Question about the ET

[Chuyang Liu]

Hi there,

I had an integrated hydrology model to simulate a snow-dominated watershed. After using the ATS converter to convert the XML file from version 1.4.2 to 1.5. I noticed that simulated ET reduced a lot, especially in the early and later months of the year. After checking the water balance, I noticed that the major source of difference in ET is from the transpiration.  As shown below, transportation using ATS 1.5 also has lower values in early and later months of the year compared to the earlier version (ATS 1.4.2). Based on the MODIS ET,  simulated ET from ATS 1.4.2. has a better alignment than the one from ATS 1.5. I would appreciate any hint you can give me.

Best,

Chuyang

[Haoyuan YU]

Hi there,

 I’m experiencing a similar issue to Chuyang’s. My ATS model (master branch) is also underestimating ET, which leads to an overestimation of discharge and surface water content. One specific issue in my case is that snow evaporation remains zero at all times.  

Best,

Haoyuan

[Wyatt Tatge]

Hi,

I also had a similar issue where I was underestimating ET in the fall and spring, and I adjusted the dessicated zone thickness to make it shallower. It seemed like in my model, I had cells that were reasonably saturated (60-80% saturated range) but were not contributing to ET. The dessicated zone thickness parameter, to my understanding, controls when cells turn on and off for providing surface/bare soil ET. I would look at some visualization files and see if some of the cells are not contributing but should be, and test changing the dessicated zone thickness.

Wyatt

[Bo Gao]

To my understanding, the output transpiration is controlled by (1) canopy potential transpiration and (2) transpiration downregulation. I remembered from earlier versions of ATS to ATS-1.5, there are multiple changes in potential transpiration, like the items in canopy radiation balance. For the downregulation, there are two options now, one is limited by root depth and the other is limited by water availability (Possibly the latter was added in ATS-1.5?). But the "dessicated zone thickness" will affect the calculation of soil resistance if using the Sakagucki-Zeng model option, and thus be used as a downregulation for surface evaporation from potential surface evaporation. I am not clear how many changes have been made to transpiration. Ethan or Saubhagya may have a detailed explanation for this. 

Best,

Bo Gao

--
You received this message because you are subscribed to the Google Groups "Amanzi-ATS Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ats-users/edaabfdb-980c-4d7e-9a36-756ef3e63087n%40googlegroups.com.

[Pin Shuai]

I am also getting zero snow evaporation using ATS v1.5, specifically ATS version 1.5.2_c4dfd539. 

Has there been any fix to this?

Thanks,

Pin

[Amanzi-ATS Users]

I don't know why people are seeing zero snow evaporation, other than having people take a closer look at the interplay of the various Priestley-Taylor terms and energy/radiation calculations.

ats-demos/05_ecohydrology/priestley_taylor_canopy_evapotranspiration_relperm_trf.xml

still shows that snow-evaporation is working as expected in 1.5.2.  See the notebook: water_balance_pt_canopy_snow_surf.ipynb  which plots this for runs of that test.

The models DID change from 1.4 to 1.5, and that did result in a reduction of transpiration, which was due to changes in how the temperature of the ground was calculated (there was a common bug in many input files around this).  In particular, the Priestley-Taylor model is sensitive to a skin temperature used to compute a ground conduction term.  In 1.5, this is determined by the air temperature - 3 degrees, which was selected after a review of the remotely sensed literature of snow temperatures.  This may be inappropriate for your area, and likely is a reasonably strong control on this model.  Also, relative humidity/vapor pressure should be a strong control, as would LAI (if LAI is large in winter, the canopy blocks radiation to the snow, so there is not much available for evaporation).

The demo appears to be working correctly, so there isn't much I can do on my end.  First, I would recommend careful comparison of your input files to the demo -- there were changes to the input file in addition to the code, and the input converter for 1.5 has an option (--transpiration?  run xml-1.4-1.5.py --help to see for sure) that does more changes for this model (changes the downregulation from ELM's downregulation to a model that is more robust for van Genuchten curves).  If someone is interested in doing more careful sensitivity studies, I'd be happy to talk through with them what parameters are likely to be important.  I'm also not sure how well these evaluators are documented, but I'd be happy to talk through the details if people are unsure of the physics here.

