search

PalEON-Project / stepps-baconizing

Data, code, and output from the Bacon age-depth models developed for use in the STEPPS prediction runs.
0 stars 0 forks source link

Review core fits #26

Closed andydawson closed 7 years ago

andydawson commented 7 years ago

Noting concern that some of the bacon age estimates do not pass through the older radiocarbon dates. Need to flag problem cores, discuss, and modify protocol accordingly.

andydawson commented 7 years ago

Following are the the handles of the sites that need further consideration:

PORTAGE - Fine UPGRAVEN - Fine WOLFCRK - Misses last date in Bacon and Bchron. MNCVTL - Fine STEELMN - Fine HELLSKIT - Varve where depths are varve ages. Breaks in Bacon and not in Bchron. DARKLAKE - Varve where depths are varve ages. Breaks Bacon. DEVILSWI - Super weird. Re-check. DISTRHFT - Fine LITTLPIN - Varve where depths are varve ages. RUBYLAKE - Varve where depths are varve ages. CAMPTWO - Fine GREENLK - Fine GLIMML1 - Fine NELSONMI - Fine GUARD - Fine KERRMI (15916) Super-fast accumuation rate that model cannot handle with generic priors. Re-run with higher accumulation rate prior. CROOKEDN - Fine

Will edit reply to include dataset ids shortly. Also need to fix bacon plotting limits so that all chron controls show up on figures.

andydawson commented 7 years ago

After reviewing the problem cores, we have made decision to accept or modify based on additional information.

The following cores have been identified as varves (some have radiocarbon dates calibrated from calendar years), and therefore age-depth models will not be run: HELLSKIT (3131) - Varve where depths are varve ages. Breaks in Bacon and not in Bchron. DARKLAKE (546)- Varve where depths are varve ages. Breaks Bacon. LITTLPIN (1643)- Varve where depths are varve ages. RUBYLAKE (2309) - Varve where depths are varve ages.

Kerr Lake (15916) has a very high accumulation rate and needs to be re-run with a larger accumulation rate prior. Associated publications were checked and no evidence suggesting anything unusual about this core was found, so we accept the radiocarbon and core top dates.

Finally, DEVILSWI needs to be rechecked. The Bacon age-depth model looks suspicious for the older part of the core. We see reduced uncertainty in between the three oldest radiocarbon dates. Bacon does not pass through any of these three older dates.

andydawson commented 7 years ago

All fixed!

Bacon model no longer running for cores with varve ages as depths. Varves can be added to the prediction dataset with zero age-depth error if desired.

Kerr Lake - set the accumulation rate mean prior to 100 yr/cm which resulted in a reasonable fit.

Devils Lake - There were some reservoir corrections that were not included in the dates in the chron control table in Neotoma but were in the notes associated with the bacon metadata for the age model on Neotoma. I re-ran the Bacon model with these reservoir corrections and things look much better.