Reservoirs, ponds and wetlands

[Bill]

Hello,

I have several questions related to modeling lakes, ponds, wetlands,
and reservoirs in SWAT 2005 using ArcMap 9.3. I have been lurking in
the SWAT forums for some time and have found many good answers to some
of my SWAT questions in prior posts, but still have a few that remain
unanswered.

I am using SWAT to study a 1,140 km2 drinking water supply watershed
in southern Maine. It contains a large (121 km2) reservoir, and quite
a few large ponds (i.e. > 0.2 km2).

1. In which subbasin should I locate the reservoir? I found a
previous posting in the SWAT-user group that indicated the reservoir
should be placed in the farthest downstream reach on the main stem,
below all tributaries contributing to the reservoir.

I manually added an outlet at the dam & spillway for the reservoir.
If I follow the above method using the default subbasin discretization
settings, this results in a 121 km2 reservoir being added to a 6.3 km2
subbasin.

Should I manually create the stream network and delineate subbasins to
reflect upland areas that directly contribute runoff to the reservoir,
thus creating a subbasin large enough to contain the reservoir? Aside
from needing to modify the upper limit to SWAT reservoir surface area
and volume and so on, I saw no requirement in the SWAT documentation
that a reservoir cannot be larger in area than the subbasin in which
it is located. If I create a larger subbasin for the reservoir,
though, how would the model handle multiple reaches from multiple
upstream subbasins entering that large subbasin? Also, the dominant
land use within that large subbasin or multiple smaller subbasins
would be water, creating an HRU with water as a land use - which
should be avoided according to the documentation (why?). I'm
wondering what best practices and user experience dictate.

2. Other questions I have relate to modeling lakes/ponds and
wetlands.
a. Should water bodies be modeled as ponds or wetlands depending on
whether they intersect the main reach being modeled in a subbasin?
b. Should I manually delineate subbasins for larger lakes so that I
do not have a lake which physically overlaps several subbasins?
c. The documentation indicates that one pond and one wetland may be
included per subbasin. Could I aggregate multiple ponds/wetlands? Or
does my discretization scheme need to ensure there is only one of each
per subbasin? Is there some threshold size below which ponds/wetlands
don't need to be modeled?
d. I have a forestry, rather than a hydrology background, and am
wondering if someone can point to me to a relatively simple but
defensible method to calculate pond/lake volume. I have access to
some (limited) bathymetry data, but am wondering if anyone knows a
reasonable method.

Thank you,
Bill VanDoren

[Jim Almendinger]

Bill -

Your questions are right on the mark.  I have some answers, but not all.  

On 19 Aug 2010, at 11:08 AM, Bill wrote:

1.  In which subbasin should I locate the reservoir?  I found a
previous posting in the SWAT-user group that indicated the reservoir
should be placed in the farthest downstream reach on the main stem,
below all tributaries contributing to the reservoir.  

I manually added an outlet at the dam & spillway for the reservoir.

Right -- manually add an outlet point at your dam, to assure there is a subbasin corresponding to your reservoir.  

If I follow the above method using the default subbasin discretization
settings, this results in a 121 km2 reservoir being added to a 6.3 km2

subbasin...

...I saw no requirement in the SWAT documentation

that a reservoir cannot be larger in area than the subbasin in which
it is located.  If I create a larger subbasin for the reservoir,
though, how would the model handle multiple reaches from multiple
upstream subbasins entering that large subbasin?  

Right -- this is a problem, at least conceptually.  SWAT may not crash (I don't know -- it might), but in any case this configuration just doesn't make sense, as you know.  But it can be avoided.  The way to do this is to add outlet points at the point where each stream enters the reservoir, which will force ArcSWAT to delineate subbasins above each of those points.  Next, manually delete those linking-point outlets that were automatically located inside your reservoir, where ArcSWAT thinks those streams come together.  The end result should be (after subbasin delineation) that the reservoir sits in its own subbasin, larger than the water surface, and with the subbasins of all inlet streams terminating at the reservoir edge.  

     When I did all this, I created a point shapefile in ArcMap identifying all my lake outlets, and all the inlets to each of these lakes.  I exported a table with coordinates for those points as a table, and then I entered those points as "table-added outlets" during the delineation part of ArcSWAT's set-up.  (If there's a way of adding outlets directly from a shapefile I'd like to know... I'm still using ArcGIS 9.2 so I'm a bit out of date).  Finally I looked at each reservoir, and manually deleted the linking-outlets that ArcSWAT had created between the inlets and outlet for each reservoir.  

    I had 18 lakes in my model, and it took me quite a few tries before I finally was satisfied with my configuration.  

Also, the dominant
land use within that large subbasin or multiple smaller subbasins
would be water, creating an HRU with water as a land use - which
should be avoided according to the documentation (why?).  I'm
wondering what best practices and user experience dictate.

The problem of what happens when water land-use dominates your HRU is one I haven't fully gripped yet.  The "problem" is that SWAT does not calculate any water yield from the WATR land use.  It calculates a yield from all other HRUs in the subbasin, as though they cover 100% of the subbasin, and then calculates a total subbasin yield by area-weighting the yields from those HRUs.

