Initial turbidity data exploration#100
Conversation
KECognac
left a comment
KECognac
left a comment
There was a problem hiding this comment.
Hi B,
I left my main comments in the document. Overall complex relationships that will need to be teased out. Interesting.
-Kristen
| select(datetime, canyon_mouth_cfs = value) %>% | ||
| mutate(date = date(datetime), | ||
| year = year(date), | ||
| doy = yday(date)) |
There was a problem hiding this comment.
I played with a smoothed q during my review - %>% mutate(smooth_q = frollmean(canyon_mouth_cfs, 7))
| mutate(deltaQ_24h = canyon_mouth_cfs - lag(canyon_mouth_cfs), | ||
| deltaQ_48h = canyon_mouth_cfs - lag(canyon_mouth_cfs, n = 2), | ||
| deltaQ_72h = canyon_mouth_cfs - lag(canyon_mouth_cfs, n = 3), | ||
| .by = location) |
There was a problem hiding this comment.
Nothing special here - adding v simple code to look at smooth lags also - deltaQ_24hs = smooth_q - lag(smooth_q), deltaQ_72hs = smooth_q - lag(smooth_q, 3), deltaQ_1wk = smooth_q - lag(smooth_q, 7),
There was a problem hiding this comment.
Hey B -- Interesting relationships (and at times lack thereof). Definitely, upon looking, key drivers for high values are likely always allochthonous. As we discussed, the hysteresis relationship between smooth_q and turbidity (p95 slightly less so but definitely apparent) is a something that the model will need to resolve. I have no other major thoughts for now, but happy to keep exploring this over your shoulder :D.
-Kristen
There was a problem hiding this comment.
rockstar. I started an adaptation of the kemper method - super curious if it will transfer at all to a mountain west/snowmelt dominated system (it was developed in NYS) and definitely gets at some of these additional parameters (but in a different way). Thank you for reminding me that paper existed.
2 participants
Hey KC, having you look over since Sam is out of town and I forced you to talk about this yesterday. Sam, I'd also like your feedback when you're back if you have any!
This is purely a data exploration exercise - I've knit the document, it is in the docs/uclp_dss/docs/knit_markdowns/data_exploration folder with the same name as this Rmd.
I think the primary question is, is there anything that I'm fundamentally missing at a high level? If we were going to do a full watershed scale assessment and a really intensive model, we'd be getting into spatial analysis, but I don't think that's really the vibe here for this particular application and for the intention of the Engine in general.
Thanks y'all.
Progress on #57