c© 2011 Phong Vu Viet Le

EXPANSION OF BIOENERGY CROPS IN THE MIDWESTERN UNITED STATES:IMPLICATIONS FOR THE HYDROLOGIC CYCLE UNDER CLIMATE CHANGE

BY

PHONG VU VIET LE

THESIS

Submitted in partial fulfillment of the requirementsfor the degree of Master of Science in Civil Engineering

in the Graduate College of theUniversity of Illinois at Urbana-Champaign, 2011

Urbana, Illinois

Adviser:

Professor Praveen Kumar

ABSTRACT

To meet emerging bioenergy demands, significant areas of the large-scale agricultural land-

scape of the Midwestern United States could be converted to 2nd generation bioenergy crops

such as miscanthus and switchgrass. Their high biomass productivity in a longer growing

season linked tightly to water use highlight the potential for significant impact on the hydro-

logic cycle in the region. This issue is further exacerbated by the uncertainty in the response

of the vegetation under elevated CO2 and temperature. This study attempts to capture

the eco-physiological acclimations of bioenergy crops under climate change and predict how

hydrologic fluxes are likely to be altered from their current magnitudes. A mechanistic mul-

tilayer canopy-root-soil model is applied for capturing the emergent vegetation responses to

environmental change. Observed data and Monte Carlo simulations of weather for recent

past and future scenarios are used to characterize the variability range of the predictions.

We found that, under present weather conditions, miscanthus and switchgrass utilized

more water than maize for total seasonal evapotranspiration by approximately 58% and

36%, respectively. Projected higher concentration of atmospheric CO2 (550 ppm) is likely to

decrease water used for evapotranspiration of miscanthus, switchgrass, and maize by 12%,

10%, and 11%, respectively. However, when climate change with projected increases in air

temperature and reduced summer rainfall are also considered, there is a net increase in

evapotranspiration for all crops, leading to significant reduction in soil-moisture storage and

specific surface runoff. These results highlight the critical role of the warming climate in

potentially altering the water cycle in the region under extensive conversion of existing maize

cropping to support bioenergy demand.

ii

...to my family

iii

ACKNOWLEDGMENTS

I would first like to thank my advisor, Prof. Praveen Kumar, whose valuable thoughts and

vision have made this thesis possible. His guidance and encouragement are also a powerful

inspiration to me. I would like to thank Dr. Darren Drewry for his advices and comments

on this thesis. I am also thankful to our team members, Prof. Murugesu Sivapalan, Sayo

Chaoka, Ciaran Harman, Hongyi Li, Juan Quijano, Venkatraman Srinivasan, Sheng Ye, and

Mary Yeager for their invaluable discussions and comments. I appreciate the support of

Prof. Carl Bernacchi and Andy VanLoocke for providing me an opportunity to gain field

experience at the Energy Farm. Thanks to the wonderful people in Hydrosystem Lab for

their friendship and supports. Finally, thanks to my family members for all their endless

love and encouragement. This thesis is dedicated to them.

This work is supported by Vietnam Education Foundation (VEF) and National Science

Foundation (grant ATM 06-28687).

iv

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

LIST OF ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

LIST OF SYMBOLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

CHAPTER 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

CHAPTER 2 METHODS AND DATA DESCRIPTIONS . . . . . . . . . . . . . . . 52.1 Multi-layer canopy root soil model . . . . . . . . . . . . . . . . . . . . . . . . 52.2 Weather Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.3 Data sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.4 Climate change projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.5 Water balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.6 Calculation of albedo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

CHAPTER 3 GRAPHIC USER INTERFACE FOR MLCAN . . . . . . . . . . . . . 153.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153.2 Graphic User Interface for MLCan . . . . . . . . . . . . . . . . . . . . . . . 163.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

CHAPTER 4 RESULTS AND DISCUSSIONS . . . . . . . . . . . . . . . . . . . . . 274.1 MLCan model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274.2 Weather Generator validation . . . . . . . . . . . . . . . . . . . . . . . . . . 304.3 Within-canopy vertical variation . . . . . . . . . . . . . . . . . . . . . . . . . 364.4 Impacts of crop type on energy balance . . . . . . . . . . . . . . . . . . . . . 384.5 Vegetation response to elevated CO2 and increased air temperature . . . . . 404.6 Sensitivity to Seasonal Variation in Photosynthetic Capacity . . . . . . . . . 414.7 Impacts on hydrology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

CHAPTER 5 CONCLUSIONS AND FUTURE WORKS . . . . . . . . . . . . . . . 49

CHAPTER 6 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

v

APPENDIX A WEATHER GENERATOR . . . . . . . . . . . . . . . . . . . . . . . 57A.1 Precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60A.2 Cloud cover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61A.3 Shortwave radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63A.4 Air temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67A.5 Humidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68A.6 Wind speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

APPENDIX B SIMULATIONS FOR OTHER LOCATIONS IN MIDWESTERNUNITED STATES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71B.1 Ozark Forest, Missouri . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74B.2 Brookings, South Dakota . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

vi

LIST OF TABLES

2.1 Projected climate change scenarios for Midwestern U.S. for 2050 during the

summer for driving MLCan model predictions [33] . . . . . . . . . . . . . . . . . 13

4.1 Value of model parameters for maize, miscanthus, and switchgrass used in the

multilayer canopy-root-soil model (MLCan) . . . . . . . . . . . . . . . . . . . . 294.2 Parameters used in the Weather Generator. . . . . . . . . . . . . . . . . . . . . 314.3 Comparison of total evapotranspiration (ET) and its component contribution

maize, miscanthus, and switchgrass under present climate condition in 2005 . . . 384.4 Comparisons of total evapotranspiration (ET) alterations (mm) under climate

change between maize, miscanthus, and switchgrass . . . . . . . . . . . . . . . . 414.5 Change of total evapotranspiration (ET ) and specific surface runoff (R) for the

two cases shown in Figure 4.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

A.1 Parameters for precipitation model . . . . . . . . . . . . . . . . . . . . . . . . . 57A.2 Parameters of the wind speed model . . . . . . . . . . . . . . . . . . . . . . . . 58A.3 Parameters for cloud cover, air temperature, and humidity model . . . . . . . . . 59

B.1 Comparison of total evapotranspiration (ET) and its component contribution

maize, miscanthus, and switchgrass under present climate condition in 2005 at

three locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

vii

LIST OF FIGURES

2.1 Schematic of the canopy, root, and soil system model . . . . . . . . . . . . . . . 62.2 Leaf-level model component coupling . . . . . . . . . . . . . . . . . . . . . . . 82.3 Application of Weather Generator for developing a forcing ensemble in Bondville, IL 92.4 Ameriflux tower at Bondville, Illinois, USA. (A) Map of Ameriflux tower located

at Bondville, Central Illinois; (B) 10-meter eddy covariance flux tower in Bondville 102.5 Key meteorological forcing data observed in 2005 overlaid on an ensemble ob-

tained using the stochastic weather generator . . . . . . . . . . . . . . . . . . . 112.6 Normalized canopy leaf area density profiles (A) and normalized root fraction

in each soil layer (B) for miscanthus and switchgrass . . . . . . . . . . . . . . . 12

3.1 Main window of MLCan graphic user interface. The GUI is organized by func-

tion and task on the left. Corresponding instructions for each sub-components

are presented on the right of the main window . . . . . . . . . . . . . . . . . . . 173.2 View of Model Setup in MLCanFace. The Model Setup is comprised of 3 subsec-

tions: (i) Specifying site information; (ii) Root structure; (iii) Canopy structure.

The root and canopy structure windows are graphically supported . . . . . . . . 193.3 View of model Options in MLCanFace . . . . . . . . . . . . . . . . . . . . . . . 203.4 View of model Forcing and Initial Conditions in MLCanFace . . . . . . . . . . . 213.5 View of Leaf & Canopy (A) and Soil & Radiation (B) parameters in MLCanFace . 233.6 View of Photosynthesis (A) and Conductance & Respiration (B) parameters in

MLCanFace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.7 Result viewer in MLCanFace . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4.1 Comparison of half-hourly net photosynthetic leaf CO2 uptake (An) of upper

canopy sunlit leaves for miscanthus and switchgrass modeled by MLCan and

observed data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284.2 Observed and simulated mean monthly rainfall at Bondville, Illinois . . . . . . . 304.3 Observed and simulated cloud cover distribution at Bondville, Illinois . . . . . . . 334.4 Observed and simulated daily cycles of air temperature in 12 months at Bondville,

Illinois . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.5 Diurnally averaged observed and simulated air temperature over 10 years (1997-

2006) at Bondville, Illinois . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354.6 The histogram of hourly wind speed from the observed and simulated data over

10 years (1997-2006) at Bondville, Illinois . . . . . . . . . . . . . . . . . . . . . 35

viii

4.7 Diurnally averaged profiles of photosynthetic rate An; latent heat LE; sensi-

ble heat H; total absorbed shortwave included photosynthetically active and

near-infrared bands Qabs; and stomatal conductance for vapor gsv for maize,

miscanthus and switchgrass under present climate condition in August 2005 . . . 374.8 Diurnal variation of mean canopy temperatures (A), temperature difference be-

tween the top layers and mean canopy (B), and albedo during the day (C) for

maize; miscanthus; and switchgrass in August 2005. Diurnally averaged change

of net-canopy fluxes and variables obtained from the MLCan model with ver-

tical bars representing ± SD over growing season of Photosynthetic rate ∆An(D); Latent heat ∆LE (E); Sensible heat ∆H (F); Leaf temperature ∆Tl (G);

