Sensitivity tests
Post-processing of sensitivity tests
Integration ranges
- Depth layers (m) defined on deptht_bounds: 0-102.4202, 102.4202-482.4563, 482.4563-995.6885. NOTE: to compute vertical integrals, concentrations or rates have to be weighted by the thickness of their layer. It's basically a weighted sum.
- Periods: in diad files with monthly output: months 1-4, 5-8, 9-12. In ptrc files with 5-day output: time indices 1-24, 25-29, 30-73. For next runs it might be better to save monthly output of ptrc variables. NOTE: use only data from year 5, no need to compute multiyear averages because years 5-10 are practically identical.
List of variables
- From ptrc files: POC, PHY, ZOO, POC+PHY+ZOO. Original units: mmol m-3. Vertically integrated units: mmol m-2.
- From diad files: Terms directly available in file: Z1NoAssim, GOC2POC, Z2FragmGOC, Z1IngestPOC, POCRemin, POC2GOCAgg. Terms that have to be computed from others: Mort2POC = NMortNoCAL + DLinMort50 + Z1LinMort + Z1QuaMort; DOC2POC = DOC2POCAgg + DOC2POCBro; Z2ff_IngestPOC = Z2IngestPOC + Z2ffPOC. Original units: mol m-3 s-1. Vertically integrated units: mmol m-2 d-1.
- Also from diad files: next transport by gravitational sinking. This cannot be computed in the same way as other budget terms because POC sinking rates, EXPPOC, have units of mol m-2 s-1. Instead of vertical integrals, here we just need the NET transport, calculated as EXPPOC_top (what enters the layer through its top cell) minus EXPPOC_bottom (what leaves the layer through its bottom cell). Final units: mmol m-2 d-1.
Plot design
For each variable (I counted 15 of them), we can make a 3x3 panels plot, with time periods in columns and depth layers in rows. Each panel will contain parameter "a" in x axis, parameter "b" in y axis, and the analyzed variable as z (image plot, color scale).
To save time: since we are likely to use this kind of figure for publications, you can adjust resolution to 300 dpi and plot width to 16 cm (height can be similar to width). Font size to 8-10. Regarding color scales, better avoid rainbow. Something like viridis would do the job.