tetra tech, inc

uilding, Suite 105

l Hill

4409

riangle Park, NC 27709

19) 485

) 485

k Cr

ll Schindler (MPCA)

Medicine River Detailed Model

scribes the

Medicine River, and other minor tributaries of the Minnesota River

020004

ough the center of this HUC8; however, it could readily be linked to

sota River.

ently labeled as

l Actions implemented to account for sediment transport through tile

esent tillage and bluff collapse)

d must be applied to a

at the

ration of less than 2 percent at all analyzed locations in the

erefore,

s adequate for most anal

sing a text editor (we suggest the shareware TextPad program) or

has advantages for different situations. The model may be run using

ote the following:

d be updated by copying the September 2009 version of hass

he Windows

i

Cape Fear

3200 Chap

P.O. Box

Research

Telephone: (

Telefax: (91

Cory Netland (Ha , 201

Hafiz Munir, Darr

Hawk Creek/Yello

This memorandum transmits and d t HUC

Hawk Creek, Yello Minnesota River

0 ulate the p

of the Minnesota River that runs th dily be linked to

scale model of the Minn

The model exists in two forms, cur ullSA

Speci port through tile

(in addition to those that rep ons

sediment transport are complex an out 15 minutes to

, a ations.

Experiments with the model show t y and

ge in average pollutant concen in the

T pecial Actions

for tile drain transport of sediment s.

The model uci files may be edited program) or

using the WinHSPF program. Eac may be run using

Please

The WinHSPF engine sho n of hass

provided) into greater number of

cannot be run rations.

2

dig

Hawk Creek, Yellow Medicine River, and other minor tributaries of th

The detailed model does not si

of the Minnesota River that runs through the center of this HUC8; however, it could re

Special Actions implemented to account for sediment tran

. These Special Act

, increasing run time from a

number of HRU combi

se Special Actions have no effect on hydrolo

ge in average pollutant concentration of less than 2 percent at all analyzed location

it is recommended that the version without

ysis purposes in these basi

The model uci files may be edited using a text editor (we suggest the shareware TextPa

using the WinHSPF program. Each has advantages for different situations. The model

The WinHSPF engine should be updated by copying the September 2009 versi

System32 directory. This allows for a

th the DOS version of HSPF due to number of op

TETRA TTECH, INC.

Cape Fear B Building, Suite 105

3200 Chapeel Hill-Nelson Hwy.

P.O. Box 1 14409

Research T Triangle Park, NC 27709

Telephone: (9 919) 485-8278

Telefax: (919 9) 485-8280

MEMORANDUM

To: Cory Netland (Haw wk Creek Watershed Project) Date: June 25 5, 2011

Hafiz Munir, Darre ell Schindler (MPCA)

From: Jonathan Butcher Project: Hawk Creek

Subject: Hawk Creek/Yellow w Medicine River Detailed Model

This memorandum transmits and de escribes the development and calibration of a 12-digi it HUC-scale

HSPF model of Hawk Creek, Yelloww Medicine River, and other minor tributaries of the e Minnesota River

contained within the 8-digit HUC 077020004 (Figure 1). The detailed model does not simmulate the portion

of the Minnesota River that runs thr rough the center of this HUC8; however, it could rea adily be linked to

the larger-scale model of the Minne esota River.

The model exists in two forms, curr rently labeled as HawkYM24.uci and HawkYM24-FFullSA.uci. The

latter version contains HSPF Specia al Actions implemented to account for sediment transsport through tile

drains (in addition to those that repr resent tillage and bluff collapse). These Special Actiions for tile

sediment transport are complex and d time-consuming to run, increasing run time from ab bout 15 minutes to

6 hours for a 10-year simulation, an nd must be applied to a large number of HRU combin nations.

Experiments with the model show thhat these Special Actions have no effect on hydrolog gy and contribute a

change in average pollutant concent tration of less than 2 percent at all analyzed locationss in the

Hawk/Yellow Medicine basins. Th herefore, it is recommended that the version without S Special Actions

for tile drain transport of sediment i is adequate for most analysis purposes in these basin ns.

The model uci files may be edited u using a text editor (we suggest the shareware TextPadd program) or

using the WinHSPF program. Each h has advantages for different situations. The model may be run using

WinHSPFLt or WinHSPF. Please n note the following:

• The WinHSPF engine shoululd be updated by copying the September 2009 versio on of hass_ent.dll

(previously provided) into t the Windows\System32 directory. This allows for a greater number of

operations.

• The uci files cannot be run wwith the DOS version of HSPF due to number of opeerations.

wq-ws4-13c

e River/Hawk Creek Basin

yonunty

The Yellow Medic

NAD_1983_StatePlane_Minnesota_Central_FIPS_2202_metersMil

Hawk - Yellow Medicine Watersheds

NAD_1983_StatePlane_Minnesota_Central_FIPS_2202_meters Map produced 06-15-2011 - P. Cada

SO

UT

H D

AK

OTA

MIN

NE

SO

TA

L Co

Lac Qui Parle County

Lincoln County

Yellow Medicine County

75

212

Legend

Highway

Major Waterway

Major Waterbody

Elevation

(meters)

High : 601

Low : 249

s

Lyon County

Redwood County

Chippewa County

Swift County

59

212

59

12

0 10 20 30 5 Mile

0 10 20 30 5 Kilometers

Yellow Medicine River

Hawk Creek

Renville County

Kandiyohi County

Brown County

71 12

212

es

Figure 1. The Yellow Medicin ine River/Hawk Creek Basin

2

response unit (HRU) basis, which takes into account land use,

ope

code

oup (HSG) combinations (although n

ed as (

precipitation station assignment.

ich is useful for parameter entry.

from multiple data sources.

spatial distribution of developed, agricultural, and undeveloped land

se NLCD lacks detailed information about types of crops and tillage

tatistical data were used to refine the agricultural land classification,

g Steps

ver in the Yellow Medicine/Hawk Creek Basin

ses representing degrees of development from rural to high intensity

ea grid was used to estimate representative percent impervious values

sses.

1)

The model is set up on a hydrologi t land use,

soil group (HSG), and s PERLND (and

IMPLND) has a three digit numeri del,

potential land use/hydrologic soil g . The three digit

code identifying an HRU is calcula e for the land

is th ERLNDs to be

ely by HRUs,

Land use/land cover was develope ased data

products were used to represent the ndeveloped land

Beca crops and tillage

level agricultural nd classification,

Processi

Land Use/Land C

NLCD provides four developed cla to high intensity

The HSPF model, on the other han impervious

The NLCD impervious a mpervious values

for each of the developed NLCD cl perv

Land Use and Land Cover Data (NLCD, 200

NAD_1983_StatePlane_Minnesota_Central_FIPS_2202_meter

Lac Qui Parl

s

The model is set up on a hydrologic response unit (HRU) basis, which takes into accou

along with weather station assignment. Each

. There are 14 weather stations used in the m

ot all are used

is the base co

This enables the

products were used to represent the spatial distribution of developed, agricultural, and

Because NLCD lacks detailed information about types of

level agricultural statistical data were used to refine the agricultural l

NLCD provides four developed classes representing degrees of development from rura

, was configured to separate developed pervious an

The NLCD impervious area grid was used to estimate representative percent

The analysis yielded the following percent i

Mil

1 Model Setup The model is set up on a hydrologic c response unit (HRU) basis, which takes into accoun nt land use,

hydrologic soil group (HSG), and sl lope – along with weather station assignment. Each PERLND (and

IMPLND) has a three digit numeric c code. There are 14 weather stations used in the mo odel, and 37

potential land use/hydrologic soil gr roup (HSG) combinations (although not all are used)). The three digit

code identifying an HRU is calculat ted as (hru – 1)*14 + wst. Where hru is the base cod de for the land

use/HSG combination and wst is the e precipitation station assignment. This enables the PPERLNDs to be

grouped consecutively by HRUs, whwhich is useful for parameter entry.

Land use/land cover was developed d from multiple data sources. NLCD 2001 GIS grid-bbased data

products were used to represent the spatial distribution of developed, agricultural, and u undeveloped land

in the watersheds (Figure 2). Becau use NLCD lacks detailed information about types of crops and tillage

practices, county-level agricultural s astatistical data were used to refine the agricultural la nd classification,

as discussed below under Processin ng Steps.

Land Use and Land Cover Data (NLCD, 200


SO

UT

H D

AK

OTA

MIN

NE

SO

TA


Lincoln County


Legend

Land Use/Land Cover (NLCD, 2001)

Open water

Developed, open space

Developed, low intensity

Developed, medium intensity

Developed, high intensity

Barren land

Deciduous forest

Evergreen forest

Mixed forest

Scrub/shrub

Grassland/herbaceous

Pasture/hay

Cultivated crops

Woody wetlands

Emergent herbaceous wetlands

1)

s

Lyon County

Redwood County

Chippewa County

e

Swift County

¯ 0 10 20 30 5 Mile



Hawk Creek

Renville County

Kandiyohi County

Brown County

es

Figure 2. Land Use/Land Co over in the Yellow Medicine/Hawk Creek Basin

NLCD provides four developed clas sses representing degrees of development from rural l to high intensity.

The HSPF model, on the other hand d, was configured to separate developed pervious and d impervious

surfaces. The NLCD impervious ar rea grid was used to estimate representative percent i impervious values

for each of the developed NLCD cla asses. The analysis yielded the following percent im mpervious values:

3

): 6.12%

7.65%

3): 58.65%

86.36%

ing county

soils were found to be rare in the watershed, and were combined

HRUs and simplify the HSPF model.

n

ons.

ed) designator was used for all other land uses.

oups

er DEM in ArcGIS.

CD grids, and smaller DEM cells would have yielded micro

consistent with the purpose of the slope classification.

ries

n areas t

inlets.

yonounty

Rural/open space (NLCD

Low intensity (NLCD 22):

Medium intensity (NLCD

High intensity (NLCD 24):

Soil HSG was developed by combi overage for the

HSG ere combined

with B soils to reduce the number ea also had a high

proportion of soils with a dual desi performance

under drained and undrained condit signator was used

for cropland and the second (undra

Hydrologic Soil G

Slope was calculated from a 30 me al for this project

because the cell size matched the N ed micro

ariations in slope that would not b on.

grid was reclassified into two categ than 1 percent).