Ethan

[Amanzi-ATS Users]

Two additional things after looking at the release notes:

1. Please don't compare to 1.4.0 - there was a bug in that code's ET.  Make sure you are comparing to 1.4.1, which fixed the bug.

2. One way to explore the differences between 1.4.1 and 1.5 ET are in input files defining how surface-temperature and snow-temperature are defined.  These would appear in the input files, but not in the code.  So looking at the different xml files between those two releases of the demos would hint at why the changes are happening.

Ethan

[Chuyang Liu]

Wyatt and Bo - Many thanks for the hint. I tested changing the dessicated zone thickness and did notice the evaporation change. Unfortunately, the transportation part is still the same.

Ethan - I will try to take a closer look at the ET based on your suggestions.

Best,

Chuyang

[Haoyuan YU]

Hi all,

I wanted to share some findings that might explain why my ATS models were showing zero values for snow evaporation, and also some insight on the dessicated zone thickness parameter.

1. Snow Evaporation Issue

The problem lies in the "surface temperature key" setting within the <ParameterList name="snow-evaporation" type="ParameterList">. It should be set to "surface-temperature" instead of "snow-temperature".

The current ATS workflow input file template uses "snow-temperature" by default, which leads to zero values in snow evaporation. This happens because the Priestley-Taylor equation receives a snow temperature of 0°C when the expected snow temperature (defined as surface temperature - 3°C) is greater than 0, effectively shutting down snow evaporation.

2. Dessicated zone thickness [m] Parameter

After running watershed-workflow, I noticed that the dessicated zone thickness parameter disappears from the input file and needs to be manually added under "WRM parameters". Although it's a bit counterintuitive since dessicated zone thickness in WRM parameters are assigned per soil type (cell-based) rather than land use type (face-based), it seems that only the value from the uppermost soil layer is used in calculating soil resistance, which influences surface evaporation.

If no value is set for the top soil layer, ATS defaults to 0.1 m, which can result in underestimated surface evaporation. So if anyone finds that surface evaporation seems too low, it might help to reduce the dessicated zone thickness value in the WRM section for the surface soil layer.

Hope this helps clarify things!

Best,
Haoyuan

[Bo Gao]

Thanks Haoyuan for the investigation. Yeah, surface-temperature shall be used instead of snow-temperature for the first point. For dessicated zone thickness, I'd like to add something.

1) There are two soil resistance model options available in ATS since v1.5, one is "Sakagucki-Zeng" and the other is "Sellers". "Sakagucki-Zeng" is the default model because it has been used in ATS for a long time, while "Sellers" was newly added to v1.5. Dessicated zone thickness is a parameter required for "Sakagucki-Zeng".

2) Before ats-1.5, the dessicated zone thickness is determined by land cover type, which means one land cover corresponds to one dessicated zone thickness. However, dessicated zone thickness is a parameter that helps to hold water to avoid evaporation. Theoretically, it should be somehow related to unsaturated zone thickness, though it was considered as an empirical value in the model. A larger value should somehow mean a thicker unsaturated soil layer, higher soil resistance, and less evaporation. So from this point, it should be soil-type determined instead of land cover determined. So we moved dessicated zone thickness out of the land cover module but added it to the soil parameter - related part. The default value of dessicated zone thickness is 0.1m. This is from a previous study in which 0.1m gave a better match (I am not sure whether it is a better match for evaporation or discharge), but it may not work well universally. In some cases, you may find a larger soil evaporation when you use the default value and you can increase it, but I think there is not much difference if just increase a little (for example from 0.1 to 0.2) if the gas saturation is not very high, because soil resistance increases with this thickness linearly, unless it is increased to a certain value. It seems there is not a theoretically suggested range for this parameter. Another thing that may need to clarify is that we used the saturation of the first cells layer to calculate soil resistance. 

3) The other soil resistance option ''Sellers" just accepts saturation as input but no other external parameters. This model tends to result in a higher soil resistance (less evaporation) than "Sakagucki-Zeng" at the same saturation condition.

4) Since both are empirical models, I can not say which one is better, but you can have some experiments based on your case.

