Plots of minimum and maximum values of 2D model fields during a specified time period.

[BernardClaxton]

What problem does your feature request solve?

Currently CSET can produce 2D plots of surface fields e.g. visibility for every required model timestep. It would be useful to be able to produce plots of the minimum , or maximum, values of a model field over a given time span. Examples could be (i) a plot of the minimum visibility diagnosed over the 24 hour period noon to noon next day to capture all overnight fog or (ii) a plot of the minimum nighttime temperatures over the model domain or (iii) a plot of the daytime maximum temperatures or (iv) a plot of rainfall rates over a day to gauge the extent of downpours. These plots could be compared to similarly time aggregated plots of observations for model evaluation. This method of comparison mitigates the "double penalty" associated with a slight mistiming or misplacing of a model event e.g. shower or fog patch, compared to observations.

Here is an example of using this technique to plot some visibility observations. The left panel plots the observed visibility at 05Z and the right panel plots the minimum reported visibilities from each station overnight. The right panel gives a good indication of the extent of the overnight fog at the expense of washing out the temporal evolution of the fog.

Describe the solution you'd like

I'd like to see an option in CSET to select these time aggregated plots that implement a condition or function on the model field e.g. maximum or minimum value over the time span. As well as specifying the condition, it would be useful to specify the time span perhaps as a start time and end time, or as a time offset .

Describe alternatives you've considered

@Sylviabohnenstengel @ukmo-huw-lewis @jfrost-mo @BernardClaxton

[ukmo-huw-lewis]

Agreed that this is useful capability for evaluation, and mirrors some of the functionality existing in ukep_plot (e.g. see Fig 3 and others in https://rmets.onlinelibrary.wiley.com/doi/full/10.1002/qj.3488).
Issue #1285 and PR #1377 begins to introduce some of the required functionality. Once this is lodged I propose testing either extent to which that can already deliver required outputs (potentially requiring different chunking of input data), or (perhaps likely needed) the MEAN, MAX, MIN standard iris.analysis methods might be supplemented to include further derived 'collapse' functions such as to compute series of DAY_MIN, DAY_MEAN, DAY_MAX etc for all 'DAY chunks' in an input data series etc. Some of these are considered in ukep_plot (e.g. see https://code.metoffice.gov.uk/trac/utils/browser/ukeputils/trunk/ukep_plot/ukep_mod/plot_ukep.py#L173 or other instances of self.accum).

Propose to look at this again once PR #1377 reviewed and merged.