This was done to distinguish betwe er slopes) versus

those that are not likely to have suc ulated only for

NAD_1983_StatePlane_Minnesota_Central_FIPS_2202_meters

level SSURGO GIS data into a unified

HSG A soils were found to be rare in the watershed, and

The study a

The two designators represen

During HRU processing, the first (drained) d

Thirty meter grid cells were id

because the cell size matched the NLCD grids, and smaller DEM cells would have yiel

ariations in slope that would not be consistent with the purpose of the slope classificat

Low (less than 1 percent), and High (greater

hat are likely to have surface tile inlets (lo

Sediment transport through tile drains is si

Mil

u212

• Rural/open space (NLCD 2211): 6.12%

• Low intensity (NLCD 22): 227.65%

• Medium intensity (NLCD 2 23): 58.65%

• High intensity (NLCD 24): 86.36%

Soil HSG was developed by combin ning county-level SSURGO GIS data into a unified c coverage for the

entire study area (Figure 3). HSG A A soils were found to be rare in the watershed, and w were combined

with B soils to reduce the number ooff HRUs and simplify the HSPF model. The study ar rea also had a high

proportion of soils with a dual desig gnation (e.g., “B/D”). The two designators represent t performance

under drained and undrained conditiions. During HRU processing, the first (drained) de esignator was used

for cropland and the second (undrainined) designator was used for all other land uses.

Hydrologic Soil Group (HSG, SSURGO)


SO

UT

H D

AK

OTA

MIN

NE

SO

TA

L C


Lincoln County


tu75

t

Legend

Hydrologic Soil Group

(SSURGO)

Water/Rock

A

A/D

B

B/D

C

C/D

D

s


Hawk Creek

Lyon County

Redwood County

Chippewa County

Swift County

tu59

tu212

tu59

tu12

¯ 0 10 20 30 5 Mile


Renville County

Kandiyohi County

Brown County

tu71

tu12

tu212

es

Figure 3. Hydrologic Soil Gr roups

Slope was calculated from a 30 met ter DEM in ArcGIS. Thirty meter grid cells were ide eal for this project

because the cell size matched the NLLCD grids, and smaller DEM cells would have yield ded micro-

variations in slope that would not be e consistent with the purpose of the slope classificati ion. The slope

grid was reclassified into two catego ories – Low (less than 1 percent), and High (greater than 1 percent).

This was done to distinguish betwee en areas that are likely to have surface tile inlets (low wer slopes) versus

those that are not likely to have suchh inlets. Sediment transport through tile drains is sim mulated only for

4

h slope categories were applied only to tilled

ow/High slope grid were combined in ArcGIS to produce a grid with

of NLCD land cover, HSG, and slope.

fol

uous, evergreen, and mixed) and shrubland were lumped into one

.

parately

lands were lumped into a single wetland category.

were lumped into a single grass/pasture category.

soils.

re found to be

relatively rare, and were lumped with C soils.

es were assumed to behave like D soils with poor infiltration, so all

e two catego

ion of the watershed (about 0.1 percent), so its HSG classes were

ry.

ed for all land except tilled land and grass/pasture.

BASINS4

us land retained the HSG subcategories.

use/land cover table with each HRU and weather station assignmen

re imported into the Land Use Processing spreadsheet, and the

ped in the

rocessed to further categorize by tillage practice and man

e and manured tillage land use classes were assigned based on

nd information on CRP/CREP/RIM acres as shown in the Land Use

ere is no explici

proximate proportions within each subbasin.

D water land use area was corrected for the surface area of lakes that

in the Land Use Processing spreadsheet

marized in

et.

Low and Hi d

NLCD land use, soil HSG, and the oduce a grid with

unique values for each combinatio

The resulting grid was simplified a

The three forest types (deci mped into one

category representing fores area in the study

s

Woody and herbaceous we

Grassland and Pasture/Hay

A soils were lumped with

For untilled land, C soils w category.

For tilled land, D soils wer

Water and wetland categor filtration, so all

HSGs were lumped for the

Barren land was a tiny frac classes were

lumped into a single categ

Slope classes were elimina

The simplified grid was processed nto pervious and

Developed pervi l result of the

BASINS4 operation provided a lan ation assignmen

Data w et, and the

processed developed lands were lu

man

The conservation tilla based on

level agricultural statistics in the Land Use

As such, t icultural land use.

Rather, they should be viewed as a

To avoid double counting, the NL rea of lakes that

y simulated as reache

The HRU numbering scheme is su d by the HRU

spreadsh

land a

NLCD land use, soil HSG, and the Low/High slope grid were combined in ArcGIS to p

The three forest types (deciduous, evergreen, and mixed) and shrubland were l

Shrubland made up a fraction of a percent of lan

B

Water and wetland categories were assumed to behave like D soils with poor i

Barren land was a tiny fraction of the watershed (about 0.1 percent), so its HS

, splitting the developed NLCD classes

The overa

BASINS4 operation provided a land use/land cover table with each HRU and weather s

Data were imported into the Land Use Processing spreadsh

processed to further categorize by tillage practice an

The conservation tillage and manured tillage land use classes were assigne

level agricultural statistics and information on CRP/CREP/RIM acres as shown

t spatial location for the types of ag

To avoid double counting, the NLCD water land use area was corrected for the surface

. Setup of the HRUs is controll

the lower slope areas. Low and Hig gh slope categories were applied only to tilled land annd

grassland/pasture.

Processing Steps

NLCD land use, soil HSG, and the L rLow/High slope grid were combined in ArcGIS to produce a grid with

unique values for each combination n of NLCD land cover, HSG, and slope.

The resulting grid was simplified as s follows, according to NLCD land cover classes:

• The three forest types (decidduous, evergreen, and mixed) and shrubland were lu umped into one

category representing forestt. Shrubland made up a fraction of a percent of land d area in the study

area and is not simulated se eparately.

• Woody and herbaceous wet tlands were lumped into a single wetland category.

• Grassland and Pasture/Hay were lumped into a single grass/pasture category.

• HSG classes

• A soils were lumped with B B soils.

• For untilled land, C soils we ere found to be rare, and were lumped with the A-B category.

• For tilled land, D soils were e relatively rare, and were lumped with C soils.

• Water and wetland categori ies were assumed to behave like D soils with poor in nfiltration, so all

HSGs were lumped for thes se two categories.

• Barren land was a tiny fract tion of the watershed (about 0.1 percent), so its HSG G classes were

lumped into a single catego ory.

• Slope classes were eliminat ted for all land except tilled land and grass/pasture.

The simplified grid was processed iinn BASINS4, splitting the developed NLCD classes iinto pervious and

impervious area. Developed pervio ous land retained the HSG subcategories. The overal ll result of the

BASINS4 operation provided a land d use/land cover table with each HRU and weather station assignment t,

as well as model subbasin. Data we ere imported into the Land Use Processing spreadshe eet, and the

processed developed lands were lum mped in the RawLU worksheet.

The tabulated tilled land was post-pprocessed to further categorize by tillage practice and d manure

application. The conservation tillag ge and manured tillage land use classes were assignedd based on

county-level agricultural statistics a and information on CRP/CREP/RIM acres as shown in the Land Use

Processing spreadsheet. As such, th here is no explicit spatial location for the types of agrricultural land use.

Rather, they should be viewed as ap pproximate proportions within each subbasin.

To avoid double counting, the NLC CD water land use area was corrected for the surface aarea of lakes that

were explicitly simulated as reaches s in the Land Use Processing spreadsheet

The HRU numbering scheme is sum mmarized in Table 1. Setup of the HRUs is controlle ed by the HRU and

Land Use Processing.xlsx spreadshe eet.

5

cheme

ed to model subbasins based on proximity.

e NLCD stations, and the model uses disaggregated hourly time

ech for the Minnesota River Turbidity TMDL project

HRU Numbering

The precipitation stations are assig ions are identified

stations hourly time

series previously created by Tetra t

. stations obtained

The sta

stations are NLCD stations, and the model uses disaggregate

series previously created by Tetra Tech for the Minnesota River Turbidity TMDL proje

, “SWCD” stations are additional SO

Table 1. HRU Numbering S Scheme

Land Use HSG Slope Base Code

Water all all 01

Barren all all 02

Wetland all all 03

Forest A,B,C all 04

Forest D all 06

Grass/Pasture A,B,C L 10

Grass/Pasture A,B,C H 11

Grass/Pasture D L 14

Grass/Pasture D H 15

Conventional Tillage A,B L 16

Conventional Tillage A,B H 17

Conventional Tillage C,D L 18

Conventional Tillage C,D H 19

Conservation Tillage A,B L 22

Conservation Tillage A,B H 23

Conservation Tillage C,D L 24

Conservation Tillage C,D H 25

Manured Tillage A,B L 28

Manured Tillage A,B H 29

Manured Tillage C,D L 30

Manured Tillage C,D H 31

Developed Pervious A,B,C all 34

Developed Pervious D all 36

Impervious all all 37

The precipitation stations are assign ned to model subbasins based on proximity. The stat tions are identified

in Table 2. The last eight stations aarre NLCD stations, and the model uses disaggregated d hourly time

series previously created by Tetra T Tech for the Minnesota River Turbidity TMDL projecct (through 2006),

with augmented data through 2009. The first six, “SWCD” stations are additional SOD D stations obtained

and processed by MPCA.

6

ons

d

outh

e numbering. The reaches and associated precipitation station

hile the spatial distribution of subbasins is shown in

ns, and Weather Station Assignments

edicine River mouth

dicine River at Gage

Trib 3 to Yellow Medicine River


dicine River


nch Yellow Medicine River


Trib 1 to South Branch Yellow Medicine River

Precipitation Stat

subbasins a llow Medicine

River from the USGS gage to the t drainage). The

hes have the sa on station

,

Reaches, Subbas

Yellow

Yellow M

Unname

Unname

ellow M

Unname

South Br

South Br

Unname

Y

dire

hes have the same numbering. The reaches and associated precipitat

, while the spatial distribution of subbasins is shown i

Table 2. Precipitation Stati ions

Met Station WST# Name

SWCD1 1 253879-4947441

SWCD2 2 270561-4962989

SWCD3 3 297650-4952225

SWCD4 4 302088-4940775

SWCD5 5 329702-4946407

SWCD6 6 341200-4980208

MN213311 7 Granite Falls

MN215204 8 Marshall

MN215482 9 Minneota

MN215563 10 Montevideo 1 SW

MN216152 11 Olivia 1 SE

MN218429 12 Tyler

MN219004 13 Willmar RTC

SD390422 14 Astoria, SD 4S

The model contains 75 subbasins an nd 76 stream reaches (RCHRES 100, representing Yeellow Medicine

River from the USGS gage to the m mouth, is a routing reach only, with no assigned direc ct drainage). The

subbasins and reaches have the sam me numbering. The reaches and associated precipitatiion station

assignments are shown in Table 3, w while the spatial distribution of subbasins is shown inn Figure 4.

