Assessment of sediment loads for a Great Lakes mixed ... · Buffalo River Watersheds Cazenovia Creek Buffalo Creek Cayuga Creek sediment . Key Lessons 6. SWAT cannot simulate ice-scour

Assessment of sediment loads

for a Great Lakes mixed-landuse

watershed

Lessons from field data & modeling

Shreeram Inamdar & Anthony Friona SUNY Buffalo State College

Motivation

1. Determine the sediment yield at the

watershed outlet (Tons/ha/yr or mg/L)

USCOE – dredging of navigable portion of Buffalo

River

2. Identify source catchments and reaches

USDA-NRCS – BMP implementation

TMDL plan for the watershed

Funded by Great Lakes Commission & COE Contract grant

Model selection

Model:

• SWAT – Soil Water Assessment Tool

– Spatially-distributed continuous simulation model

– EPA promoted public domain GIS model

– Simulate watershed conditions by providing appropriate data and parameter values

– Daily time step

– Sediment, nutrients, pesticides, organics, biological variables

Watershed description

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Buffalo

Active USGS

discharge gage

142 mi2

96 mi2

424 mi2

136 mi2

Model implementation

1

8

5

7

23

9

3

15

2

21

4

22

24

19

20

11

25

18

1214

6

13

17

10

16

26

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Model implemented

first for

Cazenovia

Once calibrated,

values

will be extended

for the full

Buffalo watershed

Model – GIS layers

10 DEM &

NHD drainage

LULC – 2002 DOQs

STASGO

SSURGO

Sediment data collection

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

discharge

sediment

Sediment data collection

• Turbidity recorded as a surrogate

for suspended sediment

• Continuously logging (15 minutes)

data sondes

• Grab samples of sediment collected

to generate sediment-turbidity

relationship

• Provides – important event sediment

dynamics

Sediment data collection

Turbidity versus sediment

relationship

Model calibration -data

Period – 1997 to 2003 (daily time step)

• Comparison of simulated versus observed time series –

• Annual

• Seasonal

• Daily

• Frequency analysis

Model calibration methods

1. Monte Carlo simulations – 1000 model runs – to identify sensitive

parameters – dotty plots

2. Once parameter values were constrained – ArcView based model

runs

Discharge results

Annual totals – within 10%

Discharge Results

Daily discharge comparisons

- Visual

- fits (Nash Sutcliffee, bias, …)

Discharge Results

Frequency analysis

Sediment Calibrations

Sediment calibration:

1. Much more difficult that discharge

2. First objective - constrain model predictions within the same order

of magnitude as observed

Sediment Results

Daily sediment comparisons

Sediment Results

Frequency analysis:

20

50

Sediment Yields

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Buffalo

0.8 t/ha/yr (96-2003)

Sediment Source Areas

Subcatchment sediment conc. in mg/L Ag land

slope

Key Lessons

1. Monte-Carlo simulations were very important – narrowed us down to

few key parameters and ranges

Key Lessons

2. Recommended (default) C & P values generated very high sediment

yields!

3. C & P values had to be reduced considerably to match observed

sediment yields (simulation based on LULC)

Key Lessons

4. When only “active” cropland was used, C & P values were not

reduced!

Resolution and accuracy of LULC extremely important!

LULC – 2002 DOQs Active cropland - NRCS

Key Lessons

5. Sediment calibrations against a single observed station at

watershed outlet may not guarantee correct values for internal

nodes.

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creek

Buffalo River Watersheds

Cazenovia

Creek Buffalo Creek

Cayuga Creeksediment

Key Lessons

6. SWAT cannot simulate ice-scour bank erosion – could be an

important contributor for Great Lakes tributaries!

Ice scour

Conclusions

1. SWAT provides a tool to generate yields, however realistic

predictions are only generated after considerable calibration.

2. Multiple measurements (in time and space) and data quality are

very critical.

Questions?