Stomatal conductance for vapor ∆gsv (H) for miscanthus and switchgrass . . . . . 394.9 Diurnally averaged net-canopy fluxes and variables obtained from MLCan model

with vertical bars representing ± one SD over one growing season of (A) Pho-

tosynthetic rate An; (B) Latent heat LE; (C) Sensible heat H; (D) Leaf tem-

perature Tl; and (E) Stomatal conductance for vapor gsv for miscanthus and

switchgrass under present climate condition in 2005 . . . . . . . . . . . . . . . . 424.10 To understand the impact of seasonality of leaf photosynthetic capacity on the

results presented earlier, Vmax for maize was varied as shown here . . . . . . . . . 434.11 Diurnally averaged change of net-canopy fluxes and variables obtained for maize

during 2005 using the MLCan model with vertical bars representing ± one SD

over one growing season of (A) Photosynthetic rate An; (B) Latent heat LE;

(C) Sensible heat H; (D) Leaf temperature Tl; and (E) Stomatal conductance

for vapor gsv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444.12 Weekly mean water balance, soil water storage change, and total specific surface

runoff for 3 crops in several weeks in growing season 2005 under present CO2

condition (370 ppm) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.13 Ensemble of total evapotranspiration (A), total soil water storage change (B),

and total specific surface runoff (C) for maize (green), miscanthus (red), and

switchgrass (blue) under present climate and eight combinations of climate

change scenarios involving expected CO2 concentration, higher air tempera-

ture, and altered precipitation magnitudes [33] . . . . . . . . . . . . . . . . . . 48

A.1 Flowchart of the dependence of variables in the Weather Generator. . . . . . . . . 58

B.1 Map of Ameriflux towers located Bondville, Illinois (red point); Ozark, Missouri

(blue point); and Brooking, South Dakota (green point) in Midwestern U.S. . . . 72B.2 Key meteorological forcing data observed in 2005 at Ameriflux tower located in

Ozark, Missouri. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74B.3 Diurnally averaged profiles of photosynthetic rate An; latent heat LE; sensi-

ble heat H; total absorbed shortwave included photosynthetically active and

near-infrared bands Qabs; and stomatal conductance for vapor gsv for maize,

miscanthus and switchgrass in Ozark, Missouri under present climate condition

in August 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75B.4 Key meteorological forcing data observed in 2005 at Ameriflux tower located in

Brookings, South Dakota. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

ix

B.5 Diurnally averaged profiles of photosynthetic rate An; latent heat LE; sensi-

ble heat H; total absorbed shortwave included photosynthetically active and

near-infrared bands Qabs; and stomatal conductance for vapor gsv for maize,

miscanthus and switchgrass in Brookings, South Dakota under present climate

condition in August 2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

x

LIST OF ABBREVIATIONS

ET Evapotranspiration

FACE Free Air CO2 Enrichment

GUI Graphical User Interface

HR Hydraulic Redistribution

LAD Leaf Area Density

LAI Leaf Area Index

MLCan Multilayer canopy-root-soil model

NIR Near-infrared Radiation

PAR Photosynthetically Active Radiation

RHC Root Hydraulic Conductivity

RuBP Ribulose-1,5-bisphosphate

Rubisco RuBP carboxylase-oxygenase

SD Standard Deviation

U.S. United States

UV Ultraviolet band

VIS Visible band

WG Weather Generator

xi

LIST OF SYMBOLS

Roman

a, b Shape parameters of beta distribution

An [µmol·m−2·s−1] Photosynthetic leaf CO2 uptake

b [mol·m−2·s−1] Stomatal intercept parameter in Ball Berry model

bi (i = 0, 1, ...4) regression coefficients for deterministic hourly

temperature increment

C [mm] Condensation on the foliage

Ca [ppm] Atmospheric CO2 concentration

Ci [µmol·mol−1] Internal CO2 concentration

do [m] Leaf width

E [mm] Evaporation from the canopy

ET [mm] Total evapotranspiration

ea [kPa] Vapor pressure

fsunlit Sunlit fraction

fshaded Shaded fraction

gbh [mol·m−2·s−1] Boundary layer conductance for heat

gbv [mol·m−2·s−1] Boundary layer conductance for vapor

gs [mol·m−2·s−1] Stomatal conductance

gsv [mol·m−2·s−1] Stomatal conductance for vapor

H [W·m−2] Sensible heat

h [mm] Storm depth

hcan [m] Canopy height

Jmax,25 [µmol·m−2·s−1] maximum rate of electron transport at 25◦C

Kd Diffuse extinction coefficient

xii

k4 [mol·m−2·s−1] Initial slope of C4 photosynthetic CO2 response

kn Decay coefficient for leaf nitrogen content

LE [W·m−2] Latent heat

LWnet [W·m−2] Net longwave energy

LWP0 [g·m−2] Cloud total liquid water path for overcast conditions

M0 Mean of fair weather cloudiness

m Stomatal slope parameter in Ball Berry model

N(t) Cloud cover

P ∗ann [mm] Precipitation index in dew temperature model

PPT [mm] Precipitation

R [mm] Total specific surface runoff (per unit area)

Rd [µmol·m−2·s−1] Reference value fo leaf respiration

Rg [MJ·m−2·day−1] Global radiation

Rlw [MJ·m−2·day−1] Incident longwave radiation from atmosphere

Q10,4 Temperature sensitivity of temperature-dependent C4 pa-

rameters

Q(abs) [W·m−2] Total absorbed shortwave radiation

S [mm] Soil water storage

SE [mm] Soil evaporation

SW [W·m−2] Shortwave radiation

So [W·m−2] Solar constant

Sp [mm] Seepage

sf [MPa−1] Stomatal sensitivity parameter

Ta [◦C] Air temperature

Tl [◦C] Leaf temperature

TR [mm] Transpiration

tb [hour] Time between two successive storms

tr [hour] Storm duration

U [m·s−1] Wind speed

uo [cm] Ozone amount in a vertical column

Vmax [µmol·m−2·s−1] Reference value for substrate saturated Rubisco capacity

xiii

Ws [m·s−1] Mean wind speed

x Leaf angle distribution parameter

z [m] Vertical coordinate

Greek

α Albedo

α4 [mol·mol−1] Intrinsic quantum yield C4 photosynthesis

αL−PAR Leaf absorptivity to photosynthetic active radiation

αL−NIR Leaf absorptivity to near-infrared

β Angstrom turbidity parameter

δT [◦C] Mean of random temperature deviates

εv Leaf emissivity

γs(t) Skewness of wind speed

µb [hour] Mean time between storms

µd [mm] Mean storm depth

µr [hour] Mean storm duration

ωAi Aerosol single-scattering albedos (VIS and NIR band)

Ψf [MPa] Ψl at which half potential gs is lost

Ψl [MPa] Leaf water potential

ρg Spatial average regional albedo

ρm Lag-1 autocorrelation of fair weather cloudiness

ρs(1) Lag-1 autocorrelation of wind speed

ρδT Lag-1 autocorrelation of random temperature deviates

σ2m Variance of fair weather cloudiness

σ2δT [◦C] Variance of random temperature deviates

σ2s [m2·s−2] Variance of wind speed

τ [hour] Time lag

θ [m3·m−3] Soil moisture

ς, γ [hour−1] Cloudiness decay rate from the end of rainfall to fairweather

xiv

CHAPTER 1

INTRODUCTION

Rapidly growing energy demand, worldwide depletion of fossil fuels, and global warming are

raising an interest in expanding clean and renewable bioenergy production. In the United

States (U.S.), the current starch-based bioethanol production only contributes a small por-

tion of total energy needs [1, 2], but it is raising new challenges related to environmental

issues [3–6] and a competition with food production on available fertile land [7]. Bioenergy

extracted from lignocellulosic feedstocks offers the possible use of marginal land [8], along

with many energy, environmental, and economic advantages over current biofuel sources [9],

and is being considered as a promising alternative to sustainably meet the U.S. Department

of Energy target for bioenenergy and biobased products in the future [10]. At present, mis-

canthus (Miscanthus × giganteus) and switchgrass (Panicum virgatum) are considered as

the two perennial grasses with the highest potential for lignocellulosic bioenergy production

in the Midwest with high biofuels yield per unit land area, reduced requirement of nutrient

inputs [11, 12], and low net CO2 emissions [13–16]. However, if large portions of the land-

scape in the Midwestern U.S. are converted to these crops for meeting bioenergy demands,

for example, by using land that supports maize (Zea mays L.) production, it is likely to

significantly impact the hydrologic cycle.

A number of studies have been conducted to compare the water use associated with

bioenergy crop production in the Midwest. Much of this work has estimated that the total

evapotranspiration (ET) of miscanthus and switchgrass is higher relative to that of maize

using methods such as the residual energy balance method [17], water budget estimation [18],

and model-based approaches [19]. Each of these studies highlighted the role of higher leaf

area index (LAI) and longer growing season as the primary reason for the increase, but

estimates of water use increase vary considerably. For instance, Hickman et al. [17] estimated

1

that miscanthus and switchgrass increase total growing season ET by 343 and 153 mm

relative to maize, respectively, while McIsaac et al. [18] showed that miscanthus increases

total ET by 104 mm relative to maize, with switchgrass and maize having comparable

total ET.