Table 3. Reaches, Subbasi ins, and Weather Station Assignments

Reach Number Name WST #

100 Yellow M Medicine River mouth 3

101 Yellow Me edicine River at Gage 3

102 Unnamed d Trib 3 to Yellow Medicine River 3


104 Yellow Me edicine River 8


106 South Bra anch Yellow Medicine River 9


108 Unnamed d Trib 1 to South Branch Yellow Medicine River 12

7


Trib 2 to South Branch Yellow Medicine River

dicine River


dicine River

dicine River

Lake

nch Yellow Medicine Ri


k

ek

Trib 3 to Spring Creek

ek


ek


ring Creek

e Creek

Trib 1 to Wood Lake Creek

Trib 2 to Wood Lake Creek

e

k

k

Trib 2 to Minnesota River

South Br

Unname

Yellow M

Unname

Yellow M

Yellow M

Shaokota

North Br

North Br

Mud Cre

Spring Cr

Unname

Spring Cr

Unname

Spring Cr

Unname

Rice Cree

Boiling S

Wood La

Unname

Unname

Wood La

Hazel Cre

Hazel Cre

Unname



110 Unnamed d Trib 2 to South Branch Yellow Medicine River 9





115 Shaokotan n Lake 14

116 North Bra anch Yellow Medicine River 1

117 North Bra anch Yellow Medicine River 14

118 Mud Cree ek 1

119 Spring Cre eek 2

120 Unnamed d Trib 3 to Spring Creek 2





130 Rice Creekk 5

140 Boiling Sp pring Creek 4

150 Wood Lak ke Creek 3

151 Unnamed d Trib 1 to Wood Lake Creek 4

152 Unnamed d Trib 2 to Wood Lake Creek 4

153 Wood Lak ke 3

160 Hazel Creeek 3

161 Hazel Creeek 2

170 Unnamed d Trib 2 to Minnesota River 7

8

Creek


k

k

tch 11

Trib 2 to Hawk Creek

k

k

Trib 1 to Hawk Creek

s Lake

mon Lake

k

k

k

tch 2

Creek

tch 1

Creek

Creek

tch 8

eek

Beaver Creek

Stony Ru

Unname

Hawk Cre

Hawk Cre

County D

Unname

Hawk Cre

Hawk Cre

Unname

Saint Joh

West Sol

Long Lak

Hawk Cre

Hawk Cre

Hawk Cre

Foot Lak

Eagle Lak

Judicial D

Chetomb

Judicial D

Chetomb

Chetomb

County D

Beaver C

West For


180 Stony Run n Creek 10

190 Unnamed d Trib 1 to Minnesota River 10

201 Hawk Creeek 7

202 Hawk Creeek 7

203 County Di itch 11 7

204 Unnamed d Trib 2 to Hawk Creek 7

205 Hawk Creeek 7

206 Hawk Creeek 7

207 Unnamed d Trib 1 to Hawk Creek 13

208 Saint John ns Lake 13

209 West Solo omon Lake 13

210 Long Lake e 13

211 Hawk Creeek 6

212 Hawk Creeek 13

213 Hawk Creeek 13

214 Foot Lake e 13

215 Eagle Lake e 13

216 Judicial Diitch 2 6

217 Chetomba a Creek 7

218 Judicial Diitch 1 7




230 Beaver Cr reek 5

231 West Fork k Beaver Creek 5

9

Beaver Creek

tch 59

Beaver Creek

Beaver Creek

ek

art Creek, East Branch

art Creek


eek

West For

County D

East Fork

East Fork

Timms Cr

Sacred H

Sacred H

Palmer C

esot

esot

esot

esot

esot

esot

esot

esot


232 West Fork k Beaver Creek 6


234 East Fork Beaver Creek 11

235 East Fork Beaver Creek 11

240 Timms Cre eek 5

250 Sacred He eart Creek, East Branch 5

260 Sacred He eart Creek 5

270 Unnamedd Trib 3 to Minnesota River 7

280 Palmer Cr reek 7

301 Minnesotaa River direct drainage 301 5








10

on

ere developed from

, 104, and 111)

was not able to provide these.

es generated by BASINS4.

akes are explicitly represented in the model

ns, Eagle, Swan, Willmar, Foot, Shaokotan

ted for these

arge characteristics.

ssing spreadsheet reassigns agricultura

nd categories, and also removes the water area represented by

adsheet is also set up with functionality to represent a variety of

me procedures as for the Minnesota River model. This is set up on

ustments are implemented at this time.

sition of nitrogen are included in the model based on information

odel

ns are shown in

n the model using the generic, aggregated approach employed in the

e are summarized in

Model Segmentat

Hydraulic FTables for the reaches s where available

(Yellow Medicine reaches 100, 10 for Beaver Creek

and lower Hawk Creek, but MDN covered by HEC

default FTab portant in the

Long, Solomon,

West Solomon, Lindgren, Saint Jo nd Mud Lakes

and separate FTables have been cre d on available

information about storage and disc

As noted above, the Land Use Proc entional tillage,

conservation tillage, and manured l sented by

explicitly simulated lakes. This spr t a variety of

using the s his is set up on

the Scenario tab, but no scenario a

Point sources and atmospheric dep information

developed for the Minnesota River ct

simulated major point source locati rges from

ted employed in the

Minnesota River basin model. The


1

21

10

s

RAS mode

RAS models apparently exis

Flowing reaches not

Lakes are particularly i

Ringo

, Wood,

bas

l land in con

conservation tillage, and manured land categories, and also removes the water area repr

explicitly simulated lakes. This spreadsheet is also set up with functionality to represe

using the same procedures as for the Minnesota River model.

Point sources and atmospheric deposition of nitrogen are included in the model based o

and supplemented through 2009 for this proj

. In addition, minor disch

ted in the model using the generic, aggregated approach

2

22

Mil

Model Segmentation


SO

UT

H D

AK

OTA

MIN

NE

SO

TA


Lincoln County


118

113 117

109

112

123

105

108

116

107 110

111

124

122

115

114

106

Legend

Major Waterway

Major Waterbody

Model Subwatershed


s

Lyon County

Redwood County

Chippewa County

Swift County

180

219

104

161

305

232

307

23

308

103

211

2

140

204

160

151

280

207

306

101

220

130

119

5

303

201

301

212

120

216

302

217

240 304

260

203

170

190

202

250

102

205

152

206

270

213

218

209

121

153

150

208

¯ 0 10 20 30 5 Mile


Hawk Creek

Renville County

Kandiyohi County

Brown County

234 31

221

0

230

235

215

233

214

210

es

Figure 4. Model Segmentati ion

Hydraulic FTables for the reaches w were developed from Lyon County HEC-RAS modells where available

(Yellow Medicine reaches 100, 101 1, 104, and 111). HEC-RAS models apparently exist t for Beaver Creek

and lower Hawk Creek, but MDNR R was not able to provide these. Flowing reaches not covered by HEC

RAS models have the default FTabl les generated by BASINS4. Lakes are particularly im mportant in the

Hawk Creek headwaters. Thirteen llakes are explicitly represented in the model (Ringo, , Long, Solomon,

West Solomon, Lindgren, Saint Joh hns, Eagle, Swan, Willmar, Foot, Shaokotan, Wood, aand Mud Lakes)

and separate FTables have been creaated for these lakes (or sets of connected lakes) base ed on available

information about storage and disch harge characteristics.

As noted above, the Land Use Proce essing spreadsheet reassigns agricultural land in convventional tillage,

conservation tillage, and manured la and categories, and also removes the water area repreesented by

explicitly simulated lakes. This spreeadsheet is also set up with functionality to represen nt a variety of

management measures, using the sa ame procedures as for the Minnesota River model. TThis is set up on

the Scenario tab, but no scenario adjdjustments are implemented at this time.

Point sources and atmospheric depo osition of nitrogen are included in the model based onn information

developed for the Minnesota River mmodel and supplemented through 2009 for this proje ect. Explicitly

simulated major point source locatio ons are shown in Figure 5. In addition, minor discha arges from

stabilization ponds are represented i in the model using the generic, aggregated approach employed in the

Minnesota River basin model. Thes se are summarized in Table 4.

11

d Point SourcesExplicitly Simulat

NAD_1983_StatePlane_Minnesota_Central_FIPS_2202_metersMil

Point Sources Included in Model


SO

UT

H D

AK

OTA

MIN

NE

SO

TA


Lincoln County


Legend

#0 Point Source

Major Waterway

Major Waterbody

Model Subwatershed


s

#0

#0

#0

#0

#0

WILLMAR WWTP

Lyon County

Redwood County

Chippewa County

Swift County

OLIVIA WWTP

MAYNARD WWTP

RENVILLE WWTP

CLARA CITY WWTP

SOUTHERN MINN. SUGAR COOP.

¯ 0 10 20 305 Mile

0 10 20 305 Kilometers

Hawk Creek

#0

#0

Renville County

Kandiyohi County

Brown County

es

Figure 5. Explicitly Simulateed Point Sources

12

s in the Model

lls WWTP

od WWTP

WWTP

WTP

WTP

TP

TP

WTP

TP

e WWTP

WWTP

WWTP

WWTP

WTP

d WWTP

t pipe d

communities”) are explicitly represented on a population basis

k Creek basin, Hazel Run and Delhi in the Yellow Medicine basin).

ed to stabilization ponds prior to 2007, but this is not yet represented

eas the model assumes a regional rate of direct

sing the m

on and Validation Report

should be reviewed before using the model for bacteria simulati

ich larger

ek, Bunde).

Stabilization Pon

odel Subbasin

Hanley F

Cottonw

Minneot

Taunton

Ivanhoe

Porter W

St Leo W

Belview

Echo W

Wood La

Clarkfiel

Pennock

Raymon

Danube

Bird Isla

The model also accounts for straig communities

with direct discharges (“unsewered ation basis

(Blomkest and Prinsburg in the Ha edicine basin).

