Master’s Degree Thesis SeminarAgricultural and Biological Engineering

Sarah RutkowskiMay 11th, 2012

Role of Climate Variability on Subsurface Drainage and Streamflow Patterns in

Agricultural Watersheds

Ale and Bowling (2010)

Tile Drainage

• Subsurface (tile) drains lower high water table levels in poorly drained soils.

•Large portions of drained agricultural land in the Midwest were once wetlands (Du et al., 2005)

• Numerous impacts on water quality and hydrology

Water Quality Impairments

Nitrate losses at the field scale have been measured by Kladivko et al. (2004): Annual nitrate losses ranged from

18 to 37 kilograms per hectare Best Management Practices to

reduce nutrient pollution: cover crops, drainage water management, grassed waterways

• Tile drains facilitate the transport of nutrients to surface water

Iowa Natural Resource Conservation Service (2008)

Hydrologic Changes in Tile Drained Watersheds

Influences flashiness and flow variability: Streamflow recession occurs more

rapidly as tile drainage extent increases (Ale et al., 2010)

Alters low flow and peak flow Increasing low flow and decreasing

peak flows as the extent of tile drainage increases (Schilling and Libra, 2003)

Kumar et al. (2009) found increasing trends in low, median, and high flow metrics in Indiana. Precipitation highly influences these trends.

Ale and Bowling (2010)

Streamflow recession as influenced by potential tile drained area.

Climate Change Effects on Streamflow Hydrology and Water Quality

Precipitation and soil moisture will increase in the winter and spring and decrease in the summer (Wuebbles and Hayhoe, 2004)

Water quality issues arise in the spring, prior to planting: Timing of fertilizer application

Conservation practices such as Drainage Water Management conserve water during dry seasons

Tools Available to Estimate Tile Drainage and Water Quality Impacts

Hydrology models are available at varying spatial scales Evaluate non-point source pollution and hydrologic effects from

drainage

Provide drainage volume estimates which can aid decisions made regarding water, nutrients, crop management, etc.

Field Scale Models: DRAINMOD Root Zone Water Quality Model (RZWQM)

Watershed Scale Models: Soil and Water Assessment Tool (SWAT)

Variable Infiltration Capacity Model Many large scale hydrology models still lack a

component for climate change analysis in tile drained river basins Variable Infiltration Capacity (VIC) Model has been used

for regional and continental climate change studies

Motivation for updates: Capable of simulating tile drainage for areas greater

than just a single watershed Could be used to potentially model drainage for the

entire Midwest Quantify drainflow and nitrates to estimate impact to the

Hypoxic Zone in the Gulf of Mexico

VIC Model Processes

Divides study area into grid cells Multiple soil layers (usually 3) Vegetation scheme which varies

sub-grid. Driven by meteorological data

(precipitation, wind-speed, temperature)

(Cherkauer et al., 2003)

Calculates lateral flow from the bottom soil layer using

the baseflow curve

Model Equations for Drainflow Addition

Arno Baseflow Equation by Todini (1996): Baseflow curve is divided into linear and non-linear baseflow response and is defined by three parameters (WS, Ds, and DSmax) Maximum baseflow out of the bottom soil layer, DSmax Baseflow shape changes at the soil moisture threshold, WS Fraction of the maximum baseflow where response shifts, Ds

Ellipse Equation: Used to adjust the Arno equation to calculate subsurface drainage Used to solve for new maximum baseflow (DSmax) out of the

bottom soil layer

Original and Modified Baseflow Curve

• New baseflow parameters are calculated based on the original, user defined values

•WS’ is calculated first based on drain depth, followed by Ds’

• Maximum baseflow rate DSmax is calculated last using Ds’ and WS’ and the ellipse equation

Equilibrium Soil moisture value when water table first rises above the drain depth

Calibration: Study Site Data and Setup

Southeast Purdue Agricultural

Center (SEPAC)

Located inButlerville,

Indiana

Drain depths at 0.75 meters

Observations from plots with drains

spaced at distances of 20, 10, and 5 meters

West Block

East Block

Water Table and Drainflow Data

Kladivko et al. (2003)

Model Input Data

Meteorological forcing file: hourly precipitation, temperature, and wind speed from the SEPAC weather station (Naz, 2006)

Soil physical properties from measurements at SEPAC (Kladivko, 1999 )