     For example -- consider an HRU that is 10 km2, with 8 km2 of WATR, 1 km2 of forest, and 1 km2 of grassland.  SWAT would not calculate yield from the WATR area.  It would calculate yields from the forest and grassland -- and it would know that, of the HRUs it used, they composed 50% each of the water yield.  It would then calculate a subbasin yield as though 50% of the subbasin (5 km2) were forest and the other 50% (5 km2) were grassland.  

    If you've got smatterings of ponds amounting to a small (5%?  no rules here...) area of your HRU, this is a perfectly reasonable way of handling the situation.  Yield is dominated by those uplands contributing runoff, and total yield is incorrect only to the degree that yield from the small area of water would be different from the uplands.  

    What I haven't fully checked is what happens in exactly the case you have (and I have it, too), where a reservoir (and its WATR land use designation) cover well over half of the subbasin.  I think -- but I'm just not quite sure on this -- that SWAT does account for the area of the reservoir (and/or ponds and wetlands, if you have them), so it doesn't double-count areas.  

   I think the moral of the story is that if you've got substantial surface-water bodies (lakes, ponds, and wetlands), you really should be adding those features (reservoirs, ponds, and wetlands) into your SWAT model.  SWAT does not use the WATR land use to calculate much of anything (except evaporation, for some reason).  

2.  Other questions I have relate to modeling lakes/ponds and
wetlands.
a.  Should water bodies be modeled as ponds or wetlands depending on
whether they intersect the main reach being modeled in a subbasin?

It's a bit of a judgment call whether to model some areas as ponds, or whether to call them reservoirs.  Technically, all water bodies on the main channel are "reservoirs" in SWAT, but you cant include every little swamp or backwater.  In SWAT, reservoirs have a huge influence on flow (your hydrograph), so you don't want to ignore those lakes whose outflow would definitely influence flow.  

b.  Should I manually delineate subbasins for larger lakes so that I
do not have a lake which physically overlaps several subbasins?

Yes, as explained above -- control the subbasin delineation by adding and deleting the outlet points as needed.  

c.  The documentation indicates that one pond and one wetland may be
included per subbasin.  Could I aggregate multiple ponds/wetlands?  Or
does my discretization scheme need to ensure there is only one of each
per subbasin?  Is there some threshold size below which ponds/wetlands
don't need to be modeled?

Aggregate small ponds and wetlands into one conceptual pond and/or wetland in each subbasin.  I would base my overall subbasin delineation on where my main channels (perennial reaches) were, and where my big lakes (reservoirs) werr -- and then aggregate other water bodies based on which subbasin they occupy.  

    There are no rules about threshold of sizes....  all I can say is to try to come up with an objective rule.  

     In our case (Minnesota -- with glaciated topography similar enough to Maine's), I tend to think landscape depressions have a huge control on runoff and nonpoint-source pollution.  I definitely want to account for relatively "closed" depressions on the landscape that don't contribute much overland flow (i.e., they rarely if ever spill).  And, I've got big wetlands that my stream may meander through, that certainly modify stream hydraulics and trap some incoming sediment.  I think SWAT Ponds and Wetlands are important tools to address these issues.  

     In one project we used Ponds to model closed depressions -- open-water areas were aggregated and given arbitrarily deep depths and large infiltration rates to stop them from ever spilling (thereby trapping al sediment and phosphorus).  Wetlands were aggregated from areas of riparian wetland, that would trap sediment but not water or phosphorus. I think we made the wetlands 0.3 m deep.  Contributing areas to depressions and wetlands were hand-delineated from topo maps (from DRGs in ArcView).  

    In a second project, we used ArcHydro to identify all depressions and their drainage areas; we calculated runoff required to fill each depression (depression volume divided by drainage area).  For all depressions with a runoff-fill depth of 1 cm or larger, we modeled depressions whose drainage area intersected the main channel as Ponds, and those disjunct from the drainage network as Wetlands.  We used the areas and volumes as determined from ArcHydro.

d.  I have a forestry, rather than a hydrology background, and am
wondering if someone can point to me to a relatively simple but
defensible method to calculate pond/lake volume.  I have access to
some (limited) bathymetry data, but am wondering if anyone knows a
reasonable method.

Thank you,
Bill VanDoren

No defensible generalization is possible, I suppose.  Ignoring that concern for the moment:  

When in doubt, try to estimate a sensible value, even if it's arbitrary, and be honest that it's a guess.  We can get the areas and volumes of most of our lakes from our Dept  of Natural Resources.  Lakes too shallow to have any bathymetry on topo maps seem to be somewhere in the neighborhood of 1-1.5 m (3-5 ft).  The active depth in wetlands (the acrotelm) is, oh, about 0.5 m, maybe.  Emergent marshes, maybe 1 m?  Make a guesstimate of your maximum (emergency) volume and area -- say, 15-30% larger than normal volume and areas.  (I looked at the bathymetry of 11 lakes my last study area, and calculated an average change in volume of 36% and area of 26% for a 1-m change in lake level.  I decided a 0.5-m change was my "emergency" level.  Who knows how transferable these values are.)  Perhaps some limnologist studying the morphometry of Maine lakes can give you an average depth and volume for a known area.  