Both Prinsburg and Hazel Run shif t yet represented

in the model. For unincorporated a le individual

sewage treatment systems (ISTS), 2

River Basin Model, Model Calibrat are estimated on a

per capita basis. These assumption eria simulati

in w s

(e.g., Svea, Roseland, Gl

ischarges to tile drain systems. Incorporate

with direct discharges (“unsewered communities”) are explicitly represented on a popu

(Blomkest and Prinsburg in the Hawk Creek basin, Hazel Run and Delhi in the Yellow

Both Prinsburg and Hazel Run shifted to stabilization ponds prior to 2007, but this is n

t

ethodology described in Tetra Tech’s 20

, in which pollutant loads

per capita basis. These assumptions should be reviewed before using the model for bac

concentrations of direct discharging syste

Table 4. Stabilization Pond ds in the Model

Permit No. Name Receiving Water MModel Subbasin

MNG580122 Hanley Fa alls WWTP Yellow Medicine 101

MNG580010 Cottonwo ood WWTP JD #10 103

MNG580033 Minneota a WWTP SB Yellow Medicine 106

MNG580090 Taunton W WWTP unnamed to YMR 112

MNG580103 Ivanhoe W WWTP Yellow Medicine 113

MNG580128 Porter WW WTP Mud Cr 118

MN0024775 St Leo WWWTP unnamed to YMR 123

MNG580003 Belview W WWTP CD #12 130

MNG580059 Echo WW WTP Boiling Springs Ck 140

MNG580107 Wood Lakke WWTP JD #10 152

MNG580093 Clarkfield d WWTP CD #9 161

MNG580104 Pennock WWTP trib to Hawk Cr, ds St. Johns lake 207

MN0045446 Raymond d WWTP Hawk Cr 211

MNG580057 Danube W WWTP WF Beaver Ck 231

MN0022829 Bird Islan nd WWTP CD #66 234

The model also accounts for straigh ht pipe discharges to tile drain systems. Incorporated d communities

with direct discharges (“unsewered communities”) are explicitly represented on a popul lation basis

(Blomkest and Prinsburg in the Haw wk Creek basin, Hazel Run and Delhi in the Yellow MMedicine basin).

Both Prinsburg and Hazel Run shift ted to stabilization ponds prior to 2007, but this is no ot yet represented

in the model. For unincorporated ar reas the model assumes a regional rate of direct-to-ti ile individual

sewage treatment systems (ISTS), u using the methodology described in Tetra Tech’s 200 02 Minnesota

River Basin Model, Model Calibratiion and Validation Report, in which pollutant loads are estimated on a

per capita basis. These assumptions s should be reviewed before using the model for bactteria simulations,

particularly in regard to areas in wh hich larger concentrations of direct discharging system ms have been

identified (e.g., Svea, Roseland, Glu uek, Bunde).

13

(This page left intentionally blank.)

14

ibrationlong period of record in

ith the Minnesota River basin model revealed this as a difficult gage

somewhat difference rainfall

to mid

Beaver Creek watersheds by the Hawk Creek Watershed Project

pera

the spring runoff may be missed, complicating efforts to fit an overall

ve been concerns with rating curves at some of

esults for 2000 yielded flows that are significantly higher than those at

th gage, later determined to be a result of deficiencies in the rating

ating curves have been field measured and adjusted is apparently less

d Water Quality Monitoring Locations

calibration was to foc

age were then extended to the various Hawk Creek and Beaver Creek

alidation test

Hydrologic CaThere is only one USGS gage with ne River near

and previous work s a difficult gage

to calibrate, which apparently has a he last decade

seen in the 1980s and earl en conducted at

multiple locations in the Hawk and shed Project

. These gages, however, ugh September),

which means that a large portion of ts to fit an overall

water balance. In addition, there h gage locations

(for instance, the Chetomba Creek igher than those at

the downstream Hawk Creek at mo es in the rating

ncy at which is apparently less

Stream Gaging a

The general strategy for hydrologic e gage for 2001

2009. Parameters derived for that nd Beaver Creek

gages as a spatial corroboration or unified


-341

03-867

s

Yellow Medic

and previous work with the Minnesota River basin model revealed this

runoff response over

1990s. Since 1999 additional gaging has b

multiple locations in the Hawk and Beaver Creek watersheds by the Hawk Creek Wate

te only on a seasonal basis (generally April thr

which means that a large portion of the spring runoff may be missed, complicating effo

thes

(for instance, the Chetomba Creek results for 2000 yielded flows that are significantly

the downstream Hawk Creek at mouth gage, later determined to be a result of deficienc

ncy at which rating curves have been field measured and adjuste

us first on the Yellow Medici

2009. Parameters derived for that gage were then extended to the various Hawk Creek

, with some iterative adjustments to th

S00

S

Mil

2 Hydrologic Callibration There is only one USGS gage with aa long period of record in the basin – Yellow Mediciine River near

Granite Falls – and previous work w with the Minnesota River basin model revealed this a as a difficult gage

to calibrate, which apparently has a somewhat difference rainfall-runoff response over t the last decade

than was seen in the 1980s and early y to mid-1990s. Since 1999 additional gaging has beeen conducted at

multiple locations in the Hawk and Beaver Creek watersheds by the Hawk Creek Water rshed Project

(Figure 6). These gages, however, o operate only on a seasonal basis (generally April throough September),

which means that a large portion of the spring runoff may be missed, complicating effor rts to fit an overall

water balance. In addition, there ha ave been concerns with rating curves at some of thesee gage locations

(for instance, the Chetomba Creek r results for 2000 yielded flows that are significantly h higher than those at

the downstream Hawk Creek at mou uth gage, later determined to be a result of deficienciies in the rating

curve), and the frequency at which r rating curves have been field measured and adjusted d is apparently less

than the USGS standard.

Flow Monitoring Stations


SO

UT

H D

AK

OTA

MIN

NE

SO

TA


Lincoln County


Legend

#* Flow Monitoring Station

Major Waterway

Major Waterbody

Model Subwatershed


s

#* #*

#* #* #*

#*#*#*

#*

Lyon County

Redwood County

Chippewa County

Swift County

S002-140

S002-148

S002-152

S002-136

S002-012

S000-159

S002-316 S001

S0

¯ 0 10 20 30 5 Mile


Hawk Creek

#*

#* Renville County

Kandiyohi County

Brown County

S000-405

1-341

003-867

S000-666

es

Figure 6. Stream Gaging an nd Water Quality Monitoring Locations

The general strategy for hydrologic calibration was to focus first on the Yellow Medicin ne gage for 2001

2009. Parameters derived for that g gage were then extended to the various Hawk Creek aand Beaver Creek

gages as a spatial corroboration or v validation test, with some iterative adjustments to the e unified

15

rmance on the earlier gage records for Yellow Medicine was

tion test.

rameters was provided by the Hawk/Yellow Medicine section of the

el and refined to reflect the more detailed representation of HRUs.

were to the lower zone nominal soil moisture storage (LZSN, varied

meter (INFILT, assigned according

adjusted to (1) reflect lessons learned in the development of the

2) to better conform

r model was that snowmelt was shifted from an energy balance

OP=1), depending only on air temperature. This was done because

eld better results for the simulation of s

into account solar radiation, wind, relative humidity, cloud cover)

be very sensitive to uncertainty in the meteorological inputs. In

implemented in HSPF provides little user control over the time

ajor factor in the uptake of heat by the snowpack.

uired to achieve acceptable fit. This was in the exer

required for portions of the Hawk Creek (exclusive of Chetomba

d number

libration are provided in A

ine gage for 2001

s good, except that there is considerable uncertainty regarding the

daily Nash

round 0

s been disaggregated, not by true hourly precipitation.

ations, flows at the Yellow Medicine gage for 1994 to 2000 tend to

all seasons and for both high and low flows, although the NSE values

epancy are not fully understood.

r set. Finally, model perf cine was

examined as an additional corrobor

The starting point for hydrologic p e section of the

er Basin mo ation of HRUs.

The primary calibration adjustment ge (LZSN, varied

by land use) and the infiltration par ic group).

Various other parameters were also ment of the

LeSueur River detailed model, and in BASINS

change from the pri gy balance

method to a degree day method (S s done because

the degree day method appears to y elt. In theory, the

energy balance method (also taking , cloud cover)

should be superior; however, it can l inputs. In

gy balance metho ver the time

course of snow albedo, which is a

Only one spatial adjustment was re

which a somewhat lower factor wa of Chetomba

Creek) and East Fork Beaver Cree f Penman Pan

Evaporation estimates from a limit surements.

Detailed results of the hydrologic c s obtained to all

aspects of flow at the Yellow Medi e various

and Beaver Creek gages also appea regarding the

For these stations, th fair, while the

excellent ( el is driven

mostly by daily precipitation that h n.

As was noted in earlier model appli to 2000 tend to

in h the NSE values

are good. The reasons for this disc

r set. Finally, model performance on the earlier gage records for Yellow Med

The starting point for hydrologic parameters was provided by the Hawk/Yellow Medici

er Basin model and refined to reflect the more detailed represen

The primary calibration adjustments were to the lower zone nominal soil moisture stor

to soil hydrolo

Various other parameters were also adjusted to (1) reflect lessons learned in the develo

recommendations containe

change from the prior model was that snowmelt was shifted from an ene

method to a degree day method (SNOP=1), depending only on air temperature. This w

pring snow

energy balance method (also taking into account solar radiation, wind, relative humidit

should be superior; however, it can be very sensitive to uncertainty in the meteorologic

gy balance method implemented in HSPF provides little user control

Only one spatial adjustment was required to achieve acceptable fit. This was in the exe

which a somewhat lower factor was required for portions of the Hawk Creek (exclusiv

This likely reflects uncertainties in extrapolation

of stations with wind and temperature me

. A good fit

2009. Fit to observed flows at t

and Beaver Creek gages also appears good, except that there is considerable uncertaint

Sutcliffe coefficients (NSE) tend to be

.9). This likely reflects the fact that the mo

mostly by daily precipitation that has been disaggregated, not by true hourly precipitati

As was noted in earlier model applications, flows at the Yellow Medicine gage for 199

in all seasons and for both high and low flows, althou

parameter set. Finally, model perfo ormance on the earlier gage records for Yellow Medi icine was

examined as an additional corrobora ation test.

The starting point for hydrologic pa arameters was provided by the Hawk/Yellow Medicinne section of the

existing Minnesota River Basin moddel and refined to reflect the more detailed represent tation of HRUs.

The primary calibration adjustments s were to the lower zone nominal soil moisture stora age (LZSN, varied

by land use) and the infiltration para ameter (INFILT, assigned according to soil hydrolog gic group).

Various other parameters were also adjusted to (1) reflect lessons learned in the develop pment of the

LeSueur River detailed model, and ((2) to better conform to recommendations contained d in BASINS

Technical Note 6.

One important change from the prio or model was that snowmelt was shifted from an enerrgy balance

method to a degree day method (SN NOP=1), depending only on air temperature. This wa as done because

the degree day method appears to yiield better results for the simulation of spring snowm melt. In theory, the

energy balance method (also taking into account solar radiation, wind, relative humidity y, cloud cover)

should be superior; however, it can be very sensitive to uncertainty in the meteorologica al inputs. In

addition, the energy balance method d implemented in HSPF provides little user control o over the time

course of snow albedo, which is a m major factor in the uptake of heat by the snowpack.

