dates: Bolivia typically Sep to Aug, Colombia and Ecuador typically Jan to Dec, Kenya typically Mar to Feb.

Also in the polar regions, things are a bit complicated. As such, mass balance measurements in the Canadian Arctic are carried out in spring only. This means that the measure Bw of the current and Bs of the past year and can only report Ba with one year delay...

One the Subantarctic islands, measurements are carried out between Jan and March for logistic reasons.

In the FoG database, the Table MB_OVERVIEW provide information about the observational dates and the reporting system used (FLO, FXD, STR, cf. Cogley et al. 2011), referring to the data as reported in the Table MB. For the point measurements (MB_POINT), we recommend to report the effective observation dates (FLO).

I hope this helps to clarify the situation more than to complicate it... :-)

best, Michael

[fmaussion]

Action items on the OGGM side:

• try to be as close as possible to the convention used by observers for the model validation, but report our data in calendar years for simplicity
• add a column to the WGMS metadata table we are using with a "start month", specified for each glacier based on the info we can gather from WGMS
• use this convention for the calibration / validation and cross-val tool
• run and output in this convention per default, but add an option to run OGGM on calendar years as well

Quite a lot of work again...