Vegetation properties: leaf area indices, root depths: Land Use History of North America (LUHNA) by Cole et al., 1998

Calibration Parameters:Baseflow: Ds and DSmaxWater Table: Brook’s and Corey Water Retention Curve EXP and

soil bubbling pressure: BUBBLESoil Infiltration Parameter: Bi

Model Sensitivity

“One at a Time” sensitivity analysis Relative Sensitivity of each parameter:

y= predicted drainage output

x = base parameter value

xhigh and xlow correspond to the high and low parameter values

yhigh and ylow are the corresponding response variable values at the high and low parameter values

Relative Sensitivity of Calibration Parameters

Calibration Methods

Calibrated using drainage and average water table data between the 1988 and 1990 water years

Simulate observed data from study site using a single grid cell

Compare output to average water table and drainage from the West 20 meter Plot Nash-Sutcliffe Efficiency (NSE) Percent Error (PE) Coefficient of Determination (R2)

Validated the model using drainage data Water table measurements were not collected after 1990

Calibration ResultsDrainage Water TableCalibration Period:

1988 to1990 Water Years

__ Simulated Data__ Observed Data __ Depth of Tile Drain (0.75 meters)

Drainage StatisticsNS = .34PE = 2.10% R2 = .34

Water Table StatisticsNS= -.08 PE = -22.7 R2 = .26

Drainage Efficiency

Conclusions from Field Scale Analysis

VIC model is primarily used for large scale analyses

Water table depths calculated by the model were not as dynamic as the observed data

Evidence of preferential flow in observed data could also account for lower model efficiency

The drainage model predicts total drainage within 21 % of the data between the water years of 1988 and 1994 Reasonably predicts drainage depths suitable for

watershed scale tests

Objectives and Hypotheses

2.) Water conservation from DWM during the growing season will decrease under future climate conditions from the levels seen now.

1.) Indiana watersheds have experienced higher annual low flows due to increased water storage capacity in the soil from conventional subsurface drainage.

Watershed Scale Study

The White River watershed extends across the majority of central Indiana: Delineated upstream of Indianapolis to

avoid urban influence

Hypothesis #1

Model Setup: Creating Input Files using Spatial Data

Hypothesis #1

NASS Cropland Data Layer

Watershed Boundary and VIC grid cells

Potentially Tile Drained Land (Ale, 2009)

Indiana Drainage Guide Recommendations

Methods

Model simulations from 1930 through 2005 water years

Two Model Scenarios Drainage Algorithm ON (Calibration Case) Drainage Algorithm OFF

Most recent 20 years used for calibration and validation

Preliminary soil parameters and constants were taken directly from the field scale calibration Ds and DSmax were changed during calibration for grid cells

containing less than 50% tile drainage.Hypothesis #1

Metrics

Hypothesis #1

The following metrics were used to compare the predicted streamflow from each model

simulation.

• Low Flow: Seven-Day Minimum (Low) Flow• High Flow: Seven-Day Maximum (Peak) Flow• Mean Flow: Mean Annual Flow (MAF)

• Streamflow Variability:• Richards-Baker Flashiness Index (RBI)

Calibration and Validation Results

Hypothesis #1

CalibrationStatistics:NSE = .45 PE= -11.5 % R2= .59

Validation Statistics: NSE = .70 PE= - 12.8 % R2= .75

Legend:

Model ___Observed ___

Hypothesis #1

Comparing Hydrographs between the Model Simulations

Model ___ Observed ___

Compare Flow Metrics Between the Drainage and No Drainage

Model Simulations

Hypothesis #1

Mean Annual FlowSeven Day Minimum Flow

Seven Day Peak Flow RBI

Conclusions and Answer to First Hypothesis

Hypothesis: Tile drainage systems have increased annual low flow due and streamflow

flashiness

Conclusions: Streamflow flashiness is higher in drained

conditions Peak flows are larger while low flows are

reduced

Climate Variability Effects: Observed Trends in the White River Watershed

Seven Day Low FlowAverage Annual Precipitation per 15

Year Time Period

• Overall increasing trend in all flow metrics largely due to precipitation• How will precipitation and temperature continue to affect tile drained landscapes in the future?