The present study evaluates potential impacts of biofuel-based land use changes on the

hydrologic cycle through simultaneous considerations of (i) above-ground canopy structure

and function as a result of changes in crop type, and (ii) vegetation response to climate

change as manifested through elevated atmospheric CO2, higher temperature, and altered

precipitation magnitude. Land use conversion from maize to bioenergy crops significantly

modifies above-ground canopy structure, affecting surface hydrological cycling in several

ways. Higher LAI allows these perennial crops to intercept more rainfall before reaching

the soil, which is then lost through evaporation, in combination with evaporative losses of

increased condensation moisture on leaf surfaces [20,21]. Denser foliage will also modify the

canopy radiative regime and within-canopy micro-climate [22], impacting ET, for example,

by way of reduced soil-evaporation as a result of the reduced energy flux reaching the ground

surface [23]. While alterations in canopy structure affect energy and water partitioning above

ground, climate change is expected to trigger acclimatory responses in vegetation that lead

to the modification of eco-hydrological responses [23]. In the context of the plant acclimation

categorization presented by Drewry et al. [23], these C4 crops do not show any significant

structural (leaf area) or biochemical acclimation (photosynthetic down-regulation), with the

primary response to elevated CO2 being ecophysiological acclimation (decreased stomatal

conductance), and associated decreases in canopy-scale transpiration. This conclusion is

drawn based on Free Air CO2 Enrichment (FACE) experiments which have demonstrated a

lack of response of photosynthesis, biomass accumulation, and yield of maize under elevated

CO2 [24,25]. These experiments have also shown insensitivity of key photosynthetic enzymes

of this C4 crop to elevated CO2 [24], and have pointed to the alleviation of water stress as the

primary impact of elevated CO2 on maize productivity [26, 27], in agreement with previous

hypotheses on the impact of elevated CO2 on the functioning of C4 plants [28]. Published

results for response of miscanthus and switchgrass to elevated CO2 are not yet available. We

have therefore adopted maize response as prototypical of all three crops. Increases in air

2

temperature will likely increase ET losses, potentially offsetting the conservative impact of

reduced stomatal conductance on transpiration. The combined impact of these counteracting

effects is a complex function of the biophysical functioning of each crop type, resulting

in potentially significant changes in canopy-integrated water and heat exchange with the

atmosphere.

In this thesis, we explore the potential hydrologic change associated with simultaneous

land use conversion to bioenergy crops and projected climate change in the U.S. Midwest.

Specifically, we contrast the eco-hydrological responses of maize, the main feedstock for

current starch-based biofuel production, with miscanthus and switchgrass, through the ap-

plication of a vertically-resolved model of canopy biophysical processes. The simulations

are conducted by parameterizing a multi-layer canopy model (MLCan; [22, 23]) to account

for canopy structural and biophysical functional characteristics of miscanthus and switch-

grass. This work is performed in four stages. First, the model is run for the year 2005 when

field observations of leaf photosynthetic CO2 uptake (An), strongly correlated with water

utilization [29, 30], is available for miscanthus and switchgrass, providing data for model

validation. Second, we examine the alterations in the energy balance and canopy temper-

ature that result from the land use conversion from maize to miscanthus and switchgrass

under present climate in 2005. Third, as a single year of data does not capture the entire

range of meteorological variability in the recent past, we use a stochastic Weather Generator

(WG; [31,32]) to provide an ensemble of forcing for the model. This ensemble enables us to

examine the range of crop responses to potential meteorological forcing. In the fourth stage,

meteorological forcing ensembles are generated which capture climate variability associated

with a number of climate change scenarios projected for the U.S. Midwest for 2050 [33]. The

model is forced using each of the climate scenarios to produce variability range correspond-

ing to the hydrologic predictions associated with these future climate scenarios. We then

estimate the water use of bioenergy crops and the impact on the hydrologic cycle which is

characterized through the change in soil-water storage and specific surface runoff (runoff per

unit area).

This thesis is organized in five chapters:

3

• Chapter 2 describes the methods, models and datasets utilized in this study. We briefly

review the theoretical background of the MLCan model [22] and the WG [31, 32].

Climate change projection for Midwest region for 2050 and data are also discussed in

this chapter.

• Chapter 3 presents a graphic user interface (MLCanFace) designed for facilitating the

use of MLCan model and making it easy to use for all interested users. We show

features and processing steps in this interface to conduct simulations in MLCan.

• In Chapter 4, numerical results of the WG and MLCan model for bioenergy crops

are presented. We discuss the simultaneous impacts of land use conversion, eco-

physiological acclimation of bioenergy crops under climate change associated with

climate variability on hydrologic cycle in the Midwest.

• Chapter 5 concludes with summary and research contribution of this thesis, including

open questions that can be addressed in future work.

4

CHAPTER 2

METHODS AND DATA DESCRIPTIONS

This chapter reviews the materials and methods used in this study. In the first section

formulation of multi layer canopy-root-soil model (MLCan) described by Drewry et al. [22]

is briefly summarized and discussed. This model has been extensively validated for both

ambient and elevated CO2 conditions for maize (C4) and soybean (C3) [22,23] and is applied

for bioenergy crops throughout this research. The aim is to resolve accurately the vertical

light and thermal regimes within the canopy and representing the tight coupling between the

leaf ecophysiology, energy balance and soil moisture state required to predict the vegetation

response to environmental perturbations. We then discuss the use of a stochastic Weather

Generator (WG; [31,32]) to deal with the variability in climate. Following the presentation

of the WG, a description of data sets used for model forcing, validation and structural

information for canopy and root systems is presented. In the last three sections, climate

change projection for the Midwest for 2050 and methods for water balance and albedo

estimation are discussed.

2.1 Multi-layer canopy root soil model

The multi layer canopy-root-soil model (MLCan) resolves the canopy radiation and meteo-

rological microenvironment and leaf-level ecophysiological states at multiple canopy levels to

determine canopy-atmosphere scalar fluxes (See [22] for further details). The model is driven

by above-canopy observations of shortwave (Rg) and longwave (Rlw) incident radiation, air

temperature (Ta), vapor pressure (ea), ambient CO2 concentration (Ca), wind speed (U)

and precipitation (PPT ) as shown in Figure 2.1. Consideration of photosynthetically ac-

tive (PAR) and near-infrared (NIR) shortwave bands separately is required to account for

5

U

!"#$%&'()*

+),-.)*

/#0,&'()*

+),-.)*

z/h = 1!Ca! Ta! ea!

!"'1)1*

2$'3%-#0*

4$'-0',)*

+##%*5'%)$*67%'8)*

/'9)$*!"

/'9)$*#"

Fc,i!

LE,i!

Hi!

/'9)$*:*

/'9)$*;*

<$)3-7-%'%-#0*

'01*4)&*

2#$.'%-#0*

/'9)$*.*

/'/'

='0#79**

5'%)$*

!%#$',)*

=

/'9)$*:*

"!

!"

2$'

!>0?-%*

2$'3%-#0*

@$%*A*,B*A*&!CB>0"

@$%*A*,B*A*&!CB"'1)"

/'9)$*:*

##%*5'%%%%)$*6666667%%%%'88888)*

/'/'99)$*;*)$*;

###%%%%%% 555555#+#++++++#

/'/'999)$*.*)$*.

Figure 2.1: Schematic of the canopy, root, and soil system model. Shortwave radiation compo-nents (PAR and NIR) are attenuated through the canopy, accounting for absorbed, transmitted,and reflected fractions at each layer. Direct (red) and diffuse (pink) components of the shortwavestreams are considered separately to account for sunlit and shaded leaf fractions at each canopylevel. The longwave radiation regime accounts for absorption and emission by the foliage in eachcanopy layer. Incident radiation fluxes force the soil heat budget model. Wind speed is resolvedwithin the canopy space, as are gradients in concentrations of CO2, water vapor, and heat. Rainfalland dew accumulate on foliage, resulting in evaporation and a reduction in throughfall to the soilsurface. Rainfall and dew replenish the subsurface moisture store that supplies the root systemmoisture for use in daytime evapotranspiration. A multilayer soil and root system model are usedto compute moisture uptake through the root zone. (Reproduced/modified from Drewry et al. [22]by permission of American Geophysical Union. Copyright 2010 American Geophysical Union).

the much higher absorptivity of PAR for green leaves [34]. Both PAR and NIR attenua-

tion are in large part determined by the vertical distribution of foliage, using a Beers law

approach [35]. The two shortwave bands are decomposed into direct and diffuse streams,

which allows for separate consideration of sunlit (fsun) and shaded (fshade) leaf fractions,

as sunlit leaves receive both direct and diffuse radiation while the shaded fraction receives

only diffuse one [36]. The canopy longwave regime is a function of both incident longwave

from the atmosphere and the soil, and also that emitted by the foliage at each canopy level.

6

The net gain in longwave energy (LWnet) at each canopy layer is the difference between the

absorbed and emitted longwave flux magnitudes.

Canopy-top observation of wind speed, air temperature and vapor pressure are applied

as upper boundary conditions for the solution of the within-canopy environment. Water

storage on foliage, as a function of dew or rainfall interception, is likewise considered [22].

Energy intercepted by wet surfaces is used for evaporation, requiring separate consideration

of the energy balance of wet leaves. Throughfall and soil surface evaporation provide the

surface flux boundary condition for the solution of soil moisture [22].

MLCan model incorporates explicit coupling between leaf-level ecophysiological processes

(photosynthesis and stomatal conductance), physical processes (energy balance and bound-

ary layer conductance), and below-ground water status which incorporates a hydraulic redis-

tribution (HR) model [37] (Figure 2.2). The dashed arrow connecting stomatal conductance

and photosynthesis signifies that this connection may not generally be important for the

determination of An in C4 plants, due to the CO2 concentrating mechanism of C4 physiology

that effectively decouples An from gs across a wide range of conditions typically experienced.