     I don't think Pond and Wetland morphology is going to be the "big error" in a model -- reservoirs have a much bigger influence and their location and morphometry are more critical, IMHO.  Though, Ponds really have about the same ability to control outflow...  

You can download the our project reports from our website, which covers these details, including how we set nearly every parameter we altered from default.  We started our modeling here without much instruction, so we may not model conventionally.  If contribution from seepage is an important contributor to baseflow (as I expect it could be in Maine, at least in outwash deposits), you might want to try our swat2005 code, which we've modified to allow for seepage from ponds, wetlands, and reservoirs to contribute to baseflow.  

Our web page is:

www.smm.org/scwrs/tapwaters/

and the project reports are near the bottom of the page.

Cheers,

-- Jim

Dr. James E. Almendinger, Senior Scientist

St. Croix Watershed Research Station

Science Museum of Minnesota

16910  152nd St. N

Marine on St. Croix, MN  55047

tel: 651-433-5953 X 19

fax: 651-433-5924

email: din...@smm.org

web: www.smm.org/SCWRS/

Experience a once-in-a-lifetime opportunity: see authentic 2,000-year-old manuscripts, including the earliest biblical writings. The Dead Sea Scrolls. Once forgotten. Now unforgettable. Visit www.smm.org/scrolls.

[Bill]

Jim,

Thanks for your prompt reply. I haven't had a chance to implement
your suggestions yet but they seem to make sense and most are in the
direction I've been leaning anways - using ArcHydro, for example, to
get a better handle on surface hydrology.

I had started out using SWAT 2009 but was experiencing problems on my
computer so switched to SWAT 2005, but I haven't seen a way to add
inlets/outlets from a shapefile - it seemed like it still had to be a
table with the XPR/YPR . . .etc. format.

Thanks again,
Bill

[Noel Forest]

1. I have 15 lakes and about 20 wetlands in the basin. I did the same
way suggested by Jim to take care of multiple streams entering a
downtream watershed. But, is there a way to check if all the outflows
from upstream watersheds are considered inflows to the downstream
watershed that contains the reservoir (i.e. sum of OUTFLOW-upstream =
INFLOW-downstream; is this a valid assumption?).

2. Method to compute max area and volume: I used landsat scenes to
delineate water bodies - did an unsupervised classification on the
subset of the scene that contain the lake/reservoir. I selected two
scenes, one during the dry period to get the min area and one during
the wet period to get the max area. For most of the lakes in my basin
the average depth or volume of the lakes were known. If you know the
average water depth/volume of the lake then you can use these two
areas to compute the max volume and max area (you will still have to
make a guess about the lake level fluctuations and I think Jim's
suggestions are good enough). Addiotnally, there are several studies
on hydropower potential of Maine, done mostly in late 1800s and early
1900s. These books must have information about the lakes and ponds of
Maine. Check the hydrology book by Daniel Mead.

3. You can model the lakes as unregulated reservoirs and calibrate the
lake level/volume fluctuations based on outflows and inflows. This
would also be a good validation check of your initial lake
calculations.

4. There is an option in the ArcSWAT preprocessor that can prevent
certain land cover types from being used in the HRU thresholding. I
usually do not include water, wetlands and other land cover types that
I consider important within certain subasins.

5. The inputs for wetlands and lakes are the same. In the case of
wetlands, there is E and ET taking place at the same time which SWAT
does not distinguish (correct me if I am wrong). One way correctly
account for lake and wetland E and ET is to consider the evaporation
coefficient as a calibration parameter instead of a constant values of
0.6. The evaporation from some of the lakes in my model are highly
underestimated.

[Jim Almendinger]

On 20 Aug 2010, at 09:01 AM, Noel Forest wrote:

1. I have 15 lakes and about 20 wetlands in the basin. I did the same
way suggested by Jim to take care of multiple streams entering a
downtream watershed. But, is there a way to check if all the outflows
from upstream watersheds are considered inflows to the downstream
watershed that contains the reservoir (i.e. sum of OUTFLOW-upstream =
INFLOW-downstream; is this a valid assumption?).

The inflow to any reach is the sum of outflow(s) of the next-upstream reach(es), plus what the local subbasin delivers to that reach as surface runoff, lateral flow, and groundwater flow.  

Outflow from a reach is inflow minus any losses to evaporation and seepage.  

Inflow to a reservoir = outflow from the local reach (i.e., the same reach as the subbasin where the reservoir is located).  This is true conceptually in SWAT even for those subbasins where there is no channelized reach in reality, because the reservoir occupies the entire flow path between inlet stream(s) and the lake outlet.  

5. The inputs for wetlands and lakes are the same. In the case of
wetlands, there is E and ET taking place at the same time which SWAT
does not distinguish (correct me if I am wrong). One way correctly
account for lake and wetland E and ET is to consider the evaporation
coefficient as a calibration parameter instead of a constant values of
0.6. The evaporation from some of the lakes in my model are highly
underestimated.

This option (to change the 0.6 or 0.7 coefficient) is not available now, is it?  I agree it could be very useful.