Only one spatial adjustment was req quired to achieve acceptable fit. This was in the exerrted PET, for

which a somewhat lower factor was s required for portions of the Hawk Creek (exclusive e of Chetomba

Creek) and East Fork Beaver Creekk.. This likely reflects uncertainties in extrapolation o of Penman Pan

Evaporation estimates from a limite ed number of stations with wind and temperature meaasurements.

Detailed results of the hydrologic ca alibration are provided in Attachment 1. A good fit iis obtained to all

aspects of flow at the Yellow Medic cine gage for 2001-2009. Fit to observed flows at th he various Hawk

and Beaver Creek gages also appear rs good, except that there is considerable uncertainty y regarding the

lowest flows. For these stations, the e daily Nash-Sutcliffe coefficients (NSE) tend to be fair, while the

monthly NSE values are excellent (aaround 0.9). This likely reflects the fact that the moddel is driven

mostly by daily precipitation that ha as been disaggregated, not by true hourly precipitatio on.

As was noted in earlier model appliccations, flows at the Yellow Medicine gage for 1994 4 to 2000 tend to

be consistently underestimated – in all seasons and for both high and low flows, althoug gh the NSE values

are good. The reasons for this discr repancy are not fully understood.

16

alibrationthe Minnesota Rive

nt, temperature, BOD/DO, nutrients, and bacteria. However, full

r sediment, inorganic N, total N, inorganic P, and total P at this time.

e

Note 8.

the Minnesota River Basin model, cri

t h

the 15

e

tress (Tau) Distribution for Yellow Mnite Falls)

t with the Minnesota River model using Special Actions to replenish

hese loads consisted of 1.

he lowermost reach of Hawk Creek (201). A reach

t was conducted to ensure reasonable sediment accumulation/loss

ed

7

nd in the lake reaches (

is a temporary phenomenon, reflecting accumulation during 2008

bed storages in preceding years.

clusive of bluff loads) had a median of

segments, consistent with observations that a significant portion of the

annel

Water Quality The model is set up consistent with ange of water

quality simulation, including sedim However, full

calibration has been pursued only f tal P at this time.

Sediment calibration is the most ti onducted

consistent with BASINS Technical ased on

As wit esses for

cohesive sedime ated distribution.

shear stress,

was set to deposit below mple of the shear

Figu

Simulated Shear ch 101 (Yellow Medicine at USGS Gage near Gr

Bluff loads were assigned consiste ions to replenish

the sediment available in the bed. reach of Yellow

in

balance analysis of channel sedime mulation/loss

behavior and approximately stable tes of channel

degradation over the course of the et accumulation

gain represented in the bluff reache n during 2008

and 2009 after large events reduce ency (depth/scour

a fraction of influent sediment, e with positive

trapping (> 80 percent) for the lake cant portion of the

total load in this basin arises from

r Basin model to provide a full

quality simulation, including sediment, temperature, BOD/DO, nutrients, and bacteria.

calibration has been pursued only for sediment, inorganic N, total N, inorganic P, and t

consuming part of model development, and was

Upland sediment erodibility rates were set

tical shear st

are set equal to percentiles of the simu

percentile dail

percentile. An ex

edicine Re

Bluff loads were assigned consistent with the Minnesota River model using Special Ac

mos

balance analysis of channel sediment was conducted to ensure reasonable sediment acc

The final simulation shows slow r

, with the exception of

210, 21

gain represented in the bluff reaches is a temporary phenomenon, reflecting accumulati

Net trapping effic

11 percent

trapping (> 80 percent) for the lake segments, consistent with observations that a signif

3 Water Quality C CalibrationT

au

(lb

/ft

2)

The model is set up consistent with the Minnesota River Basin model to provide a full r range of water

quality simulation, including sedime ent, temperature, BOD/DO, nutrients, and bacteria. However, full

calibration has been pursued only fo or sediment, inorganic N, total N, inorganic P, and to otal P at this time.

Sediment calibration is the most tim me-consuming part of model development, and was cconducted

consistent with BASINS Technical Note 8. Upland sediment erodibility rates were set b based on

SSURGO USLE K factors. As with h the Minnesota River Basin model, critical shear str resses for scour

and deposition of cohesive sedimen nt in channels are set equal to percentiles of the simullated distribution.

Silt was set to deposit below the 20tth percentile and scour above the 95

thpercentile dailyy shear stress,

while clay was set to deposit below the 15th and scour above the 90

thpercentile. An exa ample of the shear

stress distribution is shown in Figur re 7.

0.3

0.25

0.2

0.15

0.1

0.05

0

0.1 1 10 100 1000 10000

Flow (cfs)

Figure 7. Simulated Shear S Stress (Tau) Distribution for Yellow Medicine Rea ach 101 (Yellow Medicine at USGS Gage near Gra anite Falls)

Bluff loads were assigned consisten nt with the Minnesota River model using Special Act tions to replenish

the sediment available in the bed. T These loads consisted of 1.2 tons/hr in the lowermost t reach of Yellow

Medicine (100) and 0.97 tons/hr in tthe lowermost reach of Hawk Creek (201). A reach--by-reach mass

balance analysis of channel sedimen nt was conducted to ensure reasonable sediment accu umulation/loss

behavior and approximately stable b bed composition. The final simulation shows slow rates of channel a

degradation over the course of the 1 17-year simulation (Figure 8), with the exception of n net accumulation

in the bluff reaches (100 and 201), aand in the lake reaches (111, 115, 153, 208-210, 214 4-215). The large

gain represented in the bluff reaches s is a temporary phenomenon, reflecting accumulatio on during 2008

and 2009 after large events reduced d bed storages in preceding years. Net trapping efficiiency (depth/scour

as a fraction of influent sediment, ex xclusive of bluff loads) had a median of -11 percent,, with positive

trapping (> 80 percent) for the lake segments, consistent with observations that a signifi icant portion of the

total load in this basin arises from chchannel processes (Figure 9).

17

in Sediment Bed Depth over 1996

ping Efficiency by Reach

same approach used in the Minnesota River Basin models. Ammonia,

d generalized organic matter are simulated on the land surface, with

ildup

e Mass

ganic phosphorus, and BOD at entry to the stream.

has c

this effort, the calibration focused on six stations with long periods of

hetomba Creek (S002

est Fork Beaver Creek (S000

iver at Mouth (S000

esota

etomba Creek and Yellow Medicine River at Mouth as these are not

in Prinsburg, Svea, Blomkest, and Roseland; Yellow Medicine River

ystems but no major discharges

Simulated Chang on

Net Sediment Tra

llows the odels. Ammonia,

nitrate nitrogen, orthophosphate, a d surface, with

the first two being represented by b ted as sediment

T rganic matter to

organic carbon, organic nitrogen, o

The Hawk Creek Watershed Projec ity calibration.

Given the limits of the schedule for th long periods of

record and a variety of conditions: rd (S002

012), ine River at Gage

316), and Yellow Medicine ations was used

for calibration in the previous Min

The initial calibration focused on C h as these are not

strongly impacted by point sources k

system w Medicine River

has a number of stabilization pond ing for the 2006

11

81

20

2009 Simulat

llows the same approach used in the Minnesota River Basin

nitrate nitrogen, orthophosphate, and generalized organic matter are simulated on the la

two simul

Link table is used to apportion generalized

ollected a wealth of data useful for water qua

Given the limits of the schedule for this effort, the calibration focused on six stations w

n

405), Yellow Medi

159). Only the last of these s

The initial calibration focused on Chetomba Creek and Yellow Medicine River at Mou

point sources on Chetomba Cre

systems in Prinsburg, Svea, Blomkest, and Roseland; Yell

). Intensive monito

0.9

0.8

0.7

30

7

Figure 8. Simulated Change e in Sediment Bed Depth over 1996-2009 Simulatiion

120%

100%�

80%�

60%�

40%�

20%�

0%�

-20%�

-40%�

-60%�

-80%�

Ch

an

ge

in

De

pth

(ft

)

10

0

10

21

00

10

4

10

6

10

8

11

0

11

2

11

4

11

6

11

8

12

0

12

2

12

4

14

0

15

1

15

3

16

1

18

0

20

1

20

3

20

5

20

7

20

9

21

1

21

3

21

5

21

7

21

9

22

1

23

1

23

3

23

5

23

5

25

0

10

3

10

6

10

9

11

2

11

5

11

8

12

1

12

4

15

0

15

3

17

0

20

1

20

4

20

7

21

0

21

3

21

6

21

9

23

0

23

3

24

02

40

27

02

70

30

23

01

30

33

05

3

05

0.6

0.5

0.4

0.3

0.2

0.1

0

-0.1

-0.2

Figure 9. Net Sediment Trap pping Efficiency by Reach

The nutrient simulation follows the same approach used in the Minnesota River Basin m models. Ammonia,

nitrate nitrogen, orthophosphate, an nd generalized organic matter are simulated on the lannd surface, with

the first two being represented by buuildup-washoff processes and the second two simula ated as sediment-

associated with potency factors. Th he Mass-Link table is used to apportion generalized o organic matter to

organic carbon, organic nitrogen, or rganic phosphorus, and BOD at entry to the stream.

The Hawk Creek Watershed Projectt has collected a wealth of data useful for water quallity calibration.

Given the limits of the schedule for this effort, the calibration focused on six stations wi ith long periods of

record and a variety of conditions: C Chetomba Creek (S002-152), Hawk Creek at Mayna ard (S002-148),

Hawk Creek at Mouth (S002-012), WWest Fork Beaver Creek (S000-405), Yellow Medic cine River at Gage

(S002-316), and Yellow Medicine R River at Mouth (S000-159). Only the last of these st tations was used

for calibration in the previous Minn nesota River modeling effort.

