GENIE:Biogeochem:FAQs:2

What is tracer 'auditing' and do I really need it switched on?
When developing the model it is of fundamental importance that track is kept of the total tracer inventory of the system in light of internal mass transfer and any inputs (e.g., prescribed restoring or flux boundary conditions) or outputs (e.g., sedimentation) to ensure that spurious gain or loss of tracer mass does not occur, either through buglettes in the code or through numerical truncation. The tracer inventories of the ocean is therefore periodically re-calculated and compared to that predicted to have occurred on the basis of any net input (or output) that has occurred in the intervening time.

The simplest audit would be to calculate the tracer inventory at system start-up (before any movement of tracer mass), and at the very end (after the last movement of tracer mass has occurred). However, to integrate what might be relatively small net fluxes over an extended time period will invariably lead to the accumulation of numerical truncation errors. Instead, the audits are carried out periodically throughout the total model run-time, so that the integration of the net tracer fluxes into the ocean is carried out over much shorter intervals.

The entire audit procedure is as follows;


 * First, an initial inventory is calculated, achieved by summing the product of the concentration of each (selected) tracer with the mass of each each cell, across all wet cells.
 * The net spatially- and time-integrated transfer of tracer mass arising from all transfers across the external reservoir boundaries is calculated.
 * At a periodic pre-defined interval, the inventories are re-calculated. The difference between old and new inventories should be equal to the integrated net flux. If the relative difference between re-calculated inventory and estimated (on the basis of net flux) differs by more than a predefined threshold (set in biogem_config.par) then an error message is raised (and the model halted if requested).
 * The integrated net flux variable is then re-set to zero and the procedure repeated.

In short - if you are not modifying the code then you can take it on trust that the model distribution is free of (major) bugs and spurious gain or loss of tracers will not occur. It you don't trust it, then stiwch the auditing feature on.