Net leaf fluxes at each canopy level are calculated as a weighted average of the sunlit and

shaded leaf fluxes and those from the wet leaf fraction. Photosynthetic capacity, as repre-

sented by the maximum rates of ribulose-1,5-bisphosphate (RuBP) regeneration and RuBP

carboxylase-oxygenase (Rubisco) catalyzed carboxylation at 25◦C (Jmax,25 and Vcmax,25, re-

spectively), declines exponentially through the canopy, affecting An and leaf respiration at

each canopy level. Canopy-scale fluxes are computed through the vertical integration over

each of the foliage layers [22].

Moisture uptake from the soil by the root system provides water for transpiration, with

soil water deficits acting to constrain leaf-level fluxes through a control on leaf water po-

tential (Ψl) that results in decreased stomatal conductance and plant gas exchange. Root

system structure, in particular the distribution of conductive elements through the soil col-

umn, acts in concert with the vertical distribution of soil moisture to determine the avail-

ability of moisture to the plant. MLCan utilizes a vertically resolved soil column, with root

conductivity distributed according to observed profiles of root biomass. Radial root con-

ductivities determine the ability of the root system to absorb moisture from the soil. The

7

Met

Forcing Fluxes

An, LE, H

Leaf EnergyBalance

Boundary Layer Conductance

BiochemicalPhotosynthesis

StomatalConductance

Leaf Ecophysiology

Tl

An

Ci

gbvgbh

Leaf Physics

Root Water Uptake

gbvTl

gs

P

Rg, Rlw

ea, Ta, U

Figure 2.2: Leaf-level model component coupling. Observed meteorological conditions drive theleaf-level ecophysiological (photosynthesis and stomatal conductance) and physical (boundary layerconductance and energy balance) component models. The components are coupled by the boundarylayer conductances to vapor and heat (gbv and gbh), leaf temperature (Tl), net CO2 uptake (An),internal CO2 concentration (Ci), and stomatal conductance (gs). Reduced root water uptake underperiods of stress can constrain stomatal conductance. Outputs include the flux densities of CO2,latent and sensible heat for the canopy layer. (Reproduced/modified from Drewry et al. [22] bypermission of American Geophysical Union. Copyright 2010 American Geophysical Union).

root distribution-weighted pressure potential (Ψr,wgt) and the transpiration requirement of

each canopy layer determine the distribution of Ψl through the canopy. Plant hydraulic

conductance is reduced as Ψr becomes more negative, further regulating Ψl potentially to

guard against xylem cavitation [22].

In this study, the MLCan model is first run using meteorological forcing and observed LAI

data for 2005 for miscanthus and switchgrass reported in Heaton et al. [11] (See details in

section 2.3). For the present conditions, atmospheric concentration of CO2 is set equal to 370

ppm [39]. This simulation is used for model validation using observed An of upper canopy

sunlit leaves during the growing season [38]. Second, we use an ensemble of generated weather

data consisting of 30 independent years with climate change projection in the Midwest [33]

(See detail in section 2.4) for driving the MLCan model to predict water uses by energy crops.

8

2.2 Weather Generator

Stochastic Weather Generator first developed by Curtis and Eagleson [31] and later modified

by Ivanov et al. [32] is used for developing a forcing ensemble and driving the MLCan model

(see Figure 2.3). Details on the formulation of this WG is presented in Appendix A. The

WG provides Monte Carlo simulation for hourly data which are then linearly interpolated to

obtain half-hour values corresponding to the MLCan model time step. Parameters for the

generator are obtained from 10-year (1997 - 2006) observation time series at the Bondville

flux tower (See section 2.3.1 for further detail). The stochastic WG should be expected

to capture the range of variability observed during this time period (See Figure 2.5). An

ensemble of 30 independent years of weather simulation is used for each case in the study.

Observed Data

Model Parameters

Possible Weather Conditions

Parameter

WeatherGenerator

Climate ChangeProjection

Estimation

Figure 2.3: Application of Weather Generator for developing a forcing ensemble in Bondville, IL.Half-hourly observed meteorological data in 10 years (1997-2006) is used for parameterization.Along with climate change projections [33], the obtained model parameters are used to providepossible weather conditions in the future using a Weather Generator [31,32].

2.3 Data sets

2.3.1 Meteorological Data

Half-hourly meteorological data are obtained from AmeriFlux tower (See the official website

at: http://public.ornl.gov/ameriflux) located in Bondville, Illinois (40.01◦N, 88.29◦E). A

map showing the location of the Bondville flux tower is presented in Figure 2.4A. Measure-

ments were obtained from an instrumented 10 meter tower since observations began in 1997

(Figure 2.4B). Data in the growing season 2005 is used for MLCan model validation (Fig-

ure 2.5(A-D)). Observed data from 1997−2006 at this tower, and cloud cover data at the

9

0 40 80 120 km40

92° W 91

° W 90

° W 89

° W 88

° W 87

° W

37° N

38° N

39° N

40° N

41° N

42° N

43° N

Legend

Counties

Bondville

tower

B

A

Source: http://public.ornl.gov/ameriflux/

Figure 2.4: Ameriflux tower at Bondville, Illinois, USA. (A) Map of Ameriflux tower located atBondville, Central Illinois. Green area represents Champaign county; (B) 10-meter eddy covarianceflux tower in Bondville. Net radiation (Rn), downward shortwave (SW ↓) and longwave (LW ↓)energy fluxes were measured with a net radiometer and quantum sensor at the top of the tower. Atipping bucket rain gauge produced estimates of precipitation. Meteorological conditions, includingwind speed, air temperature and humidity were observed from instrumentation on the tower.

nearby University of Illinois Willard Airport obtained from National Climatic Data Cen-

ter (http://www.ncdc.noaa.gov/oa/ncdc.html) are used for parameterizing the stochastic

WG [31,32].

2.3.2 Ecophysiological Data

Ecophysiological data is also collected in central Illinois. LAI data for miscanthus and

switchgrass is obtained from a published study by Heaton et al. [11] with approximately

bi-weekly measurements from emergence to senescence at three locations very close to the

Bondville tower during 2005. Each location has four plots of 10m × 10m for both miscanthus

10

0

20

40

60

[mm

]

0

200

400

600

800

[mm

]P

PT

0

10

20

30

[M

J m

−2 d

ay−

1]

0

10

20

30

40

Tem

pera

ture

[ oC]

100 140 180 220 260 300

0

2

4

6

8

10

Miscanthus

Switchgrass

Maize

[m2 m

−2]

Day of year

PP

TR

gLA

I

A

E

B

D

C

Figure 2.5: Key meteorological forcing data observed in 2005 overlaid on an ensemble obtainedusing the stochastic weather generator. Meteorological data includes (A) Daily precipitation (PPT,black bars); (B) Cumulative precipitation (PPT, red line); (C) Daily global radiation (Rg, blueline); and (D) Mean day time air temperature (Ta, magenta line). Grey bars and lines in (A, B, C,and D) represent corresponding data obtained from stochastically generated weather ensemble of30 independent years. All observed meteorological data in 2005 is obtained from Ameriflux towerat Bondville, Illinois, USA [22]. (E) Leaf Area Index (LAI) for the maize (green circles), mis-canthus (green triangles), and switchgrass (green squares) canopies were obtained from publishedsources [11, 22]. Miscanthus and switchgrass have a longer growing season as compared to maize,both at the beginning and end which is reflected in the LAI plots.

11

and switchgrass. Observations for An of upper canopy sunlit leaves are obtained from another

published study [38]. These observations were made at 2 hour intervals from pre-dawn to

post-dusk on 8 separate days across the growing season in 2005 in the same plots as studied

by Heaton et al. [11].

2.3.3 Canopy, soil, and root system

Canopy structure is described by leaf area density (LAD) profile and the total LAI. Vertical

distributions of leaf area through the canopy for miscanthus [40] and for switchgrass [41]

are averaged and normalized by dividing the canopy LAI at the time of measurement (Fig-

ure 2.6A). Distributions of root systems through the soil column for miscanthus and switch-

grass are obtained from the study by Monti & Zatta [42] (Figure 2.6B). Canopy and root

structures of maize are obtained from the study by Drewry et al. [22]. Initial condition of

soil moisture is set equal to 0.3 [22].

0 0.02 0.04 0.06 0.080

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

[−]

Normalized LAD Profile

A.

z /

h

0 0.2 0.4 0.6 0.8−1.2

−1.0

−0.8

−0.6

−0.4

−0.2

0

Normalized Root Fraction

[m]

B.

z

Miscanthus Switchgrass

Figure 2.6: Normalized canopy leaf area density profiles (A) and normalized root fraction in eachsoil layer (B) for miscanthus (red circles) and switchgrass (blue squares). The vertical axis in (A)is normalized by the height of the canopy (3.5m for miscanthus, 2.0m for switchgrass) to facilitatecomparison between the two crops(based on data obtained from [40–42]).

12

2.4 Climate change projections

Climate change projection for 2050 in Midwestern U.S. follows the study by Wuebbles and

Hayhoe [33] and Houghton et al. [39]. Under climate change, atmospheric concentration

of CO2 is set equal to 550 ppm while precipitation pattern in the Midwest is expected to

decrease 15% in the summer and increase 10% in the winter at the middle of the 21st cen-

tury [33]. In addition, expected average increase of temperature ranges from 1 to 3◦C in

the summer and from 0.3 to 1◦C in the winter [33]. To evaluate the role of temperature

increase, we conducted an air temperature sensitivity analysis to the water uses by increas-

ing air temperature at 3 levels in both summer and winter seasons. However, vegetation

growth period mostly spans the summer season. We also conducted predictions for both

change and no change in precipitation pattern to assess the role of water availability on the

predictions. Eight scenarios (S1 to S8) corresponding to proposed climate projections in the

Midwest U.S. for 2050 are presented in Table 2.1.

