We can also look at global daily absolute temperatures (rather than anomalies), which show just how much global temperatures have shifted over time:

Reminds me of the wiki articles that students at Utrecht University wrote for an oceanography course, instead of a typical paper assignment: www.uu.nl/en/news/wave...
So much nicer than laboring for hours for something never to be revisited (except, hopefully, by your own memory)

Thanks a lot to my co-authors: Michael Denes, Siren Rühs, Øyvind Breivik, Tor Nordam and Erik Van Sebille!

We explore all of this in the manuscript. We provide technical explanations for the cause of this bias but also do some more practical experiments in the Atlantic using typical mesoscale set-ups, and find the bias is significant at timescales of less than 180d 😬 Previous research could be affected!

We find that this is due to a reversal of stability to numerical errors in divergent and convergent regions as time is reversed. But not just RK4 (used in software like Parcels) is affected; also the analytical scheme (TRACMASS and ARIANE) can suffer from biases near the surface due to the B.C.s!

This fig shows which bins in the ACC channel are over- (red) and underrepresented (blue) in particle counts after forward and backward integration. Esp Runge-Kutta 4 (with dt = 10 mins, which is on the shorter end of typical sims) shows significant biases with stationary and time-evolving flows.

In one experiment, we initialized particles uniformly in an ACC channel model (mesoscale) and tracked particles forward and then backward for 180 days. With perfect numerics, particles should then just end up at their original locations, but numerical errors shake things up a lot! (2/n)

(2/2) We track DIC changes over a year and split these into biogeochemical and mixing contributions. For 4 pathways, we compute magnitudes and timescales at which BGC and mixing alter DIC. Mixing during subduction changes DIC in a parcels the most, but most parcels persist in NASTMW over the year.