The initial calibration focused on Chhetomba Creek and Yellow Medicine River at Moutth as these are not

strongly impacted by point sources ((there are no major point sources on Chetomba Cree ek, although there

are some direct discharging systems s in Prinsburg, Svea, Blomkest, and Roseland; Yello ow Medicine River

has a number of stabilization pond s systems but no major discharges). Intensive monitor ring for the 2006

18

sis for calibration, with earlier monitori

ted by a number of factors, including the presence of large point

awk Creek) and Southern Minnesota Beet Sugar (discharged to

er Creek through August 2004, subsequently to County Ditch

entrations are generally reported only monthly, leading to

ading time series. In addition, ammonia is the

s, requiring use of approximate assumptions to specify concentrations

surprisingly, this leads to considerable uncertainty in the prediction

m of these sources. In Hawk Creek it appears that the total N loading

ng the period around 2005, leading to over

presented in more recent dat

rient simulation is evident strong interactions between nutrients and

ts planktonic algae and periphyton, but does not have a routine to

se of nutri

ty of nutrient and algal processes when stream depth falls below 3

y). In addition to preventing algal uptake of nutrients, thi

il depth again increases above the threshold. These factors tend to

utrient concentrations at very low flows, particularly for stations on

nally predicts very high concentration values (but low net loads)

e stranding phenomenon. This biases statistics based on average

edian error statistics.

nt fit generally required only small modifications to the previous

. One particular area of improvement was in the specification of

d with tile drainage. Model fit was great

oncentrations associated with the areas nearer the Minnesota River

ns 3

be characterized by more wetlands and ponds (areas associated with

and confirm

ion results is provided in Attachment 2. The quality of model fit

ing to parameter and location, and is consistent with other

ota River Basi

urce nutrient loads and the simulation of nutrients under very low

is appropriately calibrated to support scenario evaluation for

ed above, the model has not been fully calibrated for dissolved

2009 period was used as the main g a corroboration

Calibration for nutrients is complic f large point

sources at Willmar (discharging to ischarged to

t Fork Bea nty Ditch

Sacred Heart Creek). Pollutant con ing to

considerable uncertainty in actual l form of nitrogen

regularly monitored in the discharg ify concentrations

of nitrate and organic nitrogen. No n the prediction

rations downstre the total N loading

estimated dur on of instream N

concentrations, but is much better r

Another factor complicating the nu en nutrients and

algae/macrophytes. HSPF represe e a routine to

explicitly represent uptake and rele PF model code

shuts down the simulation of a vari falls below 3

(to prevent numeric instabili ts, thi

in “stranding” of pollutant mass un factors tend to

prediction of y for stations on

casi w net loads)

during low flow conditions due to t d on average

concentration, but does not affect

rove nutri the previous

setup of the Minnesota River mode cification of

interflow N concentrations associat d by assigning

two seasonal patterns, with higher innesota River

(areas associated with weather stati ociated with the

at appear t associated with

9, and 1 is apparent

difference need further investigatio

The full set of water quality simula of model fit

ranges from fair to excellent, accor other

applications of HSPF in the Minne model and data

appear to be associated with point s nder very low

Overall, the mode ation for

sediment and nutrients. As mentio r dissolved

ng providi

Calibration for nutrients is complicated by a number of factors, including the presence

sources at Willmar (discharging to Hawk Creek) and Southern Minnesota Beet Sugar (

t Fork Beaver Creek through August 2004, subsequently to Co

Sacred Heart Creek). Pollutant concentrations are generally reported only monthly, lea

only

regularly monitored in the discharges, requiring use of approximate assumptions to spe

of nitrate and organic nitrogen. Not surprisingly, this leads to considerable uncertainty

rations downstream of these sources. In Hawk Creek it appears that

estimat

Another factor complicating the nutrient simulation is evident strong interactions betw

algae/macrophytes. HSPF represents planktonic algae and periphyton, but does not ha

ents by macrophytes. In addition, the H

shuts down the simulation of a variety of nutrient and algal processes when stream dept

(to prevent numeric instability). In addition to preventing algal uptake of nutrie

in “stranding” of pollutant mass until depth again increases above the threshold. These

prediction of nutrient concentrations at very low flows, particular

casionally predicts very high concentration values (but l

during low flow conditions due to the stranding phenomenon. This biases statistics bas

rove nutrient fit generally required only small modifications t

setup of the Minnesota River model. One particular area of improvement was in the sp

ly improv

two seasonal patterns, with higher concentrations associated with the areas nearer the

11) and lower concentrations as

at appear to be characterized by more wetlands and ponds (area

). The geochemical reasons for t

The full set of water quality simulation results is provided in Attachment 2. The qualit

ranges from fair to excellent, according to parameter and location, and is consistent wit

n. The largest discrepancies between

appear to be associated with point source nutrient loads and the simulation of nutrients

Overall, the model is appropriately calibrated to support scenario eval

sediment and nutrients. As mentioned above, the model has not been fully calibrated f

2009 period was used as the main bbaasis for calibration, with earlier monitoring providin ng a corroboration

test.

Calibration for nutrients is complica ated by a number of factors, including the presence o of large point

sources at Willmar (discharging to H Hawk Creek) and Southern Minnesota Beet Sugar (d discharged to

County Ditch 37 to West Fork Beav ver Creek through August 2004, subsequently to Cou unty Ditch 45 to

Sacred Heart Creek). Pollutant conccentrations are generally reported only monthly, leadding to

considerable uncertainty in actual lo oading time series. In addition, ammonia is the only form of nitrogen

regularly monitored in the discharge es, requiring use of approximate assumptions to spec cify concentrations

of nitrate and organic nitrogen. Not t surprisingly, this leads to considerable uncertainty in the prediction i

of nutrient concentrations downstreaam of these sources. In Hawk Creek it appears that the total N loading

from Willmar is over-estimated duriing the period around 2005, leading to over-estimati ion of instream N

concentrations, but is much better re epresented in more recent data.

Another factor complicating the nut trient simulation is evident strong interactions betwe een nutrients and

algae/macrophytes. HSPF represen nts planktonic algae and periphyton, but does not hav ve a routine to

explicitly represent uptake and relea ase of nutrients by macrophytes. In addition, the HS SPF model code

shuts down the simulation of a varie ety of nutrient and algal processes when stream depthh falls below 3

inches (to prevent numeric instabilitty). In addition to preventing algal uptake of nutrien nts, this can result

in “stranding” of pollutant mass unt til depth again increases above the threshold. These factors tend to

cause consistent over-prediction of nnutrient concentrations at very low flows, particularlly for stations on

smaller streams. The model occasio onally predicts very high concentration values (but lo ow net loads)

during low flow conditions due to thhe stranding phenomenon. This biases statistics baseed on average

concentration, but does not affect m median error statistics.

Parameter updates to improve nutrie ent fit generally required only small modifications to o the previous

setup of the Minnesota River model l. One particular area of improvement was in the spe ecification of

interflow N concentrations associate ed with tile drainage. Model fit was greatly improve ed by assigning

two seasonal patterns, with higher c concentrations associated with the areas nearer the M Minnesota River

(areas associated with weather statio ons 3 – 7, and 10 – 11) and lower concentrations ass sociated with the

upland moraine areas that appear to o be characterized by more wetlands and ponds (areass associated with

weather stations 1 – 2, 8 – 9, and 12 2 – 14; see Table 3). The geochemical reasons for th his apparent

difference need further investigation n and confirmation.

The full set of water quality simulat tion results is provided in Attachment 2. The quality y of model fit

ranges from fair to excellent, accord ding to parameter and location, and is consistent with h other

applications of HSPF in the Minnes sota River Basin. The largest discrepancies between model and data

appear to be associated with point soource nutrient loads and the simulation of nutrients u under very low

flow conditions. Overall, the modell is appropriately calibrated to support scenario evalu uation for

sediment and nutrients. As mention ned above, the model has not been fully calibrated fo or dissolved

oxygen and bacteria at this time.

19

(This page left intentionally blank.)

20

sount of budget and calendar t

the Hawk Creek Watershed Project. Scenario 1 evaluates the

des to the Willmar wastewater treatment plant, which was a major

eek. Scenario 2 investigates the potential impact of widespread

cultural lands.

Rook a major upgrade to their wastewater tr

potential water quality benefits of the plant upgrade.

moval and nitrification, resulting in order of magnitude decreases in

gen l

s also been a significant decrease in biochemical oxygen demand

moved its discharge point. However, the new disch

t (213).

r quality in the Willmar effluent were provided by the Hawk Creek

with the wastewater treatment plant superintendent. T

61 mg/L

0 mg/L

mg/L

mg/L

57 mg/L

#/100 ml

g/L

ew plant is 4.73 MGD, whereas t

ffluent flow is correlated to weather conditions the scenario was

ather and effluent flow time series, scaling up the effluent flow by a

n added based on the concentration assumptions provided above.

LTURAL, designed to investigate the

the Hawk Creek watershed. Partial implementation would

n current conditions and full implementation.

ed to all NHD streamlines in the watershed. Further, the average

entire watershed are assumed to apply to croplands within each

ed. A GIS analysis indicates that

t of the

y, but not the entirety of public drainage ditches within the watershed.

Model ScenariThis task order included a limited a io analysis. Two

specific scenarios were requested b aluates the

potential impacts of the recent upgr ch was a major

t load to Hawk C widespread

acceptance of stream buffers in agr

ILLMThe City of Willmar recently under lant.

scenario is designed to evaluate the .

included phosphorus r ude decreases in

total phosphorus and ammonia nitr it is mostly

There h gen demand

As part of the upgrade, Willmar als h

falls within the same model segme

upgrade wa he Hawk Creek

Watershed Project after consultatio . T

0.

0.

17

1.

5.

59

2

The projected average flow for the the model for

2009 is 3.55 MGD. Because scenario was

constructed by using the historic w fluent flow by a

llutant loads are th vided above.

GRICThis scenario is a bounding scenari plying 50

opland withi would

approximate a linear scaling betwe

The buffers are assumed to be appl , the average

am density characteristics of th within each

subbasin of the Hawk Creek waters 0 feet of NHD

streamlines accounts for 1.57 perce ination suggests

that the NHD represents the majori in the watershed.

ime to support scena

specific scenarios were requested by the Hawk Creek Watershed Project. Scenario 1 e

potential impacts of the recent upgrades to the Willmar wastewater treatment plant, wh

t load to Hawk Creek. Scenario 2 investigates the potential impact of

eatment

scenario is designed to evaluate the potential water quality benefits of the plant upgrad

included phosphorus removal and nitrification, resulting in order of magni

oad. Nitrogen is not removed, however; rathe

There has also been a significant decrease in biochemical ox

As part of the upgrade, Willmar also moved its discharge point. However, the new dis

upgrade water quality in the Willmar effluent were provided by

Watershed Project after consultation with the wastewater treatment plant superintenden

he average flow i

2009 is 3.55 MGD. Because effluent flow is correlated to weather conditions the

constructed by using the historic weather and effluent flow time series, scaling up the e

llutant loads are then added based on the concentration assumptions pr

impact of a

opland within the Hawk Creek watershed. Partial implementatio

The buffers are assumed to be applied to all NHD streamlines in the watershed. Furthe

am density characteristics of the entire watershed are assumed to apply to croplands

area within

(A cursory exa

that the NHD represents the majority, but not the entirety of public drainage ditches wit

4 Model Scenarioos This task order included a limited ammount of budget and calendar time to support scenar rio analysis. Two

specific scenarios were requested by y the Hawk Creek Watershed Project. Scenario 1 ev valuates the

potential impacts of the recent upgraades to the Willmar wastewater treatment plant, whi ich was a major

source of pollutant load to Hawk Cr reek. Scenario 2 investigates the potential impact of widespread

acceptance of stream buffers in agri icultural lands.

