EDmodel / ED2

Ecosystem Demography Model
79 stars 114 forks source link

Boundry heat flux in bottom soil layer #243

Closed larsonej closed 4 years ago

larsonej commented 6 years ago

Hi All, I've run across a problem in soil temperatures for Arctic simulations. The bottom soil layer seems to have a minimum temperature of 0 Celcius. I initialize the simulation with soil temps of about -20 C and within a year the whole column (0-2.5m) warms to zero or more. For reference, the mean air temperature at this site during this period is around 2 C. Is this a side effect of the hydrology boundary condition? There also does not seem to be a thermal conductivity term (or heat flux) in the heat equation from the lower boundary. I plan on adding one to maintain a permafrost boundary condition. Perhaps @mpaiao or @rgknox could comment on this. boundary_heat_flux

mdietze commented 6 years ago

While allowing heat to flux downward indefinitely might create permafrost, it seems like a bit of a kluge. When we were working on getting persistent permafrost we found the following was helpful:

mpaiao commented 6 years ago

@larsonej The constant temperature is because the soil is partially frozen (it won't change temperature until water is completely frozen or completely melted). In this case during the winter, heat flux between the bottommost layer and the layer should be always positive (as the layer is warmer), are you seeing this in the results?

@mdietze's suggestion should make some difference and the first two points should be easy to test (changing NZG/SLZ and using a more reflective soil color, you can change both of them directly in ED2IN). But it would be good to verify the heat fluxes to make sure there isn't any problem in the heat flux in the partially frozen soils.

larsonej commented 6 years ago

I agree that the constant temperature is probably due to partially frozen soil in the model. The question is why do 2 meters of hard frozen soil warm to the melting point by May in the first year of the simulation? A second run with a maximum depth of 5 meters took about two years to warm the whole column to freezing. For reference, measurements at this site have permafrost below 1 meter. I will try changing the albedo to match soil temperature observations, but that seems like a bit of a kluge. I think there should be another term (conduction?) in the energy budget at the bottom boundary layer that moves the bottommost layer temperature toward some constant value at depth.

mdietze commented 6 years ago

I wouldn't 'tune' the albedo to match the soil temperature, but Carl Davidson's thesis did find that the ED2 default snow albedo was not consistent with satellite observations (e.g. papers by Michael Loranty). I think you can defensibly move that parameter up to the empirical mean and then start sorting out the remaining issues

larsonej commented 6 years ago

Thanks for the recommendation. I'll start there and see how much that helps.

larsonej commented 6 years ago

Hi All, Modifying the soil and snow albedo had small effects on the 2m and deeper soil temperatures. However, the thermal conductivity used in the model was an order of magnitude too high for peat and about a factor of 2 too high for most mineral soils. Reducing the thermal conductivity and modeling a deeper soil seems to give much more realistic values. Erik

On Thu, Mar 29, 2018 at 3:30 PM, Michael Dietze notifications@github.com wrote:

I wouldn't 'tune' the albedo to match the soil temperature, but Carl Davidson's thesis did find that the ED2 default snow albedo was not consistent with satellite observations (e.g. papers by Michael Loranty). I think you can defensibly move that parameter up to the empirical mean and then start sorting out the remaining issues

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_EDmodel_ED2_issues_243-23issuecomment-2D377346687&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=uJsb92W0cPZc-WrMGDoJaYBpyra3HRxkJwvTFq8XI_s&m=XQad2xDL6-2mvGNbwfMdhn3Z3TsCxoLCRFXHwjfgBHs&s=RaoqPRS2nF2QocYj2ynQ0sFDFpW3qF_n2xi9RyU1MMc&e=, or mute the thread https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_Af-5F4bziQGCxZW7t9olzB7HiCTZ-2D5Vigfks5tjTZggaJpZM4S5XJE&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=uJsb92W0cPZc-WrMGDoJaYBpyra3HRxkJwvTFq8XI_s&m=XQad2xDL6-2mvGNbwfMdhn3Z3TsCxoLCRFXHwjfgBHs&s=yjrQRWRtXjHxYHLC1KJkLEErq85GseL4fzulGP-kehM&e= .

-- Erik Larson Postdoctoral Fellow Department of Organismic and Evolutionary Biology Harvard University

mpaiao commented 6 years ago

@larsonej I think the thermal conductivity for peat came directly from LEAF-3, and as far as I remember no one has changed since then. For the mineral soil conductivities, I used the numbers from Monteith and Unsworth (2008) book and applied a de Vries model to get the conductivities for different textures. Back then I couldn't find a better approach and I couldn't even find thermal conductivity for silt.

It looks like you have found information on the soil conductivity for different textures. Should we update the defaults in ED2?

larsonej commented 6 years ago

Currently, ED calculates thermal conductivity in two ways; 1) look up table based on specified soil type and 2) calculating thermal conductivity based on specified sand and clay fractions. So far, I've changed the lookup table values (ed_params) for soil type 12, peat. These values were way off (factor of 10 in thermal conductivity). The default values in this table for other soil types could probably be modified as well, although they are at least closer to observed estimates (factor of ~2 or less depending on soil moisture).

Method 2, calculating the thermal conductivity from composition will never work for peat, since peat is mostly organic material (conductivity of ~0.08) and this method requires the fractions of clay, sand, and silt to sum to 1. This could be fixed by adding a fourth constituent of organic material and requiring the sum of all 4 fractions be 1. I think this would be a useful future update and I'll put this on my to-do list.

On Tue, Apr 17, 2018 at 12:55 PM, Marcos Longo notifications@github.com wrote:

@larsonej https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_larsonej&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=uJsb92W0cPZc-WrMGDoJaYBpyra3HRxkJwvTFq8XI_s&m=gFVc7D3FWCWAVaU9-sg-65olvHch0GTzqnWSF8pzof0&s=zVUpfz1iWj7nVA2-8tq-l1dR0u4OZQLjc-Phr_YO1yk&e= I think the thermal conductivity for peat came directly from LEAF-3, and as far as I remember no one has changed since then. For the mineral soil conductivities, I used the numbers from Monteith and Unsworth (2008) book and applied a de Vries model to get the conductivities for different textures. Back then I couldn't find a better approach and I couldn't even find thermal conductivity for silt.

It looks like you have found information on the soil conductivity for different textures. Should we update the defaults in ED2?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_EDmodel_ED2_issues_243-23issuecomment-2D382065818&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=uJsb92W0cPZc-WrMGDoJaYBpyra3HRxkJwvTFq8XI_s&m=gFVc7D3FWCWAVaU9-sg-65olvHch0GTzqnWSF8pzof0&s=ll-OPoTFO4pQ3A-zPyEwoARlhSaBDLv_TkGeqNDDAF8&e=, or mute the thread https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_Af-5F4b5Jd2jUHD9EDagAViCwBtTuiaw7cks5tph5lgaJpZM4S5XJE&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=uJsb92W0cPZc-WrMGDoJaYBpyra3HRxkJwvTFq8XI_s&m=gFVc7D3FWCWAVaU9-sg-65olvHch0GTzqnWSF8pzof0&s=_kIGkI1oumFUb5PE9JdiDyp8ezpQvt_Dc3_pP5Kq6rA&e= .

-- Erik Larson Postdoctoral Fellow Department of Organismic and Evolutionary Biology Harvard University