Context
We aim to simulate the spatio-temporal stream temperature of a river reach using TELEMAC-2D coupled with [name removed] (THERMIC). The hydraulics are calibrated. Solar radiation was reduced to account for site-specific shading, and additional calibration parameters (e.g., coefficient for calibrating surface water radiation) were tuned so that simulated and observed temperature amplitudes agree well at the logger locations.
Problem.
There is a systematic time shift between simulated and observed temperature peaks. The lag increases with distance from the inflow boundary; by the downstream end the simulated signal is ~3 h late. Examples (inflow-near, mid-reach, outflow-near) are all from loggers in the main channel (see attachments).
What we have tried (and outcomes).
(1) Run without WAQTEL coupling
→ No change. Suggests the issue is not the heat-exchange module, as the heat fluxes at the air–water-interface primarily affect heat input/output and the source term, not the transport timing).
(2) Increase advective transport via higher velocities ([name removed] kst)
-
Hydraulics already calibrated: with kst = 35, depth-averaged velocity profiles agree well with three ADCP transects.
-
kst = 40 → negligible change in velocity and tracer advection.
-
kst = 45 → very slight shift of simulated temperature curves toward observations (still insufficient).
(3) Increase tracer diffusion coefficient
→ No significant effect on timing.
(4) Change numerical scheme for tracer advection–diffusion
*Baseline: NERD (scheme #14).
*Scheme #1 (Method of Characteristics) reduces the time lag but yields unrealistic amplitudes and shape of the time series.
Questions to the community:
- What else could we test to reduce the observed time-lag, e.g. schemes for advection, turbulence model, etc…?
Hi [name removed],
Clearly from your tests this appears to be an issue more on the numerical side, specifically on advection. Have you tried to run other schemes for advection ?
There is a strong sensitivity to time steps and CFL condition in those schemes, so maybe try reducing the TIME STEP parameter. You may also want to have a look at the keyword MASS-LUMPING ON TRACERS, full lumping (1) is very diffusive, so if you have set it to a non zero value you may want to decrease it.
Since your velocities are already well calibrated, it may also be interesting if you have data for these to replace the temperature by a passive scalar like salinity and see if you reproduce this issue. That would help to see if this is purely a transport issue.
Otherwise, this is a bit of an obvious tip but double check the PRESCRIBED TRACERS VALUES or the conditions related to the diffusion of tracers.
Hopefully this helps, curious to hear what you find.
[name removed]
Hello,
Can you tell which release you use?
Can you upload the 2 steering files you use (TELEMAC-2D + WAQTEL) + potential Fortran file?
Can you show the extent of your model and tell where it is located?
[name removed]
Hello,
thank you very much [name removed] and [name removed].
In response to [name removed] last message:
-
I am working with Telemac v8p4.
-
I have attached the two steering files from one of the example simulations.
-
The study area is a 3–4 km river reach of the Lower Lech in Germany, located between [name removed] and Ellgau, just north of Augsburg (see attachment).
Lech_FA1_winter25_therm_rad_reduced04.cas (8.88 KB)
waqtel_thermic-2.cas (2.77 KB)
I look forward to your insights.
[name removed]
Hello [name removed],
Sorry for the delay.
There is an issue with the choice DENSITY EFFECTS = YES. As written in the TELEMAC-2D manuals (reference + user), if using this choice, the 1st tracer has to be salinity but you do not mention it in the steering file, this is quite suspicious (you can try with default choice DENSITY EFFECTS = NO, but also = YES + adding salinity as 1st tracer).
You can also try to use default choice for LAW OF FRICTION ON LATERAL BOUNDARIES (= 0).
Anyway, it is advised to use the latest official release (currently 9.1.0) as bugs may have been fixed.
Hope this helps,
[name removed]