Best,

Bo Gao

To view this discussion visit https://groups.google.com/d/msgid/ats-users/06f9e5b9-db0a-4457-8bea-05748609ca0dn%40googlegroups.com.

[Coon, Ethan]

Thanks for jumping in on the dessicated zone part, Bo.  Note that the input spec converter cannot move this parameter for you – there is no 1-1 map of land cover to soil properties, so we can’t know how to move this for you.  This conversation has helped me realize that the dessicated zone thickness parameter may not appear in any input spec, so it may not get auto-generated in ats_input_spec.  I’ll see if I can get that fixed for 1.6.

 

As always, we try to have the input spec converter to an exact remapping of the run, but we can’t guarantee this because it isn’t always possible.  It is always incumbent on the user to diff the before and after files, and try to understand the difference.  If you don’t understand the difference, please ask!  I’m trying to get better about having release notes highlight these types of changes in particular.

 

 

I do want to comment on the surface vs snow temperature part, because this is both confusing and, in my opinion, a place that still needs work.  This may get long…

 

The snow-melt evaluator for non-Arctic runs has always been an empirical model based on PRMS, where the snow melt rate is an empirical “degree-day” approach, which says that the snow melt rate is proportional to the difference between the air temperature and the snow temperature.  The snow temperature cannot go above 0 C (it would be water).  Typically, the snow temperature is assumed to be the max of 0 and the air temperature minus some constant (e.g. 3 degrees).  So effectively, the snow started to melt when the air temp hit 3 C, and the higher above 3 C, the faster the melting.  Prior to 1.5, this model did not explicitly use a variable called “snow temperature” – it just took the air temperature, subtracted 3, and checked if this was above 0.  In 1.5, I added the “snow temperature” variable explicitly to let users change the model used to compute this value (particularly I’d been playing with using an energy balance in non-Arctic simulations).

 

Now comes the confusing part.  This led me to realize that we actually have two different, inconsistent snow temperatures.  One used for melting (which was implied before 1.5, but still there) and one used for Priestley-Taylor of snow evaporation, which has always used “surface-temperature,” (which is almost always defined as yesterday’s air temperature, this also comes from PRMS).  So (while 1.5 was still in development) I did what I thought was the natural thing and used “snow-temperature” (air temp – 3) in the Priestley Taylor model.  I found exactly what Haoyuan and Pin found – this doesn’t work very well (though I admit I didn’t remember exactly how it didn’t work, or I’d have had you check this!).  So prior to releasing 1.5, I changed it back to having the PT snow model use “surface-temperature” again (see the demos:  https://github.com/amanzi/ats-demos/blob/master/05_ecohydrology/priestley_taylor_canopy_evapotranspiration_relperm_trf.xml#L428).

 

I don’t really like this inconsistency, and I agree that it is confusing to use “surface-temperature” for PT snow evap when we have a “snow-temperature”. 

 

If you find places that use “snow-temperature” in the PT snow evap, please let me know and I’ll fix them.  I know that the demos are correct.  I know that the version 1.4 template files in exasheds-campaign2 (which is not supported at this point) are correct.  The intent is to have people stop using those template files and use the demos as templates, but this isn’t all neat and clean yet, so I’m not sure what people are using for templates in version 1.5.

 

Also, I’d love to fix the inconsistency!  To do this, the best thing would probably be to get away from Priestley-Taylor and instead use a surface energy balance, which would also then get rid of empirical melting (energy available for melting would be computed as the remainder of energy in the energy balance).  Effectively this aims to bridge the gap between the Arctic model, which is much better on this, and the temperate model.  “simple_energy” in ats_demos was a first cut at this – it took away the freeze-thaw parts of the Arctic model, but still solves an energy equation to get the ground temperature.  I’d like to also provide the choice of letting the user provide an empirical way of computing ground conduction rates, then solve for “surface” and “snow” temperatures in an energy balance calculation.  I suspect this would 1, fix the inconsistency, 2, give a more mechanistic model of surface energy balance terms, and 3, improve evaporation, snow sublimation, and total evapotranspiration predictions.  I just don’t have time to do this right now – if anyone wants to work on it I’d love to help them do so!