GFDL Model Emissions Scenarios:

A2: High Emissions: Best representation of our current GHG trajectory A1B: Mid to High Emissions: Technological advances will limit some GHG B1: Conservative Emissions: Future climate with many technological advances

Future Climate Projections

Average Annual Precipitation Relative to Historic Period from 1980-2009

Average Annual Temperature Relative to Historic Period from 1980-2009

Future Time PeriodsFuture Time Periods

Interaction of Conservation Practices with Climate Change

Conservation practices such as drainage water management could be used to mitigate seasonal variability from climate change Drainage Water Management (DWM) controls the

water table height and level of drainage seasonally

How will water conserved by DWM change in future climate conditions? Hypothesis 2: Water conservation will decrease

during the growing season in future climates

Hypothesis #2

DWM and Conventional Drainage Model Setup

The VIC model was also modified to handle monthly changes in drain depth from DWM Mimics the effect of raising and lowering the DWM control structure

Two model setups both forced with future climate data for all 3 emissions scenarios: Agriculturally drained land is using Drainage Water Management

(DWM) Conventional Tile Drainage (using the previous model setup)

Three control heights are used for DWM Case: Winter: 0.3 meters April and September: 0.9 meters Summer (Growing Season): 0.6 meters

Hypothesis #2

Field and Watershed Scale Hydrographs

Hypothesis #2

Grid Cell Flow at Watershed Outlet

Legend:

DWM _Conventiona

l _

Examine how factors (DWM and future climate) affect streamflow metrics

DWM increases low flow in historic and future climate conditions

Climate change has a greater impact on streamflow metrics than DWM

Factor Separations

Hypothesis #2

1980-2009 2070-20990.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

Moderate Emissions Model Seven Day Minimum Flow

Time Period

Flo

w (

m3/

s)

1980-2009 2070-20990.0

50.0

100.0

150.0

200.0

250.0

300.0

350.0

Moderate Emissions Model Seven Day Peak Flow

Conventional Drainage

Drainage Water Management

Time Period

Flo

w (

cms)

Annual Water Conservation Broken into 30 Year Time Periods

Hypothesis #2

Water Conserved in the soil column is difference between the streamflow from the DWM and conventional drainage simulations.

Differences in flow equals the amount of water that remains in the soil column or used as evapotranspiration

Net increase in water conservation throughout the 21st century

Water Conserved by DWM during the Growing Season

• Decreasing water conservation during the growing season in future time periods

• Similar trend in growing season

•Evapotranspiration trends are similar: less water availability in the future

•DWM case predicts higher ET than conventional drainage

Growing Season Water Conservation

Hypothesis #2

Conclusions and Answer to Second Hypothesis

Hypothesis: Growing season water conservation will decrease throughout the next century

• Conclusions: • Dry summers will decrease water availability, less water

to conserve. Growing season ET is also decreasing.

• DWM is very effective in the Spring months at maintaining high water table levels.

• DWM will become more efficient as precipitation totals increase

Overall Conclusions

Hypothesis 2: Growing season water conservation will decrease in future climates DWM is more effective at holding soil moisture during the growing season

and will be a valuable practice in future climates Water Conservation will increase during the spring and periods of high

precipitation and decrease throughout the next century during dry seasons

Hypothesis 1: Subsurface drainage has increased low flows and decreased streamflow flashiness The model proved that streamflow flashiness is increasing and low

flows are reduced There are increasing low flow trends that are likely due to

precipitation

Project Improvements

Improve field scale calibration using a more stationary dataset Reassess whether the correct parameter values were

chosen

Select a global climate model that more accurately represents future streamflow in the Upper White River watershed Decreasing low flow trends in the GFDL model in opposition

of the observed data GFDL model was chosen because it has been used for

studies in the Midwest

The End

Thank you for watching my presentation!

Acknowledgments:

I’d like to thank my advisors Drs. Keith Cherkauer and Laura Bowling

Dr. Eileen Kladivko for providing me with data from SEPAC

Srinivasulu Ale for sparking the idea for this research endeavor

My friends and family for their support!

Simulated Water Table: Validation

Validation Hydrographs

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Drainflow Flashiness

Water Year

Fla

sh

ine

ss

In

de

x (

RB

I)

Monthly Conservation Cycle• Negative Water Conservation:

- Considerable losses in April due to lowering DWM boards for Spring Planting

- Less noticeable losses in the late growing season (July through August)

• Water Conservation during seasons of higher precipitation (December through March)

Hypothesis #2