4.1 SCENARIO 1: WILLMAAR TREATMENT PLANT UPGRADES The City of Willmar recently underttook a major upgrade to their wastewater treatment p plant. This

scenario is designed to evaluate the potential water quality benefits of the plant upgrade e.

The upgrade included phosphorus reemoval and nitrification, resulting in order of magnittude decreases in

total phosphorus and ammonia nitro ogen load. Nitrogen is not removed, however; rather r it is mostly

converted to nitrate form. There ha as also been a significant decrease in biochemical oxy ygen demand

(BOD).

As part of the upgrade, Willmar also o moved its discharge point. However, the new disc charge location still

falls within the same model segmen nt (213).

Specifications for post-upgrade watteer quality in the Willmar effluent were provided by tthe Hawk Creek

Watershed Project after consultation n with the wastewater treatment plant superintendentt. The following

assumptions are used:

Total Phosphorus 0.6 661 mg/L

Ammonia-N 0.2 20 mg/L

Nitrate-N 17 mg/L

Organic N 1.6 6 mg/L

TSS 5.8 857 mg/L

Fecal Coliform 59 #/100 ml

BOD5 2 m mg/L

The projected average flow for the n new plant is 4.73 MGD, whereas the average flow in n the model for

1996-2009 is 3.55 MGD. Because eeffluent flow is correlated to weather conditions the scenario was

constructed by using the historic we eather and effluent flow time series, scaling up the ef ffluent flow by a

factor 1.214. Pollutant loads are the en added based on the concentration assumptions pro ovided above.

4.2 SCENARIO 2: AGRICU ULTURAL BUFFER IMPLEMENTATION This scenario is a bounding scenario o, designed to investigate the maximum impact of ap pplying 50-foot

stream buffers to all cropland within n the Hawk Creek watershed. Partial implementationn would

approximate a linear scaling betwee en current conditions and full implementation.

The buffers are assumed to be appli ied to all NHD streamlines in the watershed. Further r, the average

stream density characteristics of the e entire watershed are assumed to apply to croplands within each

subbasin of the Hawk Creek watershhed. A GIS analysis indicates that land area within 550 feet of NHD

streamlines accounts for 1.57 percen nt of the total area in the watershed. (A cursory exam mination suggests

that the NHD represents the majorit ty, but not the entirety of public drainage ditches within the watershed. h

21

rs based on the NHD may be slightly less than could be obtain

age ditches.)

d as vegetative filter strips (VFSs) with perennial grass cover. Thus,

ift in land use with 1.57% of cropland converting

val by buffers in HSPF presents challenges because HSPF is a

as within a subbasin do not have a specific position relative to

idded model

are present in SWAT, also a lumped model. In the recent release of

ped to address this issue, and the same approach is adopted for

.

ite and J.G. Arnold, 2009, Development of a simplistic vegetative

utrient retention at the field scale,

mpirical analyses of field studies and application of the vegetative

ate an approximation of treatment by buffers and filter strips that can

. The approach contain

n BMP efficiency in an agricultural setting, and regression equations

conditional on flow path. The approach also replaces the tradition

rnate measure, the ratio of contributing area to buffer area, which is

ncertain geometry of lumped models.

rformance is obtained when all flow i

he length of the buffer. In contrast, flow that becomes fully

the buffer with little or no pollutant removal. White and Arnold’s

orld applications of buffers occur in situations where a majority of

vely small portion of the buffer. It thus divides the flow from the

eneral loading to the b

that is directed to the most heavily loaded 10 percent of the filter

t is subject to minimal pollutant removal.

nd on the magnitude of flow and the magnitude of sediment loading

by White and Arnold. This is inconvenient to implement in HSPF

ORTRAN code; however, at the ranges of sur

n the Hawk Creek watershed, the impacts on removal rates are

in the range of uncertainty in the regression models. It is thus

tation of treatment efficiency for incorporation into HSPF.

nd Arnold approach is the determination of the different flow

surface runoff that is expected to be fully channelized.

in dispersed sheet flow over a distance of more than 300 feet from the

of 50 feet, would result in a ratio of contributing area to buffer area of

atio of contributing area to buffer area is 62.6. This suggests that 90

fer will be concentrated; however, this may fall within the 10 percent

hannelized. We assu

d, and that the remaining 50% receives some treatment (although less

uffer that receives sheet flow).

ffic

d in the buffer, stranding any particulate pollutants. However, these

zed and transported to the streams during subseq

ften saturated during wet weather events,

ble to settle on fixed (rather than flow

Thus, the impact of instituting buff be obtain

instituting buffers on all public drai

Buffers are assumed to be maintain ass cover. Thus,

the first effect of the scenario is a s grass.

The representation of pollutant rem HSPF is a

lumped model. That is, land use ar relative to

streams, as would be the case in a affect total

pollutant loading. Similar problem recent release of

SWAT2009 an approach was devel adopted for

HSPF representation of this scenari

The SWAT2009 approach (M.J. W stic vegetative

filter strip model for sediment and cesses

lops a method based on the vegetative

filter strip model (VFSMOD) to cr er strips that can

be incorporated into lumped model onceptualization

of the flow paths and their impacts ression equations

that estimate pollutant removal rate es the tradition

reliance on buffer width with an alt r area, which is

more appropriate to the varied and

The best buffer pollutant removal p he buffer as sheet

flow and evenly distributed across es fully

channelized is able to punch throug te and Arnold’s

re a majority of

the field runoff is directed to a relat flow from the

upland area into three categories: w, the fraction of

channelized) concentrated flo nt of the filter

strip, and fully channelized flow th

tionships dep ediment loading

in the regression models developed ment in HSPF

without changes to the underlying noff flow and

surface sediment loading expected l rates are

and also well wit . It is thus

HSPF.

A key factor in applying the White erent flow

especially the fraction o ed.

believed that it is difficult to maint 300 feet from the

which, with a buffer width a to buffer area of

is indicates that suggests that 90

percent of the flow reaching the bu in the 10 percent

focus area and may not all be fully ed to the

concentrated area is fully channeliz ent (although less

effective than in the portion of the

As mentioned above, the treatment or small flows,

be infiltrat However, these

may be remobil t events. Further,

soils in streamside buffer areas are or

infiltration. Therefore, it is reason moval rates.

Thus, the impact of instituting buffers based on the NHD may be slightly less than coul

Buffers are assumed to be maintained as vegetative filter strips (VFSs) with perennial g

t

The representation of pollutant removal by buffers in HSPF presents challenges becaus

lumped model. That is, land use areas within a subbasin do not have a specific positio

, so it is difficult to assess how buffers

pollutant loading. Similar problems are present in SWAT, also a lumped model. In the

SWAT2009 an approach was developed to address this issue, and the same approach is

The SWAT2009 approach (M.J. White and J.G. Arnold, 2009, Development of a simpl

Hydrological Pr

lops a method based on empirical analyses of field studies and application o

filter strip model (VFSMOD) to create an approximation of treatment by buffers and fil

s two major components: a

of the flow paths and their impacts on BMP efficiency in an agricultural setting, and re

that estimate pollutant removal rates conditional on flow path. The approach also repla

reliance on buffer width with an alternate measure, the ratio of contributing area to buff

s directed to

flow and evenly distributed across the length of the buffer. In contrast, flow that beco

channelized is able to punch through the buffer with little or no pollutant removal. Wh

world applications of buffers occur in situations wh

the field runoff is directed to a relatively small portion of the buffer. It thus divides the

uffer without concentrated fl

channelized) concentrated flow that is directed to the most heavily loaded 10 perc

tionships depend on the magnitude of flow and the magnitude of

in the regression models developed by White and Arnold. This is inconvenient to impl

face r

surface sediment loading expected in the Hawk Creek watershed, the impacts on remov

and also well within the range of uncertainty in the regression models

entation of treatment efficiency for incorporation int

A key factor in applying the White and Arnold approach is the determination of the dif

especially the fraction of surface runoff that is expected to be fully channeli

believed that it is difficult to maintain dispersed sheet flow over a distance of more tha

which, with a buffer width of 50 feet, would result in a ratio of contributing ar

is indicates that ratio of contributing area to buffer area is 62.6. This

percent of the flow reaching the buffer will be concentrated; however, this may fall wit

me that 50% of the flow direc

concentrated area is fully channelized, and that the remaining 50% receives some treat

iency of buffers varies with flow loading rate.

be infiltrated in the buffer, stranding any particulate pollutants

ue

reducing

dependent) r

Thus, the impact of instituting buffe ers based on the NHD may be slightly less than couldd be obtained by

instituting buffers on all public drainnage ditches.)

Buffers are assumed to be maintaine ed as vegetative filter strips (VFSs) with perennial grrass cover. Thus,

the first effect of the scenario is a sh hift in land use with 1.57% of cropland converting to o grass.

The representation of pollutant remo oval by buffers in HSPF presents challenges because e HSPF is a

lumped model. That is, land use are eas within a subbasin do not have a specific position n relative to

streams, as would be the case in a ggrridded model, so it is difficult to assess how buffers affect total

pollutant loading. Similar problems s are present in SWAT, also a lumped model. In the recent release of

SWAT2009 an approach was develooped to address this issue, and the same approach is adopted for the

HSPF representation of this scenarioo.

The SWAT2009 approach (M.J. Wh hite and J.G. Arnold, 2009, Development of a simpli istic vegetative

filter strip model for sediment and n nutrient retention at the field scale, Hydrological Pro ocesses, 23: 1602

1616) develops a method based on e empirical analyses of field studies and application of f the vegetative

filter strip model (VFSMOD) to cre e tate an approximation of treatment by buffers and filter strips that can

be incorporated into lumped models s. The approach contains two major components: a c conceptualization

of the flow paths and their impacts o on BMP efficiency in an agricultural setting, and reg gression equations

that estimate pollutant removal ratess conditional on flow path. The approach also replacces the traditional

reliance on buffer width with an alte ernate measure, the ratio of contributing area to buffeer area, which is

more appropriate to the varied and u uncertain geometry of lumped models.