 

TL;DR – Haoyuan is exactly right, use “surface-temperature” in your Priestley-Taylor snow evaporation models!

 

 

Ethan

 

 

 

 

 

 

Hi there,

 

I had an integrated hydrology model to simulate a snow-dominated watershed. After using the ATS converter to convert the XML file from version 1.4.2 to 1.5. I noticed that simulated ET reduced a lot, especially in the early and later months of the year. After checking the water balance, I noticed that the major source of difference in ET is from the transpiration.  As shown below, transportation using ATS 1.5 also has lower values in early and later months of the year compared to the earlier version (ATS 1.4.2). Based on the MODIS ET,  simulated ET from ATS 1.4.2. has a better alignment than the one from ATS 1.5. I would appreciate any hint you can give me.

 

Best,

Chuyang

--
You received this message because you are subscribed to the Google Groups "Amanzi-ATS Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ats-users/edaabfdb-980c-4d7e-9a36-756ef3e63087n%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Amanzi-ATS Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ats-users/06f9e5b9-db0a-4457-8bea-05748609ca0dn%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Amanzi-ATS Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-users+...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/ats-users/CAN8t3f%2BsWvH9-bJ71ahK7fz-1d3iUeZatH%3D0nobd%2BhpW65eCwg%40mail.gmail.com.

[Coon, Ethan]

> The current ATS workflow input file template uses "snow-temperature" by default, which leads to zero values in snow evaporation.

 

Ah ha, now I see what you mean Haoyuan:

https://github.com/environmental-modeling-workflows/watershed-workflow/blob/master/examples/Coweeta/input_data/transient-template.xml#L421

 

Yes, you are correct, this is a remnant of my pre-ATS-1.5 experimenting, and is incorrect.  I’ll fix this now.

 

Ethan

 

 

 

 

 

From: ats-...@googlegroups.com <ats-...@googlegroups.com> on behalf of Haoyuan YU <haoyuan...@gmail.com>
Date: Thursday, April 10, 2025 at 1:20 AM
To: Amanzi-ATS Users <ats-...@googlegroups.com>
Subject: [EXTERNAL] Re: Question about the ET

Hi there,

 

I had an integrated hydrology model to simulate a snow-dominated watershed. After using the ATS converter to convert the XML file from version 1.4.2 to 1.5. I noticed that simulated ET reduced a lot, especially in the early and later months of the year. After checking the water balance, I noticed that the major source of difference in ET is from the transpiration.  As shown below, transportation using ATS 1.5 also has lower values in early and later months of the year compared to the earlier version (ATS 1.4.2). Based on the MODIS ET,  simulated ET from ATS 1.4.2. has a better alignment than the one from ATS 1.5. I would appreciate any hint you can give me.

 

Best,

Chuyang

--
You received this message because you are subscribed to the Google Groups "Amanzi-ATS Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ats-users/edaabfdb-980c-4d7e-9a36-756ef3e63087n%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Amanzi-ATS Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-users+...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/ats-users/06f9e5b9-db0a-4457-8bea-05748609ca0dn%40googlegroups.com.

[Pin Shuai]

Thanks Ethan. I tried to follow this input file to modify my old ones:

https://github.com/amanzi/ats-demos/blob/master/05_ecohydrology/priestley_taylor_canopy_evapotranspiration_relperm_trf.xml

However, the updated model was running really slow (<100 sec per cycle) for my watershed model. I compared the differences between priestley_taylor_canopy_evapotranspiration.xml and priestley_taylor_canopy_evapotranspiration_relper. They used different transpiration models: rooting depth vs relative permeability, and capillary pressure at fully closed/open stomata vs maximum xylem capillary pressure. 

My question is: What are the main differences between those two transpiration models? which input file should I use for general purpose ecohydrology model runs? 

Note, the priestley_taylor_canopy_evapotranspiration.xml still uses "snow-temperature" in snow evaporation: https://github.com/amanzi/ats-demos/blob/2ef0d582aa5339bb40fd6ef5ca2bccfc77a14026/05_ecohydrology/priestley_taylor_canopy_evapotranspiration.xml#L427

-Pin