In 2008, I advocated for addition of seasonality to subcatchment and soils parameters. Several improvements were subsequently added to Build 5.0.014 (1/21/09) in support of this functionality. I confess that I've hardly used them since in the many models I have built or supervised over the intervening years. As one who feels that curve number methods are inferior to the more hydrologically rigorous Green-Ampt and Horton methodologies and given the limited use to which I've put those existing capabilities, it does not seem worthwhile to me to add seasonality to CN methods. In case it is of any use, I'm attaching an email chain with Lew Rossman and Wayne Huber that led to the 5.0.014 revisions.

From: Wayne Huber
To: Mitch Heineman, Lew Rossman
Sent: Friday, July 25, 2008

I agree with Lew's assessment that the soil moisture accounting scheme needs to be revised to logically connect it to ET and deeper percolation. This should probably be done by linking the infiltration routines with the SWMM groundwater routines, but exactly how, I don't know, although I have some ideas. It will need some literature research. I'd be delighted to have CDM
(or Christine Pomeroy) do this. Several seasonal parameters are needed in the model, including depression storage (or a way to simulate interception as a function of leaf canopy), ET coefficients, possibly roughness as a function of temperature, and infiltration parameters through some mechanism. I have no problem with a quick fix of the notion you suggest, i.e., monthly GA or Horton variation, since the rigorous approach will take some time.

The principal need for the rigorous approach is for continuous simulation, of course. For the many users who do event modeling, we can make it easy to continue with basically the same GA or Horton routines they use now.

From: Rossman.Lewis
Sent: Friday, July 25, 2008
To: Heineman, Mitchell, wayne.huber

If 1 involves just applying a monthly pattern factor to the GAML recovery constant (as per my suggestion), then I can make the needed changes to the engine and GUI by next week and send you an evaluation copy of the updated program for testing. If it produces satisfactory results for you, then we could include this as an optional feature in a future SWMM 5 update.

If you (i.e., CDM) would like to do more than this, then I would suggest creating a project that takes a more rigorous look at the entire soil moisture accounting (SMA) process associated with infiltration. Currently, SWMM has the infiltrated water just disappear during a storm and
uses an empirical recovery equation (either linear for GAML or 1st-order for Horton) to modify the "moisture state" of the soil during a dry period. Even when the SWMM groundwater component is included, the infiltration portion of it still depends on the empirical recovery equation. Other models, like HSPF and HEC-HMS, provide a more explicit soil moisture accounting process. As I understand it, evapotranspiration and percolation to GW zones are the main physical processes that allows the surface soil infiltration capacity in these models to "recover" during dry periods. Yes, I know that the SWMM GW module has these processes too, but they aren't used (or linked) with the surface infiltration sub-model. The seasonal effects on infiltration of a more rigorous SMA model would follow directly from measurable seasonal changes in ET and GW levels, rather than relying on empirical, seasonally adjusted "recovery" functions.

My 2 cents. I'd be interested to see what Wayne thinks.

From: Heineman, Mitchell
To Lewis Rossman, Wayne Huber
07/24/2008

Lew and Wayne - To modify SWMM's GAML to be seasonal do you think that:

1. we should just do it;
2. we should do a literature search and document what, if anything, researchers have found;
3. 2 should be supplemented by original field/lab work?

From: Rossman.Lewis
Sent: Friday, June 20, 2008
To: Heineman, Mitchell

Regarding the snow melt comment, WEPLOW is the residual depth parameter you refer to. Unfortunately, if you set it to 0 in the GUI, the program thinks that snow removal is turned off -- I need to fix this. If you want a bare surface, just set its value to a real small number.

Regarding GAML, I agree 100% with Wayne. It looks as though Mein pulled together the recovery portion of the model out of thin air. There is absolutely no supporting documentation for what he did. In a nutshell, here's how it currently works. Theta_D, the difference between current soil moisture content and saturated moisture content, is the state variable that gets tracked continually over time. During wet periods, it gets reduced as follows:

Theta_D --> Theta_D - f / Lu * Dt

where f is the G-A infiltration rate, Dt is the time step, and Lu is an "upper zone" depth that Mein defined as a function of hydraulic conductivity Ks. During dry periods, Theta_D increases as per:

Theta_D --> Theta_D + Kr * Theta_D_max * Dt

where Kr is Mein's recovery rate constant (also a function of Ks) and Theta_D_max is the maximum initial moisture deficit which is a user-supplied constant (typically saturated moisture content - wilting point). The standard G-A equations for infiltration rate "f" are based on the Theta_D value that exists at the start of the rain event. To separate one event from another, Mein defined a minimum inter-event dry period Tr that must occur before a new event (with a new Theta_D value) is begun. He made Tr = 0.06 / Kr (why? -- who knows.) The plot below shows how the three recovery parameters, Lu (solid line), 1/Kr (large dash line), and Tr (small dash line), are related to Ks:

Note that evaporation never enters the picture. To "seasonalize" GAML, one would need to apply a seasonal adjustment factor to one or more of these parameters (most likely just to Kr). A similar approach could be used for Horton, applying a seasonal pattern factor to its recovery parameter as well.

The other way to make infiltration more seasonal is via the groundwater model's ability to turn off infiltration whenever the upper zone becomes completely saturated. The GW model is already somewhat seasonally dependent since it does utilize monthly evaporation rates. However a more direct (and realistic) seasonality could be produced if the depth of the lower saturated zone were allowed to be adjusted (independently of infiltration from above) to reflect higher regional water table levels during certain times of the year.

Probably the easiest way (programming-wise) to introduce seasonality to runoff rates is to just apply a monthly pattern factor to the infiltration recovery rate parameter, since SWMM 5 already has the ability to process monthly patterns (for DWFs).

From: Heineman, Mitchell
To Lewis Rossman
06/20/2008

One thing that has bothered me for some time is SWMM's inability to differentiate runoff seasonally. I spoke with Wayne Huber recently about how Green-Ampt Mein-Larsen infiltration is formulated and how it might be modified to allow soils to stay wet in winter and dry rapidly in summer. Wayne told me that it has never sat right with him, as the Mein-Larsen infiltration recovery model does not have a strong physical basis. He was not aware of specific research that addressed the issue. It seems to me that GAML could be modified without much fuss to adjust the soil moisture recovery terms based on monthly evaporation rates. I don't know if there's any research to support this. I could see about initiating a CDM R&D project.

This has come to the fore recently as I've been involved in a debate for a client where another consultant proposed using HSPF to model I/I. CDM (and other consultants involved in the project) are advocating the RTK approach, possibly combined with the SWMM groundwater module to account for infiltration. I've been sitting on the fence between RTK and the non-linear
reservoir approach (which is getting little traction), but frustrated that HSPF does indeed seem to be doing a better job. HSPF simulates small early spring storms producing more inflow than larger summer storms, a phenomenon we've observed in numerous studies, but GAML cannot reproduce. One can use monthly RTK parameters to achieve this effect, but I prefer the hydrologic basis of the non-linear reservoir methods.

@MitchHeineman this is most helpful! Thank you for posting your deliberations.