Weekly timeseries plot not showing some data, inserting others

[sgwinder]

Data points are being dropped if they are too near the beginning of the month. For example, at Lennox Creek (siteid 181), the data for May 3, 2020 is not being displayed. I think that this could be a side effect of the "month" being 4 (since weekstart was in April), but d2p being in May? It also appears that the dashboard is displaying dates as d2p + 1. I think we should use either d2p or weekstart directly. I have been creating weekstart to be the first date in the week we are summarizing (always a Thursday), and made d2p the following Saturday. Maybe the most straightforward would be to create the plot with weekstart.

And, other data points seem to come out of nowhere, near the beginning of the year. Again at Lennox Creek, there is an extra data point between December 29, 2019 and January 5, 2020. This is not in the viz_model_www.csv table.

I have no idea where this one came from!

[aromatic-toast]

@sgwinder,
Can you refresh me on what d2p represents?
And also, which project contains lennox creek? I was looking for it under Western Washington but was unable to find the polygon.

[sgwinder]

d2p stands for, I believe, "date to plot". Right @woodsp? Originally, d2p was not actually based on the underlying dates going into the estimates, but rather started on January 1 and then just added 7 to go through the year. I wanted more precision when I was summarising estimates and social media data, so I created weekstart and have been using it in my code. I've been creating d2p to write out for compatibility with the dashboard (but see Issue #126 ).

And Lennox Creek is in the Middle Fork project, sorry for the confusion.

But, the problem I described above seems to be true for all sites, not just Lennox Creek. (If you put your mouse over the "weekly" timeseries and scroll, you can zoom in).

[sgwinder]

This is still a problem. I think the fix is probably to create the weekly time series based on the d2p column directly.

An example of weird behavior is here, at the Pilchuck Lookout Trail in the South Mountain Loop project.

[EmiliaH]

@sgwinder, is this still a relevant issue?

[sgwinder]

Yes, this is still a problematic bug on the dashboard. Would be great to add to the list of potential improvements for David to investigate.