# Query regarding accuracy of the mixed layer depth diagnostic in ACCESS-CM2 **URL:** \ **Category:** Coupled Model\ **Tags:** help, cosima, community-help\ **Created:** [16 September 2026 04:14 UTC](https://forum.access-hive.org.au/t/query-regarding-accuracy-of-the-mixed-layer-depth-diagnostic-in-access-cm2/6996 "2026-09-16T04:14:42Z")\ **Posts on this page:** 5\ **Page:** 1
**Author:** ![hrsdawson](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.access-hive.org.au/hrsdawson/32/218_2.png) [@hrsdawson](https://forum.access-hive.org.au/u/hrsdawson)\ **Post date:** [16 September 2026 04:14 UTC](https://forum.access-hive.org.au/t/query-regarding-accuracy-of-the-mixed-layer-depth-diagnostic-in-access-cm2/6996/1 "2026-09-16T04:14:42Z")
Hi all, I’ve got a query about the mixed layer diagnostic in ACCESS-CM2. The model output I’m using (pre-industrial configuration) shows very deep mixed layers in the Ross Sea (left panel below). ![Screenshot 2026-09-16 at 2.11.39 pm](https://us1.discourse-cdn.com/flex020/uploads/access1/original/2X/2/2022b417cb5656db1adfcf17de222ee2b68022c8.png) However, the temperature, salinity, and density profiles from a location within this deep MLD region (the red marker) don’t support this: ![Screenshot 2026-09-16 at 2.09.35 pm](https://us1.discourse-cdn.com/flex020/uploads/access1/original/2X/4/46a358ee20c6df1622019f710735ef90e6101483.png) I’ve recalculated the MLD using a 0.03 g/kg density criterion (blue dashed line in figure above, middle map panel) and it produces vastly different estimates than the model `mld` diagnostic (solid blue line in figure above). The recalculated mixed layer depth looks correct to me. Is this a known issue? Is the model `mld` diagnostic incorrect, or is the monthly averaging of T and S hiding deep convection events that occur on shorter timescales? Any insights would be very appreciated! ---
**Author:** ![RhaegarZeng](https://avatars.discourse-cdn.com/v4/letter/r/ea666f/32.png) [@RhaegarZeng](https://forum.access-hive.org.au/u/RhaegarZeng)\ **Post date:** [17 September 2026 04:30 UTC](https://forum.access-hive.org.au/t/query-regarding-accuracy-of-the-mixed-layer-depth-diagnostic-in-access-cm2/6996/2 "2026-09-17T04:30:20Z")
Hi @hrsdawson thanks for raising this. We’re not yet sure whether this indicates an issue with the ACCESS-CM2 `mld` diagnostic or is related to differences in how the diagnostic is calculated/averaged. Katja has recently been looking at offline calculation of mixed layer depth from temperature and salinity for ACCESS-OM2 and ACCESS-OM3. In particular, she has developed notebooks that calculate MLD manually rather than relying on the model `mld` diagnostic, partly because the diagnostic is calculated differently in OM2 and OM3. The work is available here: [Add notebooks for manually calculated mixed layer depths by KZCurtin · Pull Request #131 · ACCESS-Community-Hub/access-om3-paper-1 · GitHub](https://github.com/ACCESS-Community-Hub/access-om3-paper-1/pull/131) ---
**Author:** ![hrsdawson](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.access-hive.org.au/hrsdawson/32/218_2.png) [@hrsdawson](https://forum.access-hive.org.au/u/hrsdawson)\ **Post date:** [17 September 2026 06:01 UTC](https://forum.access-hive.org.au/t/query-regarding-accuracy-of-the-mixed-layer-depth-diagnostic-in-access-cm2/6996/3 "2026-09-17T06:01:27Z")
Thanks @RhaegarZeng. I’ve checked out Katja’s notebook and the way I’ve calculated the mld above is essentially the same, minus the last interpolation step. Would your recommendation at this stage be to use a manually calculated MLD, rather than the model diagnostic? ---
**Author:** ![RhaegarZeng](https://avatars.discourse-cdn.com/v4/letter/r/ea666f/32.png) [@RhaegarZeng](https://forum.access-hive.org.au/u/RhaegarZeng)\ **Post date:** [17 September 2026 06:34 UTC](https://forum.access-hive.org.au/t/query-regarding-accuracy-of-the-mixed-layer-depth-diagnostic-in-access-cm2/6996/4 "2026-09-17T06:34:38Z")
Thanks @hrsdawson . I’m probably not the right person to make a recommendation on which MLD diagnostic you should use, as this is outside my area of expertise. Given that your calculation is essentially consistent with Katja’s approach, you could certainly try using that method for your analysis. Otherwise, it may be worth waiting to see if someone in the community with more expertise in MOM5/MLD diagnostics can provide a more definitive answer on the model diagnostic. ---
**Author:** ![Benoit](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.access-hive.org.au/benoit/32/3115_2.png) [@Benoit](https://forum.access-hive.org.au/u/Benoit)\ **Post date:** [23 September 2026 00:55 UTC](https://forum.access-hive.org.au/t/query-regarding-accuracy-of-the-mixed-layer-depth-diagnostic-in-access-cm2/6996/5 "2026-09-23T00:55:16Z")
Is it possible that the MOM5 MLD diagnostic is a maximum rather than an average or a snapshot? (I think T, S, and ρ are time-averaged, right?)