The best buffer pollutant removal peerformance is obtained when all flow is directed to t the buffer as sheet

flow and evenly distributed across t the length of the buffer. In contrast, flow that becom mes fully

channelized is able to punch throughh the buffer with little or no pollutant removal. Whi ite and Arnold’s

approach recognizes that most real-wworld applications of buffers occur in situations whe ere a majority of

the field runoff is directed to a relatiively small portion of the buffer. It thus divides the flow from the

upland area into three categories: ggeneral loading to the buffer without concentrated flo ow, the fraction of

(non-channelized) concentrated flow w that is directed to the most heavily loaded 10 perce ent of the filter

strip, and fully channelized flow tha at is subject to minimal pollutant removal.

Pollutant removal relationships depeend on the magnitude of flow and the magnitude of ssediment loading

in the regression models developed by White and Arnold. This is inconvenient to imple ement in HSPF

without changes to the underlying F FORTRAN code; however, at the ranges of surface ruunoff flow and

surface sediment loading expected i in the Hawk Creek watershed, the impacts on removaal rates are

relatively small – and also well with hin the range of uncertainty in the regression models . It is thus

appropriate to adopt a static represeenntation of treatment efficiency for incorporation into o HSPF.

A key factor in applying the White aand Arnold approach is the determination of the diff ferent flow

fractions – especially the fraction of f surface runoff that is expected to be fully channeliz zed. It is generally

believed that it is difficult to mainta ain dispersed sheet flow over a distance of more than n 300 feet from the

buffer – which, with a buffer width of 50 feet, would result in a ratio of contributing are ea to buffer area of

6. The GIS analysis indicates that r ratio of contributing area to buffer area is 62.6. This suggests that 90

percent of the flow reaching the buf ffer will be concentrated; however, this may fall with hin the 10 percent

focus area and may not all be fully c channelized. We assume that 50% of the flow directted to the

concentrated area is fully channelizeed, and that the remaining 50% receives some treatm ment (although less

effective than in the portion of the b buffer that receives sheet flow).

As mentioned above, the treatment eefficiency of buffers varies with flow loading rate. FFor small flows,

most of the volume may be infiltrateed in the buffer, stranding any particulate pollutants. . However, these

stranded pollutants may be remobili ized and transported to the streams during subsequen nt events. Further,

soils in streamside buffer areas are o often saturated during wet weather events, reducing or eliminating

infiltration. Therefore, it is reasona able to settle on fixed (rather than flow-dependent) re emoval rates.

22

Hawk Creek watershed rates of generation of overland flow and

e low, with much of the flow proceeding through ground water and

ent being generated from scour associated w

ents. We undertook a modeling analysis of the predicted pollutant

White and Arnold under a variety of flows ranging from the median

centile overland surface flow. While removal rates are predicted to

range is generally small (due in part to the assumptions regarding

urther, the majority of pollutant

s calculated at the 95

ulting removal rates relative to the total upland field load generation,

ach employed in SWAT2009, are shown in

Rates for Agricultural Buffer Scenario

al Rates (50erl

heet and

.

ing this scenario is as follows:

the grass land use category (in the corresponding slope a

)

ble to incorporate the reduction rates for pollutant loads associated

ng ravine sediment load) as shown in the table above.

loads. It should be noted, however, that the analysis will not account

t accrue from increasing streambank stability throu

by comparing pollutant loads and concentrations at

also shown at the mouth of Beaver Creek (Scenario 1 does not affect

compared on the basis of medians, as the averages are potentially

imulating concentrations at very low flows.

icted to result in large decreases in both median concentrations and

h the median concentration of nitrate

and permeable nd flow and

overland sediment transport are qui ound water and

tile drainage, and much of the sedi drain outlets and

channel erosion during high flow e icted pollutant

removal rates using the equations o from the median

pe are predicted to

be greater at lower flow depths, the ons regarding

fully channelized or bypass flow). a few large

events. Therefore, the removal rat ow appear

appropriate for the analysis. The r load generation,

nsistent with the appr

Pollutant Remova

Range of Net Remopercentile o

Note that these rates apply only to /gully erosion are

assumed to not be treated by buffer

In sum, the approach for implemen

Shift 1.57 % of cropland to pe a

hydrologic soil group class

LINK t ads associated

with surface runoff (exclud e.

streamside

buffers in reducing upland pollutan s will not account

for any additional benefits that mig h the use of

Results of the scenarios are analyz t

Hawk Creek. Scenario 2 results ar 1 does not affect

ek). Concentrations are re potentially

biased by the model difficulties in

Scenario results are summarized in Willmar

Wastewater Treatment Plant is pre centrations and

althou cted to change.

and permeable Hawk Creek watershed rates of generation of overl

overland sediment transport are quite low, with much of the flow proceeding through g

ith tile

channel erosion during high flow events. We undertook a modeling analysis of the pre

removal rates using the equations of White and Arnold under a variety of flows rangin

percentile overland surface flow. While removal rates

be greater at lower flow depths, the range is generally small (due in part to the assumpt

loads move durin

percentile overland surface f

appropriate for the analysis. The resulting removal rates relative to the total upland fiel

rill erosion; loads generated through ravin

Shift 1.57 % of cropland to the grass land use category (in the corresponding sl

LINK table to incorporate the reduction rates for pollutant l

with surface runoff (excluding ravine sediment load) as shown in the table abo

imate of the potential benefits associated wit

buffers in reducing upland pollutant loads. It should be noted, however, that the analys

Results of the scenarios are analyzed by comparing pollutant loads and concentrations

Hawk Creek. Scenario 2 results are also shown at the mouth of Beaver Creek (Scenari

ek). Concentrations are compared on the basis of medians, as the averages

. In general, the upgrade of the

Wastewater Treatment Plant is predicted to result in large decreases in both median co

In the relatively flat and permeable Hawk Creek watershed rates of generation of overla and flow and

overland sediment transport are quitte low, with much of the flow proceeding through gr round water and

tile drainage, and much of the sedim ment being generated from scour associated with tile drain outlets and

channel erosion during high flow ev vents. We undertook a modeling analysis of the pred dicted pollutant

removal rates using the equations off White and Arnold under a variety of flows ranging g from the median

overland surface flow to the 99th

per rcentile overland surface flow. While removal rates are predicted to

be greater at lower flow depths, the range is generally small (due in part to the assumpti ions regarding

fully channelized or bypass flow). FFurther, the majority of pollutant loads move during g a few large

events. Therefore, the removal rate es calculated at the 95th

percentile overland surface fl low appear

appropriate for the analysis. The reessulting removal rates relative to the total upland fieldd load generation,

calculated consistent with the appro oach employed in SWAT2009, are shown in Table 5..

Table 5. Pollutant Removall Rates for Agricultural Buffer Scenario

Constituent Range of Net Remov 99

th percentile ov

val Rates (50th

to verland flow)

Selected Removal Rate (95th

percentile overland flow)

Sediment 48 – 55 % 51 %

Organic N 38 – 47 % 42 %

Inorganic N 34 – 54 % 43 %

Sorbed and Organic P

43 – 50 % 46 %

Dissolved P 27 – 44 % 35 %

Note that these rates apply only to s sheet and rill erosion; loads generated through ravine e/gully erosion are

assumed to not be treated by buffers s.

In sum, the approach for implement ting this scenario is as follows:

1. Shift 1.57 % of cropland to the grass land use category (in the corresponding slo ope and

hydrologic soil group class. .)

2. Modify the MASS-LINK ta able to incorporate the reduction rates for pollutant lo oads associated

with surface runoff (excludiing ravine sediment load) as shown in the table abov ve.

This should provide a realistic first--cut estimate of the potential benefits associated with h streamside

buffers in reducing upland pollutant t loads. It should be noted, however, that the analysiis will not account

for any additional benefits that migh ht accrue from increasing streambank stability througgh the use of

buffers.

4.3 SCENARIO RESULTS Results of the scenarios are analyzeedd by comparing pollutant loads and concentrations a at the mouth of

Hawk Creek. Scenario 2 results are e also shown at the mouth of Beaver Creek (Scenario o 1 does not affect

Beaver Creek). Concentrations are compared on the basis of medians, as the averages a are potentially

biased by the model difficulties in s simulating concentrations at very low flows.

Scenario results are summarized in Table 6 and Table 7. In general, the upgrade of the Willmar

Wastewater Treatment Plant is pred dicted to result in large decreases in both median con ncentrations and

loads of total P and total N – althouggh the median concentration of nitrate-N is not prediicted to change.

23

icted 11 percent decrease in TSS load in Hawk Creek and a 15

ith smaller fractional losses of N and P. The net effect of the buffers

e to the assumptions regarding the fraction of flow that can be

ed form, as well as the fact that subsurface loads

uffers. The buffers have little impact on media

ce loading during high flow events.

Hawk Creek Mouth (1996

Beaver Creek Mouth (1996

The buffer scenario results in a pre k and a 15

percent decrease in Beaver Creek, ect of the buffers

e d hat can be

concentr cluding tile

are not mitigated by the centrations

because they primarily address surf

Scenario Results,

Scenario Results,

The buffer scenario results in a predicted 11 percent decrease in TSS load in Hawk Cre

percent decrease in Beaver Creek, with smaller fractional losses of N and P. The net ef

e due to the assumptions regarding the fraction of flow

(i

n co

The buffer scenario results in a pred dicted 11 percent decrease in TSS load in Hawk Cree ek and a 15

percent decrease in Beaver Creek, w with smaller fractional losses of N and P. The net efffect of the buffers

is reduced at the watershed scale du ue to the assumptions regarding the fraction of flow t that can be

effectively treated in non-concentraatted form, as well as the fact that subsurface loads (in ncluding tile

drainage) are not mitigated by the b buffers. The buffers have little impact on median con ncentrations

because they primarily address surfaace loading during high flow events.

Table 6. Scenario Results, Hawk Creek Mouth (1996-2009 Simulation)

Baseline Scenario 1 Scenario 2

TSS median concentration (mg/L) 10.80 10.10 10.10

TSS mass (tons/yr) 8927 8921 7901

NOx median concentration (mg/L) 6.30 6.30 6.20

Total N median concentration (mg/L) 13.40 7.20 13.10

Total N mass (tons/yr) 2025 1123 1965

Total P concentration (mg/L) 1.00 0.49 0.96

Total P Mass (tons/yr) 59.9 36.3 58.8

Table 7. Scenario Results, Beaver Creek Mouth (1996-2009 Simulation)

Baseline Scenario 2

TSS median concentration (mg/L) 9.40 8.70

TSS mass (tons/yr) 4203 3571

NOx median concentration (mg/L) 3.40 3.30

Total N median concentration (mg/L) 6.20 6.10

Total N mass (tons/yr) 413 385

Total P concentration (mg/L) 0.45 0.45

Total P Mass (tons/yr) 14.2 13.5

24

tetra tech, inc

Documents