Table 2.1: Projected climate change scenarios for Midwestern U.S. for 2050 during the summerfor driving MLCan model predictions [33]. Symbol “−” represents no change in variable.

ScenarioCO2 concentration Precipitation Temperature

(ppm) change [%] increase [◦C]

S-1 550 − −

S-2 550 -15% −

S-3 550 − + 1

S-4 550 − + 2

S-5 550 − + 3

S-6 550 -15% + 1

S-7 550 -15% + 2

S-8 550 -15% + 3

13

2.5 Water balance

Change in soil water storages is important for evaluating the impact of different land covers

on the hydrologic cycle. It is given as:

dS

dt= P + C − TR − E − SE − SP −R (2.1)

where P , C, TR, E, SE , SP , and R represent precipitation, condensation, transpiration,

evaporation, soil evaporation, seepage, and specific surface runoff. All variables are in the

dimensions of [L / T].

2.6 Calculation of albedo

For the whole spectrum, the reflection coefficient of a natural surface is often called the

albedo [34, 43]. Its values are very important in shaping local and global climates through

the radiation budget. Albedo for each crop in this study is estimated based on the ratio of

total outgoing and incoming shortwave radiation during the daytime.

α =

∫SW ↓>40(W ·m−2)

SW ↑dt∫SW ↓>40(W ·m−2)

SW ↓dt(2.2)

α: albedo [dimensionless];

SW↓: downward or incoming shortwave radiation [W·m−2];

SW↑: upward or outgoing shortwave radiation [W·m−2].

14

CHAPTER 3

GRAPHIC USER INTERFACE FOR MLCAN

3.1 Introduction

In the previous chapter, we have described the formulation of multi layer canopy-root-soil

model (MLCan) which couples above ground ecophysiological processes with below ground

coupled soil-root moisture transport. Drewry et al. [22,23] also demonstrated the important

role of environmental driver and model skill of MLCan in capturing variability in canopy-

atmosphere fluxes exchange. Biological data (LAI) and high resolution of a range of me-

teorological data are required for driving the model. Moreover, the MLCan model requires

specification of structural characteristic of the canopy, root and soil systems to simulate

accurately the vertical variation in radiative states and scalar fluxes [22].

The lack of a friendly graphic user interface (GUI) and visualization capabilities, however,

is often a barrier to process raw input data and setup the MLCan model, and it is not

easy for an ecohydrologist with no specific training in MLCan model or a novice user to

operate it. Throughout this Chapter 3, we present a GUI that works as an additional

preprocessor for MLCan model. The aim of this GUI is to provide a better visualization

tool and to simplify significantly the processing steps in this model for all users. We show

important features of this GUI as well as basic steps to process and transfer raw input data

to the MLCan model. An independent interface that works as a post-processor for viewing

numerical results obtained from original MLCan model is also presented.

15

3.2 Graphic User Interface for MLCan

The GUI for multi layer canopy root-soil-model, MLCanFace, is developed and linked to

the original model code [22] in MATLAB (2009a, Mathworks, Natick, MA, U.S.) that can

be executed in many operating systems (e.g. Linux, Windows, Mac OS). It uses interactive

visualizations as an interface which supports MLCan users to organize data, and save and

load their work more easily. When data processing is completed, a set of processed data and

required parameters will be saved and transferred to the original MLCan model. Finally,

the MLCan model is activated to implement simulations.

The main window of MLCanFace is illustrated in Figure 3.1, showing five important sub-

components on the left side of the interface: (i) Model setup for providing general simulation

and vertical structure information; (ii) Options for selecting sub-models and simulation

modes; (iii) Forcings & Initial conditions for processing input data and initial conditions

required in MLCan model; (iv) Parameters for providing a diverse set of parameters required;

and (v) Result viewer for displaying simulation results in graphical modes. Additional menu

bar and toolbar are also integrated, providing other convenient ways for running the model.

To support new users, a short instruction summarizing basic steps in MLCanFace is presented

on the right of the main window (Figure 3.1). For further detail instructions, users can open

Help section on the menu bar.

3.2.1 Model Setup

The Model setup is comprised of 3 subsections, one for specification of geographic location

and simulation periods by mean of a threshold value in LAI data, and the other two for the

specification of root and canopy structures with plotting capabilities (Figure 3.2).

Specify Site Information

• Vegetation name: The name of simulated vegetation is provided in the Vegetation

name text box. Vegetation name is required and used for saving output files of the

MLCan model later.

16

Figure 3.1: Main window of MLCan graphic user interface. The GUI is organized by functionand task on the left. Corresponding instructions for each sub-components are presented on theright of the main window.

• Latitude, Longitude, and Elevation: Geographic location (degree) and site elevation

(meter) are required for zenith angle calculation.

• Minimum LAI: The minimum LAI is a user-defined threshold for identifying simulation

periods. Based on LAI data sets, simulations are conducted on the days in which LAI

was greater than this threshold.

Canopy structure

• Number of canopy layer: Vertical canopy resolution is identified by the number of

layers used to subdivide the canopy. Finer vertical resolution is better for accuracy

but requires more time for simulation. Choosing the number of canopy layers should

balance these important criteria.

17

• Leaf Area Density (LAD) layer: LAD layer is used to describe the observed, vertical

canopy structure. The observed structure is then interpolated to match the number of

canopy layer provided above

• Input LAD profile: Users can input observed LAD profile in two different ways. The

first (manual) way is using designed table built in the interface. The second (automatic)

way is importing observed data from supporting files. The supporting files include:

Text (*.txt), Comma-separated values (*.csv), and Microsoft Excel (*.xls) files.

Root structure

• Number of root layer: Similar to canopy structure, vertical root structure is identified

by the number of root layers in the soil. Choosing the number of root layer should also

balance the accuracy and simulation time criteria.

• Input root profile: The distribution of root through the soil column is provided via two

methods. First, a logistic dose-response curve used by Schenk and Jackson [44] can be

applied. The two parameters representing 50% and 95% cumulative root distribution

and one parameter for the maximum root are required for root distribution estimation.

Second, observed data can be imported to describe the root distribution. This method

uses the same procedure including two different ways to input raw data as shown in

the Input LAD profile.

3.2.2 Options

Simulation modes and sub-models are selected in Options windows (Figure 3.3). The Pho-

tosynthesis panel allows flexibility in the choice of photosynthetic pathways, C3 or C4, in

the MLCan model. The Linear option in the Root Hydraulic Conductivity panel represents

the linear increase of root hydraulic conductivity (RHC) with soil depth. If this option is

not selected, no increase of RHC along the soil depth is applied. The Root Numerical Model

is used for selecting numerical finite difference scheme in root model. Users can select either

18

Figure 3.2: View of Model Setup in MLCanFace. The Model Setup is comprised of 3 subsections:(i) Specifying site information; (ii) Root structure; (iii) Canopy structure. The root and canopystructure windows are plotting supported.

Explicit or Implicit scheme [45] for the root model. Ca is input in the CO2 concentration

panel. Two default values for ambient condition (370 [ppm]) and elevated condition (550

[ppm]) are given for reference. However, user can easily modify Ca for their specific uses.

There are five sub-models shown in the Models panel located at the bottom-left side of the

Options window. Each sub-model is activated by checking the corresponding box. Soil Heat

Model is used for simulating the heat transfer in the soil system. The Turbulence model is

used to calculate the turbulence transport of micro-environment using a first-order canopy

closure model. Hydraulic Redistribution (HR) model simulates the nocturnal transfer of

water by roots from moist to dry regions of the soil profile. The last two Nutrient models

(Carbon and Nitrogen) are used to simulate the nutrient dynamics in soil and plant in

MLCan model. These two models are temporarily unavailable.

3.2.3 Forcings and Initial conditions

Processing input data and setting up initial conditions are most important steps in model-

ing. The Forcing & Initial conditions interface is designed to support and facilitate these

processes in MLCan model. This interface is designed in two separate tab-pages (Figure 3.4):

19

Figure 3.3: View of model Options in MLCanFace.

Forcings

• Forcings page is used for processing new input data or loading existing data files. A

flexible naming system is used that allows user to save or select forcing files. Error

message boxes are also integrated to redirect user as errors occur. We also integrate

a friendly calendar application for time setting when users create new forcing files.

There are two options for processing new forcing data: (i) By choosing Create blank

file, users are allow to create blank forcing files with appropriate temporal and spatial

information obtained from other sub-components of the MLCanFace, then users can

use MATLAB to transfer the observed data to the blank forcing file; or (ii) user can

import directly data from existing file by choosing Create and Import. A browsing

windows will appear to help users select appropriate files easily.

20

A

B

Figure 3.4: View of model Forcing and Initial Conditions in MLCanFace. (A) The Forcings tabis used for data processing. (B) The Initial conditions tab is used for setting up initial conditionsin the root-soil models.

21

• Initial conditions page is used to set up initial conditions for soil moisture and heat in

the root-soil model. The spatial discretization in the soil is obtained from the verti-

cal root profile shown in Model setup section. Users can either input initial conditions

manually or import automatically from existing data files. The supporting files also in-

clude: Text (*.txt), Comma-separated values (*.csv), and Microsoft Excel (*.xls) files.

