Finite canopy radius in longwave multiple-scatter RTM

[ashiklom]

In the two-stream RTM, blackbody emissivity is scaled by the crown area index as follows:

https://github.com/EDmodel/ED2/blob/42a05a8a33c340f5aba37b5c9a33c50798c6d085/ED/src/dynamics/twostream_rad.f90#L145-L149

However, what looks like the same correction is not applied in the multiple-scatter RTM:

https://github.com/EDmodel/ED2/blob/42a05a8a33c340f5aba37b5c9a33c50798c6d085/ED/src/dynamics/multiple_scatter.f90#L184-L188

As a result, with finite canopy enabled, the two-stream RTM seems to produce hotter leaves than the multiple-scatter RTM.

Is this intentional or a bug?

[mdietze]

Bug. When the finite canopy RTM was developed the focus was on trying to correct a persistent demographic bias (canopy trees were growing too fast, understory too slow, relative to observations) so there was an immediate focus on PAR and GPP. Just never had time to apply the same changes to the LW code, as the addition of the emission term made that code more complicated to deal with (i.e. it involved more than just cut-and-pasting what worked for SW)

[mpaiao]

It may be a bug, but as far as I remember I had put a barrier to prevent using crown_mod =1 with icanrad=2 in ED/src/io/ed_opspec.F90. When crown_mod=0, then cai is always 1, so this would not be an issue.

The finite crown was never fully implemented in ED-2 (not even in the old two-stream). Back then, it was only applied to direct radiation, which also creates physical inconsistencies: crowns can't be finite in direct light but infinite in diffuse light. The equations for direct and diffuse radiation are actually the same, except that in diffuse light we integrate all possible solid angles.

I tried to make the corrections in the new two-stream model, but they are not thoroughly tested. For emissions, we may need to drop cai because LW emission happens from all surfaces.

[ashiklom]

The current logic in ed_opspec.f90 makes it impossible to run with finite crown enabled (crown_mod = 1), because icanrad < 1 is illegal (per the first part) but icanrad != 0 & crown_mod = 1 is also illegal.

https://github.com/EDmodel/ED2/blob/42a05a8a33c340f5aba37b5c9a33c50798c6d085/ED/src/io/ed_opspec.F90#L2195-L2205

FWIW, in my simulations of two-stream RTM + finite canopy (after removing the block above) starting from bare ground, I get high leaf temperatures, higher respiration, and ultimately dramatically suppressed productivity.

[ashiklom]

This may also be a good thread to mention that EDR currently ignores the setting of crown_mod and assumes it is always on.

https://github.com/EDmodel/ED2/blob/42a05a8a33c340f5aba37b5c9a33c50798c6d085/EDR/src/dynamics/radiate_driver.f90#L609-L614

I strongly suggest we change that back, since that behavior is very counterintuitive.

[ashiklom]

P.S. Also wanted to quickly point out that I think (?) you guys have a typo in the Figure 4 of the GMD model description paper. At the start of line 3 of the caption, the second Q should have a down arrow, not a circle-dot, right?

[mpaiao]

@ashiklom Thanks for catching the typo! I will make sure to correct it during the proofreading.

[ashiklom]

For emissions, we may need to drop cai because LW emission happens from all surfaces.

I think this is the correct solution. At the very least, removing the cai term from the blackbody emissivity term in the two-stream longwave RTM gives much more reasonable results. In particular, when the cai term is included, you can get leaf temperatures that drop to around 2K in the winter, which is obviously wrong. Conversely, when this correction is removed, the leaf temperature fluctuations return to close to the default values.

In the below figure, "default" is without the finite crown model, "crown" is the current implementation (with black ~ f(cai, temp)), and "crown-alt" is the finite crown model without correcting blackbody emissivity for crown radius (black ~ f(temp)). These are monthly outputs from a bare-ground initial condition run in northern Michigan.