Salinity diffusion in a hypersaline estuary

[user_17855]

Hello,

I am a PhD student and am working with TELEMAC-2D on salinity in the Casamance estuary (Senegal), which is a dry-season hypersaline inverse estuary. During the dry season, upstream discharge is almost zero, and evaporation plays an important role in salt concentration.

I am currently trying to understand the diffusion/spreading of salinity in this context, but I observe very little sensitivity to the tracer diffusion coefficient. I tested several values (0.01, 0.1, 1, 5, 10), and the salinity patterns remain quite similar overall, with only small smoothing differences.

So I would like to ask:

Is this weak sensitivity to the horizontal tracer diffusion coefficient expected in such a dry-season, quasi-stagnant hypersaline estuary?

Could salinity be controlled mainly by tidal advection, boundary conditions, initial salinity field, and evaporation rather than by diffusion itself?

In TELEMAC-2D, which parameters would you recommend testing first for this type of estuary: diffusion coefficient, initial salinity field, boundary salinity, friction, wetting/drying options, or density effects?

Since evaporation is important in this system, would you recommend activating density effects at this stage, or first calibrating salinity without them?

For reference, I attach:

my .cas file,

a BlueKenue view of map from the simulation,

a dry-season salinity mesures.

Salinity_casamance_test-4.cas (5.47 KB)

[user_438]

Hello [name removed],

My experience of modelling such tropical areas is quite low, sorry. In particular, modelling salinity with TELEMAC-2D is not very familiar to me.

Anyway, I have a few questions about the area:

• what is the water depth variation during dry or wet seasons along the domain?

• is the salinity homogeneous along the vertical or do you have significant variations/stratifications?

• how big is evaporation (be careful with the expected unit)?

Reading your steering file, a few remarks:

• use smaller values for ACCURACY FOR DIFFUSION OF TRACERS (e.g. default value 1.E-8 since release 9.0 or even smaller if possible, e.g. 1.E-15),

• NERD for the advection of tracers is quite diffusive, you can try LIPS (SCHEME FOR ADVECTION OF TRACERS = 5 + SCHEME OPTION FOR ADVECTION OF TRACERS = 4),

• OPTION FOR SOFT BOUNDARIES: I do not have feedback for the combo of this option and tracers,

• you can try to have look of the influence of OPTION FOR LIQUID BOUNDARIES = 2 (Thompson) for salinity,

• initial salinity distribution along the river could have influence,

• perhaps coupling with WAQTEL-THERMIC could improve your results.

If you can find someone who has knowledge of salinity distribution along the river and over time (dry/wet season), it may be a help for a better modelling.

Perhaps you should switch to TELEMAC-3D non-hydrostatic version (perhaps coupled with WAQTEL-THERMIC).

Hope this helps,

[name removed]

[user_17855]

[name removed],

Thank you very much for your valuable feedback. Allow me to provide some context about the study area and the progress made since my initial post.

• Study area — Physical characteristics

The Casamance estuary (southern Senegal) is an inverse estuary, meaning that salinity increases from the mouth inland during the dry season due to evaporation exceeding freshwater inputs. Water depth is highly variable: from 0.1 m on sandbanks, mudflats and banks, to 26 m near the mouth and in the vicinity of the port of Ziguinchor (63 km upstream). Despite this bathymetric range, the estuary is reported to be vertically well-mixed according to literature from the 1980s and 1990s ([name removed], 1970; [name removed], 1985; Bassel, 1992), due to strong tidal mixing.

I will use 3D modelling maybe after but now I will advance with 2D.

Regarding evaporation, it varies both spatially and seasonally. Based on long-term station data from ANACIM (1969–2021):

• [name removed] (coastal): 2.89 mm/day on average over March–May

• Ziguinchor (83 km upstream): 3.88 mm/day over March–May

• Kolda (300 km inland): 8.44 mm/day over March–May

After conducting several tests, I noticed that the model is highly sensitive to evaporation when I add more in the obs. However, I’m not sure how long it will take me to get the model up and running (1 or 3 years?) in order to reproduce a good model

This spatial gradient reflects the increasing continentality inland. ERA5 reanalysis underestimates evaporation over the estuarine section (2.13 mm/day vs. 3.88 mm/day at Ziguinchor for March–May), so a uniform correction of −2.0 mm/day has been applied via the keyword RAIN OR EVAPORATION IN MM PER DAY.

[user_17855]

-Progress since the initial post

Since posting, the model has evolved considerably:

1.WAQTEL-THERMIC coupling: is already implemented, using ERA5 fields (wind, air temperature, relative humidity, atmospheric pressure, cloud cover, and downwelling shortwave radiation) interpolated onto all mesh nodes at each time step.

2.Salinity initialisation is performed using the USER_CONDIN_H subroutine, which imposes a spatially distributed initial salinity profile based on field observations along the main channel.

3. OPTION FOR LIQUID BOUNDARIES = 2 (Thompson method) is already active.

4. Field data: I have conducted 4 seasonal field campaigns (end of dry season and end of wet season between 2024 and 2025) in the estuary, measuring the longitudinal salinity profile from the mouth to 205 km upstream using a little boat (pirogue). [name removed] were taken at 4–5 km intervals along the main channel. These data provide validation points for both the dry season peak and the wet season freshwater recovery.

5. Historical context: I have presented an oral communication on the long-term evolution of salinity in the Casamance estuary, based on archived datasets spanning from 1966 ([name removed], 1970) to 2023 (own measurements), covering the major drought period (1968–1995) and the partial pluviometric recovery since the 2000s.

–Regarding your suggestions

• I will test the LIPS scheme (SCHEME FOR ADVECTION OF TRACERS = 5, SCHEME OPTION = 4) as a replacement for NERD, which I agree is too diffusive for sharp haline fronts.

• ACCURACY FOR DIFFUSION OF TRACERS will be reduced to 1.E-15 as recommended.

I’m attaching my steering files for your reference

Thank you again for your time and expertise.

mesh_salinite_1991-2.cas (6.21 KB)
waqtel_thermic_s-2.cas (1.78 KB)

[user_438]

Hello,

A few comments from your last TELEMAC-2D steering file:

• why using OPTION FOR THE TREATMENT OF TIDAL FLATS = 3 rather than more classical option 1?

• why using a coarser SOLVER ACCURACY than previously? 1.E-4 may be quite coarse (compared to 1.E-6 or even 1.E-8).

About a previous remark of mine, I would try to see the influence of OPTION FOR LIQUID BOUNDARIES = 1 (rather than 2 = [name removed]) and see if the results are changed/improved/worse.

A few comments from your WAQTEL steering file: you have modified a few keywords which are only used with THERMIC module BUT in 3D (see helps in WAQTEL dictionary and/or WAQTEL user manual), so that your modifications are useless in 2D (e.g.: ATMOSPHERE-WATER EXCHANGE MODEL, LIGHTNESS OF THE SKY, SECCHI DEPTH, COEFFICIENT TO CALIBRATE THE ATMOSPHERE-WATER EXCHANGE MODEL).

If you want to calibrate the model, rather than COEFFICIENT TO CALIBRATE THE ATMOSPHERE-WATER EXCHANGE MODEL, use the 2 value keyword COEFFICIENTS OF AERATION FORMULA.

Hope this helps,

[name removed]