3.2.4 Parameters

Users must provide the parameters that control the simulation processes. In fact, the MLCan

model requires a large number of parameters in several categories. The Parameters window

in MLCanFace is provided to support this process in the model.

More than 40 parameters required for MLCan model are classified in four major groups:

(i) Leaf & Canopy; (ii) Soil & Radiation; (iii) Photosynthesis; and (iv) Conductance & Res-

piration (See Figure 3.5 and Figure 3.6). Each group is shown in a separate tab-page located

on the top of the interface. For example, the view of Photosynthesis parameters for both C3

and C4 plants are shown in Figure 3.5A. Only one table in photosynthesis tab-page is enable,

depending on the user’s selection for photosynthesis model in Options (See section 3.2.2). A

list of parameters used for MLCan model is described comprehensively by Drewry et al. [22].

We refer interested users to that paper for further information on model parameters.

3.2.5 Result Viewer

Presenting and plotting result are important for evaluating model simulations. An inde-

pendent Result Viewer is developed for supporting users to present the numerical results

obtained from MLCan model easily. For convenience, it is also integrated directly to the

MLCanFace (Figure 3.1).

The window of Result viewer in MLCanFace is presented in Figure 3.7. At the top of

the interface, a browsing windows is designed to allows users to select result files easily.

Information in the Load result file panel shows the name of the working file, time step, and

time that simulation is started and ended (Figure 3.7).

22

A

B

Figure 3.5: View of Leaf & Canopy (A) and Soil & Radiation (B) parameters in MLCanFace.

23

A

B

Figure 3.6: View of Photosynthesis (A) and Conductance & Respiration (B) parameters in ML-CanFace. In photosynthesis parameters, only one table (C3 or C4) is enable, depending on theselection of photosynthesis model.

24

There are four view modes for users to examine interested variables. The Time Series

Fluxes panel plots a single time series from the beginning to the end of the simulation period.

The Diurnally Averaged Flux panel plot the diurnal variations of interested variable with

vertical bar representing ± one standard deviation. The Time Series Canopy Profile plot the

canopy profile from the beginning to the end of the simulation period using Matlab “pcolor”

function. The Diurnally Averaged Canopy Profile plots the diurnally average canopy profile

based on the whole simulation period.

Figure 3.7: Result viewer in MLCanFace. Four plotting modes are provided to examine corre-sponding interested variables.

3.3 Summary

This chapter has presented a graphic user interface (GUI) for the MLCan model [22], sup-

porting users to manipulate and run it with graphical visualization capabilities and avoiding

error-prone processes in the text mode. The GUI is developed and linked to the original

MLCan model in MATLAB. The objective is to provide a friendly user interface with a flexi-

25

bility in the choice of operating systems. Indeed, users can run the MLCanFace in Microsoft

Windows, Linux, or Mac OS environments.

The designed GUI works as a preprocessor as well as a post-processor for the MLCan

model. First, raw input data and parameters required are classified and processed in separate

sub-components of the GUI. General information, processed data and parameters are saved

and transferred to the original MLCan model in appropriately formatted files. The original

MLCan model is then activated to conduct simulations. Second, the GUI are also used to

analyze output files obtained from the MLCan model. Numerical results, saved in Matlab

format (*.mat) files, are loaded and transferred easily to the GUI for graphical presentations

and analyses.

There are five main sub-components in the MLCanFace, accounting for different tasks and

functions. The Model setup is used for providing general information and vertical structure

information. The second component, Options interface, is designed to select simulation

modes and sub-models. Third, the Forcing & Initial conditions interface is provided to

support input data processing and initial conditions setting. Next, the Parameters is used

to provide model parameters that control the simulation processes. Finally, the Result viewer

is designed to facilitate the result processing with four plotting modes. Each sub-component

is designed to simplify manipulations in MLCan model, avoid the error-prone process in the

text mode, and decrease the time needed to build simulations for novice users.

In general, the application of coupled canopy, leaf, root and soil processes resolved ver-

tically through the canopy and soil domains are important for simulating the vegetation

responses to climate states and perturbation. This work requires an enormous and con-

certed effort for data processing and model setup. It is expected that the MLCanFace will

be used as an useful visualization for facilitating the modeling work in MLCan model.

26

CHAPTER 4

RESULTS AND DISCUSSIONS

In this chapter, we evaluate the applications of MLCan model [22] through its GUI, pre-

sented in the previous one, for exploring the simultaneous impacts of land use conversion,

eco-physiological acclimation of bioenergy crops under climate change associated with cli-

mate variability on hydrologic cycle in Central Illinois, characterizing much of the Midwest-

ern U.S., based on collected data described in Chapter 2. We first validate the MLCan

model for two bioenergy crops and the WG [31, 32] for Ameriflux Bondville tower site, Illi-

nois. Within-canopy vertical variation and impacts of crop type on energy balance are then

discussed. We try to understand the impact of land use change to the canopy-atmosphere

exchange and energy balance. In the next section, the vegetation response to elevated CO2

and increased air temperature are examined to evaluate the impact of climate change on

vegetation response to environmental perturbations. Finally, we explore the water use of

bioenergy crops and their impacts on hydrologic cycle which is characterized through the

change in soil-water storage and specific surface runoff.

Simulations for maize, miscanthus and switchgrass for growing season in the year 2005 at

two other locations in the Midwest region: (i) Ozark, Missouri (38.74 ◦N, 92.20 ◦E) and (ii)

Brookings, South Dakota (44.34 ◦N, 96.84 ◦E) are also conducted for comparison. Details

and simulation results for those two additional sites are presented in Appendix B.

4.1 MLCan model validation

Comparisons of modeled and observed (data obtained from [38]) photosynthetic leaf CO2

uptake (An) for several days demonstrate the ability of the MLCan model to capture the

ecophysiological functioning of both miscanthus and switchgrass throughout the growing

27

season (Figure 4.1). An for miscanthus is consistently higher than switchgrass throughout

the growing season. The fluctuations ofAn for both crops are strong on some days (e.g Jul 7th,

Aug 10th) and are an indicator of the tight link between An and environmental conditions.

Variations in solar radiation due to cloudiness and the linked air temperature fluctuations are

the primary drivers of variability in biochemical photosynthesis and stomatal conductance

which in turn controls leaf temperature through the energy balance [22] (See Figure 2.5).

Along with the study by Drewry et al. [22], a list of essential parameters and their values

for maize, miscanthus, and switchgrass is presented in Table 4.1.

0

10

20

30

4001-Jun

An [µ

mo

l m

-2 s

-1]

A Miscanthus

15-Jun 07-Jul 20-Jul

4 8 12 16 20

0

10

20

30

4010-Aug

4 8 12 16 20

01-Sep

4 8 12 16 20

22-Sep

4 8 12 16 20

06-Oct

Time [hours]

0

10

20

30

4001-Jun

An [µ

mo

l m

-2 s

-1]

B Switchgrass

15-Jun 07-Jul 20-Jul

4 8 12 16 20

0

10

20

30

4010-Aug

4 8 12 16 20

01-Sep

4 8 12 16 20

22-Sep

4 8 12 16 20

06-Oct

Time [hours]

Figure 4.1: Comparison of half-hourly net photosynthetic leaf CO2 uptake (An) of upper canopysunlit leaves for (A) miscanthus modeled by MLCan (red solid lines) and observed data (red circles)and (B) switchgrass modeled by MLCan (blue solid lines) and observed data (blue open circles).Observed data is obtained from Dohleman et al. [38] on eight separate days during the growingseason in 2005. Grey shading represents nighttime.

28

Table

4.1:

Valu

eof

mod

elp

ara

met

ers

for

mai

ze,

mis

canth

us,

and

swit

chgr

ass

use

din

the

mu

ltil

ayer

can

opy-r

oot

-soi

lm

od

el(M

LC

an).

[#]

refe

rsto

refe

ren

ces.

Des

crip

tion

Sym

bols

Un

itM

aiz

ea

Mis

canth

us

Sw

itch

gra

ss

Can

opy

hei

ght

hcan

m2.5

3.5

[40]

2.0

[41]

Lea

fw

idth

do

m0.0

80.0

30.0

1

Dec

ayco

effici

ent

for

leaf

nit

roge

nco

nte

nt

kn

-0.5

0.1

50.4

Lea

fem

issi

vit

yε v

-0.9

40.9

5[65]

0.9

5[65]

Lea

fab

sorp

tivit

yto

ph

otos

ynth

etic

acti

vera

dia

tion

(PAR

)αL−PAR

-0.8

00.8

4[40,66]

0.8

[34]

Lea

fab

sorp

tivit

yto

nea

r-in

frar

ed(N

IR)

αL−N

IR

-0.2

30.2

[34]

0.2

[34]

Diff

use

exti

nct

ion

coeffi

cien

tK

d-

0.5

50.6

8[67]

0.6

7[41]

Lea

fan

gle

dis

trib

uti

onp

aram

eter

x-

1.6

41.6

41.6

4

Intr

isic

quan

tum

yie

ldC

4p

hot

osynth

esis

α4

mol·m

ol−

10.0

35

0.0

35

0.0

34

Init

ial

slop

eof

C4

ph

otos

ynth

etic

CO

2re

spon

sek4

mol·m

−2·s−

10.7

0.7

0.7

Ref

eren

ceva

lue

fole

afre

spir

atio

nR

dµ

mol·m

−2·s−

10.8

0.8

0.6

Ref

eren

ceva

lue

for

sub

stra

tesa

tura

ted

Ru

bis

coca

paci

tyVm

ax

µm

ol·m

−2·s−

140

66

48

Tem

per

ature

sen

siti

vit

yof

tem

per

atu

re-d

epen

den

tC

4p

ara

met

ers

Q10,4

-2.0

2.5

0.5

Sto

mat

alsl

ope

par

amet

erin

Bal

lB

erry

mod

elm

-7.0

5.7

8.0

Sto

mat

alin

terc

ept

par

amet

erin

Bal

lB

erry

mod

elb

mol·m

−2·s−

10.0

08

0.0

07

0.0

08

Sto

mat

alse

nsi

tivit

yp

aram

eter

s fM

Pa−1

6.5

6.5

6.5

Ψl

atw

hic

hh

alf

pot

enti

alg s

islo

stΨ

fM

Pa

-1.3

-1.3

-1.3

aV

alu

esar

eob

tain

edfr

omD

rew

ryet

al.

