Siliceous and Delicious

Beaudry

Progress Report

Fixed the first-year problem: it was a historic data set, not a problem with the model. CSU model not running properly yet, been consumed with calibration issues. Was looking at diversions, realized this was silly, looked at acreages instead. Spent 4-5 days improving fit between acreages generated by model and historic acreages. Only simulating six of the nine or ten crops being planted in reality, so there’s simplification involved. Spent time fiddling with this. It got better. Spent a lot of time building a calibration interface allowing futzing with parameters at the start of each run.

Re-did crop insurance process, rewrote algorithm. Was reaching limit of what simulation could do in terms of crop acreages, couldn’t fit correctly. Needed to add variables or change logic. Thought about adding risk tolerance. Had been trying to reduce the number of behavioral variables, but in the end added it because there was no other way to fit insurance into existing behavioral set-up of the model.

Then discovered some errors from months ago, fixed those. For example: spent three hours on Wednesday trying to work out why the cost of crop insurance couldn’t be plotted on the graph. Finally found he was asking for cost rather than income…

Just been struggling to calibrate, improving fit of model to reality. Came upon dead ends with acreages, especially vegetables (overplanting) and grasses (underplanting, both relative to historical data). Discovered yesterday that back in January he had designed a mesh that had 1000m grid squares, but updated in March to a 250m grid squares, but forgot to update size variables. That improved things—have to check today what the consequences are (e.g. there had been so little water available to each grid square).

In general, it’s a struggle to get the fit better, and time is ticking. Doesn’t know how to work harder or more efficiently. Beau finds himself in “tunnel vision” mode at times, eating three or more hours in a flash on something that may or may not actually be crucial to the end product. Ben suggested working on different areas of the model for shorter bursts of time (three hours at a time).

Ben

Progress

1. FIB-SEM: booked Wednesday (today!). Spent Friday evening in the lab trying to get sample prepared. May have managed it (we’ll see!). First time around, next time will surely be easier.
2. Didn’t get a chance to figure out confocal prep.
3. No chance to look at the Mesozoic rad catalogue.
4. Asked Andy about sieves – should be some in the lab, otherwise EPS will have some. Strange numerals, though. Don’t know who to ask…

Deliverables

1. Figure out confocal prep.
2. Look at the Mesozoic Rad catalogue. Should only take 10 minutes.
3. Reply to Zoe’s e-mail which came during TG.
4. E-mail Dave regarding isotope measurements and ODP samples, etc.

Got an e-mail while he was gone from Zoe – she’s made some silica isotope measurements. She had some questions. This is an interesting conversation; wrote some term papers in the past, but hasn’t really looked at it since. One standard model for interpreting isotopes in diatoms. Very closed, localized system which doesn’t fit with actual geologic time and mixing in the ocean. But no easy alternative. Thought Ben had a while ago: don’t just look at diatoms; compare diatoms and radiolarians, picking those radios which live at deeper waters – allows depth comparison, which allows determination of input values for silica models. Guy at Harvard now who makes isotope measurements, and does them very well. Would need to know from Dave how viable it is to get ODP samples etc. Feels a lot sexier than the current projects. Zoe measured the diatoms.