[22]

29

4.2 Weather Generator validation

The performance of the WG are validated by reproducing hourly observations at Ameriflux

tower located in Bondville, Illinois. Statistics derived from hydrometeorological observed

data are compared with those obtained from a 10-year (1997-2006) simulation period by

the WG. A list of model parameters used in the WG adapted from [32] is summarized in

Table 4.2. Interested readers are also referred to the study by Ivanov et al. [32] for further

details on the WG parameter information. The values of these parameters obtained from

observed data in Bondville are presented in Appendix A.

The first important variables that need to be validated in the WG is precipitation. Com-

parison of observed mean monthly precipitation for 10-year period and those modeled by the

WG is shown in Figure 4.2. The precipitation model in the WG preserves well the monthly

statistical mean but slightly underestimates the variance, especially in July. Probably, the

relative short observations in 10 years used in this study is the primary reason for these dif-

ferences as this observed data can not capture the variability range of climate in general and

rainfall in particular. During the 10-year period, the mean of the total annual precipitation

from observations is 842 mm and that from simulations by the WG is 802 mm. Overall, the

WG demonstrate significant skills in preserving the statistical properties of precipitation.

1 2 3 4 5 6 7 8 9 10 11 120

50

100

150

200

month

PP

T [m

m]

Mean monthly precipitation

Simulated

Observed

Figure 4.2: Observed and simulated mean monthly rainfall at Bondville, Illinois. The verticalbars respresent ± one SD oh the monthly value.

Comparisons between observed and modeled cloud cover distribution are presented in

Figure 4.3. The WG is capable of reproducing both the U-shaped with spikes at zero and

30

Table

4.2:

Para

met

ers

use

din

the

wea

ther

Gen

erat

or.

(Rep

rod

uce

d/m

od

ified

from

Ivan

ovet

al.

[32]

by

per

mis

sion

ofA

mer

ican

Geo

physi

cal

Un

ion

.C

opyri

ght

2007

Am

eric

anG

eop

hysi

cal

Un

ion

).

Para

met

ers

Un

itD

efin

itio

nE

stim

atio

np

erio

d

µb

hou

rm

ean

tim

eb

etw

een

stor

mm

onth

-se

ason

µr

hou

rm

ean

stor

md

ura

tion

mon

th-

seas

onµd

hou

rm

ean

stor

md

epth

mon

th-

seas

on

M0

-m

ean

”fai

rwea

ther

”cl

oud

ines

sm

onth

σ2 m

-va

rian

ceof

”fai

rwea

ther

”cl

oud

ines

sm

onth

ρm

-la

g-1

auto

corr

elat

ion

of”f

airw

eath

er”

clou

din

ess

mon

thς

hou

r−1

clou

din

ess

dec

ayra

tefr

omth

een

dof

rain

fall

tofa

irw

eath

erm

onth

a,b

-sh

ape

par

amet

ers

ofb

eta

dis

trib

uti

onm

onth

uo

cmoz

one

amou

nt

ina

vert

ical

colu

mn

con

stan

t-

seas

onβ

-A

ngs

trom

turb

idit

yp

aram

eter

con

stan

t-

seas

onωAi

-ae

roso

lsi

ngl

e-sc

atte

rin

gal

bed

os(V

isib

lean

dN

IRb

and

)co

nst

ant

-se

ason

ρg

-sp

atia

lav

erag

ere

gion

alal

bed

oco

nst

ant

-se

ason

LWP

0g·

m−

2cl

oud

tota

lli

quid

wat

erp

ath

forovercast

con

dit

ion

sm

onth

b i(i

=0,

1,...4

)re

gres

sion

coeffi

cien

tsfo

rdet

erm

inis

tic

mon

thco

mp

onen

tin

tem

per

atu

reδT

◦ Cm

ean

ofra

nd

omte

mp

erat

ure

dev

iate

sm

onth

σ2 δT

◦ Cva

rian

ceof

ran

dom

tem

per

atu

red

evia

tes

mon

thρδT

-la

g-1

auto

corr

elat

ion

ofra

nd

omte

mp

erat

ure

dev

iate

sm

onth

P∗ ann

mm

pre

cip

itat

ion

ind

exin

dew

tem

per

atu

rem

od

elm

onth

Ws

m·s−

1m

ean

win

dsp

eed

mon

thσ

2 sm

2·s−

2va

rian

ceof

win

dsp

eed

mon

thρs(1

)-

lag-

1au

toco

rrel

atio

nof

win

dsp

eed

mon

thγs(t

)-

skew

nes

sof

win

dsp

eed

mon

th

31

one as well as the seasonality of the process, showing a generally good agreement between

observed and modeled data. However, the WG appears to overestimate cloud cover in late

spring and early summer months (Mar to Jun) as overcast condition decreases strongly. This

discrepancy was noted by Ivanov [46] as passing atmospheric precipitation systems sometimes

does not necessarily result in rainfall at a given location. The cloud cover, however, is non-

stationary during such periods and the estimated mean can be significantly affected [46].

Simulation of air temperature should be able to capture the diurnal cycle and average

daily variability. Figure 4.4 illustrated the comparisons of simulated and observed diurnally

averaged air temperature in 12 months. The model simulates well the daily air temperature

fluctuation and seasonality. It slightly underestimates air temperature in the afternoon for

summer months, but constantly overestimates for December. The model also underestimates

the variance, especially in winter months. At monthly scale, the air temperature statistics

are well reproduced as shown in Figure 4.5, showing small discrepancy as presented in

Figure 4.4. A reasonable explanation for this discrepancy is the short observations in data

used for parameter estimation. Generally, the WG reproduces well the air temperature

during growing season for bioenergy crops which mostly spans in the summer.

Comparisons of wind speed histograms computed from observed and modeled data are

shown in Figure 4.6. The frequency distribution of wind speed data is positively skewed [32].

Both the skewness properties and the first two moments of the distribution are preserved

well in the WG.

32

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.41002

so=0.46055

January

ms=0.46743

ss=0.47217

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.33132

so=0.44098

February

ms=0.38577

ss=0.46343

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.2545

so=0.38978

March

ms=0.39879

ss=0.45454

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.15115

so=0.31024

April

ms=0.28684

ss=0.41848

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.21999

so=0.33895

May

ms=0.34862

ss=0.41774

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.14863

so=0.27106

June

ms=0.28648

ss=0.39846

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.17021

so=0.28538

July

ms=0.27065

ss=0.37095

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.17137

so=0.29233

August

ms=0.23934

ss=0.35399

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.12278

so=0.26383

September

ms=0.21727

ss=0.36542

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.17756

so=0.33579

October

ms=0.28324

ss=0.41417

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.22213

so=0.38452

November

ms=0.32911

ss=0.44411

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8m

o=0.36589

so=0.45122

December

ms=0.47505

ss=0.47387

obs

sim

Figure 4.3: Observed and simulated cloud cover distribution at Bondville, Illinois. Symbol mand s are the mean and SD values, correspondingly, for the observed (subscript “o”) and simulated(subscript “s”) data.

33

0 6 12 18 24−10

0

10

20

30

January

[oC

]

sim

obs

0 6 12 18 24−10

0

10

20

30

February

[oC

]

0 6 12 18 24−10

0

10

20

30

March

[oC

]

0 6 12 18 24−10

0

10

20

30

April

[oC

]

0 6 12 18 24−10

0

10

20

30

May

[oC

]

0 6 12 18 24−10

0

10

20

30

June

[oC

]

0 6 12 18 24−10

0

10

20

30

July

[oC

]

0 6 12 18 24−10

0

10

20

30

August

[oC

]

0 6 12 18 24−10

0

10

20

30

September

[oC

]

0 6 12 18 24−10

0

10

20

30

October

[oC

]

0 6 12 18 24−10

0

10

20

30

November

[oC

]

hour0 6 12 18 24

−10

0

10

20

30

December

[oC

]

hour

Figure 4.4: Observed and simulated daily cycles of air temperature in 12 months at Bondville,Illinois. Vertical bars represent ± one SD.

34

0 6 12 18 24−10

−5

0

5

10

15

20

25

30Air temperature

Hour

Te

mp

era

ture

[oC

]

sim

obs

Figure 4.5: Diurnally averaged observed and simulated air temperature over 10 years (1997-2006)at Bondville, Illinois. Vertical bars represent ± one SD.

0 5 10 15 20 250

0.02

0.04

0.06

0.08

0.1

0.12

mo=4.4468

so=2.5701

Wind speed histogram

Fre

qu

en

cy

Wind speed [m/s]

ms=3.6053

ss=2.4626

obs

sim

Figure 4.6: The histogram of hourly wind speed from the observed and simulated data over 10years (1997-2006) at Bondville, Illinois.

35

4.3 Within-canopy vertical variation

The multi-layer canopy-root-soil system model, MLCan, provides insights into the impact

of the vertical distributions of leaf area and root biomass [22], presented in Figure 2.6.

The leaf area density (LAD) affects radiation attenuation through the canopy and canopy

micro-climate, while root biomass distribution dictates patterns of water uptake through

the soil column. Figure 4.7 presents the mean diurnal vertical patterns of An, latent heat

(LE), sensible heat (H), total absorbed shortwave radiation including photosynthetically

active (PAR) and near-infrared (NIR) bands (Qabs), and stomatal conductance for vapor

(gsv) through the canopy of each crop over the month of August, 2005. For each crop,

the distribution of An and LE correspond closely to Qabs as PAR is the primary driver

of An, and Qabs provides the majority of the energy partitioned into LE and H [22]. In

addition, gsv is highest at the very top of the canopy where shortwave intensity is strongest.

However, deeper distributions of positive H compared to An, LE, and Qabs reflect the ability

of NIR to penetrate deeper in the canopy (Figure 4.7). Switchgrass has a smoother LAD

profile than that of miscanthus (See Figure 2.6), which in combination with its lower canopy

density (or LAI) (See Figure 2.5E), results in a more uniformly distributed radiation regime

through the canopy, similar to that of maize [22]. The much denser foliage of the miscanthus

canopy, in combination with an upper canopy LAD maximum at z/h = 0.75 (Figure 2.6),

where z is the vertical coordinate and h is the canopy height, results in greater extraction of

shortwave radiation in the upper third of the canopy, and a greater sink of CO2 and source of

energy relative to maize and switchgrass (Figure 4.7). The denser miscanthus canopy more

effectively shades the soil column below it, making the fraction of soil evaporation under

miscanthus much smaller than for either switchgrass or maize (Table 4.3). The greater leaf

area of miscanthus likewise increase interception of precipitation, and condensation, thereby

increasing subsequent canopy evaporation (Table 4.3).

36

Figure 4.7: Diurnally averaged profiles obtained from MLCan model simulation under presentclimate condition in August 2005 for photosynthetic rate An; latent heat LE; sensible heat H; totalabsorbed shortwave radiation included photosynthetically active and near-infrared bands Qabs; andstomatal conductance for vapor gsv for maize (left column- A1, B1, C1, D1, and E1), miscanthus(center column - A2, B2, C2, D2, and E2), and switchgrass (right column - A3, B3, C3, D3,and E3).

37

Table 4.3: Comparison of total evapotranspiration (ET) and its component contribution maize,miscanthus, and switchgrass under present climate condition in 2005.

Crops Maize Miscanthus Switchgrass

Total ET [mm] 380 588 498

Transpiration [mm] 302 (82.4%) 473 (80.5%) 402 (80.7%)

Canopy Evaporation [mm] 36 (9.4%) 98 (16.6%) 66 (13.3%)

Soil Evaporation [mm] 42 (8.2%) 17 (2.9%) 30 (6.0%)

4.4 Impacts of crop type on energy balance

The structural and ecophysiological differences in the three bioenergy crops examined here

have implications for changes in surface temperature and albedo. Here we examine the rel-

ative effects of each canopy cover on mean surface temperature and albedo under present

climate conditions (2005). The diurnal variations of mean canopy temperature under present

conditions over one month (Aug 2005) for miscanthus and switchgrass are slightly lower than

those of maize (Figure 4.8A). The largest differences between the mean canopy temperature

of maize, and that of miscanthus (≈0.9◦C) and switchgrass (≈0.5◦C) occurs at approxi-

mately noon, when all three crops achieve their maximum canopy temperatures. This can

be attributed to the increased dissipation of absorbed radiation by transpiration for mis-

canthus and switchgrass. Higher LAI for miscanthus result in a stronger decrease of mean

canopy temperature than switchgrass, relative to maize.

Diurnal variation of temperature difference between top layers and mean canopy for three

crops over the same time are also compared to evaluate their temperature attenuation

through the canopy (Figure 4.8B). Compared to other crops, maize shows a smaller temper-

ature fluctuation through the canopy due to lower LAI. Temperature difference between top

layers and mean canopy for maize, miscanthus, and switchgrass ranged from -0.2 to 0.4◦C,

-0.6 to 1.1◦C, and -0.3 to 1.1◦C, respectively.

However, diurnal variations of albedo during the daytime in August 2005 for the bioenergy

crops are higher than maize (Figure 4.8C). This is because the lower LAI of maize allows

more radiation to penetrate through the canopy to the soil which has a lower reflectivity

38

than the leaves. Predicted mean values of albedo for miscanthus, switchgrass, and maize are

0.237, 0.235, and 0.220, respectively.

−2

−1

0

1

[µm

ol m

−2s

−1]

D1 D2

−90

−60

−30

0

[W m

−2]

E1 E2

0

30

60

90

[W m

−2]

F1 F2

−0.1

0

0.1

0.2

0.3

0.4

[oC

] G1 G2

2 7 12 17 22

−0.04

−0.02

0

0.02

Time [hour]

[mol m

−2s

−1]

H1

2 7 12 17 22

Time [hour]

H2

18

20

22

24

26

28

30

A

[oC

]Leaf te

mpera

ture

Tl

2 7 12 17 22−1

0

1 B

Time [hour]

∆A

n∆

LE

∆H

∆T

l

∆g

sv

6 8 10 12 14 160.2

0.21

0.22

0.23

0.24

0.25

0.26C

Alb

edo

Time [hour]

Maize

Miscanthus

Switchgrass

Figure 4.8: Diurnal variation of mean canopy temperatures (A), temperature difference betweenthe top layers and mean canopy (B), and albedo during the day (C) for maize (black dot line);miscanthus (red dash line); and switchgrass (blue solid line) in August 2005. Diurnally averagedchange of net-canopy fluxes and variables obtained from the MLCan model with vertical barsrepresenting ± one SD over growing season of Photosynthetic rate ∆An (D); Latent heat ∆LE(E); Sensible heat ∆H (F); Leaf temperature ∆Tl (G); Stomatal conductance for vapor ∆gsv (H)for miscanthus (in red-D1, E1, F1, G1, and H1) and switchgrass (in blue-D2, E2, F2, G2, and H2).

39

4.5 Vegetation response to elevated CO2 and increased air

temperature

To understand how each crop responds to elevated CO2, and in particular the impact of

reduced stomatal conductance on canopy energy partitioning, the model was run for the

2005 “present climate” forcing with two different atmospheric CO2 concentrations: one

representing present conditions (370 ppm) and one representing projected concentrations for

the year 2050 (550 ppm). Figure 4.8(D−H) show the diurnally averaged change of several

net-canopy fluxes for miscanthus and switchgrass over four months of growing season (Jun -

Sep), with each change (∆) representing the difference between the 550 ppm and 370 ppm

simulations. The diurnally averaged net-canopy fluxes over this same period under the 2005

climate forcing and atmospheric CO2 concentration of 370 ppm is presented in Figure 4.9

for comparison. The model’s ability to incorporate ecophysiological acclimation of reduced

stomatal conductance but no structural and biochemical acclimation (see Figure 2 in [22])

for these C4 crops [24, 27] to elevated CO2 results in only small changes in An (< 2%)

for both miscanthus and switchgrass. However, reduced stomatal conductance causes a

decrease in LE and a corresponding increase in H for both crops, with peak decreases in

LE of 56 and 50 W·m−2, and an increase in H of 54 and 52 W·m−2 for miscanthus and

switchgrass, respectively. The peak reduction in stomatal conductance gsv is -0.022 and

-0.025 mol·m−2·s−1 for miscanthus and switchgrass, respectively. Figure 4.8(D−H) further

shows that the diurnal variability of the net-canopy flux changes under elevated CO2 for

switchgrass are larger than those for miscanthus. These ecophysiological changes imply a

reduction in water loss through ET under elevated CO2. All three crops show a consistent

decrease in total ET ranging from 40 to 70 mm over one growing season (Table 4.4).

To evaluate the role of temperature on ET change, the model was run for the 2005 “present

climate” with projected atmospheric CO2 concentration (550 ppm), but for three scenarios

of increased air temperature during the summer, ranging from 1 to 3◦C [33]. In contrast to

the results with a modification only to CO2 concentration, as air temperature increases in

a higher CO2 environment, the advantage of reduced water use is lost due to the increase

of total ET (Table 4.4). The reason is that higher air temperature not only increases water

40

Table 4.4: Comparisons of total evapotranspiration (ET) alterations (mm) under climate changebetween maize, miscanthus, and switchgrass. ET under elevated CO2 and at different levels ofincrease of air temperature (Ta) are compared with ET under present condition for each crop.Elevated CO2 is set equal to 550 ppm.

Total ET change (mm) under elevated CO2

Crops Ta+0◦C Ta+1◦C Ta+2◦C Ta+3◦C

Maize -40 -10 15 40

Miscanthus -70 -17 19 53

Switchgrass -47 5 20 49

evaporation from the soil and canopy but also modulates transpiration rate through its effect

on vapor pressure [47].

4.6 Sensitivity to Seasonal Variation in Photosynthetic Capacity

Considerations for structural, biochemical, and ecophysiological acclimation responses under

elevated CO2 have been made following the methodology of Drewry et al. [23]. For C4 crops,

the simulations are performed for a constant value for Vmax for the growing season (Table 4.1).

We have examined the impact of seasonal variations in Vmax for maize following observations

presented in a study by Markelz et al. [48]. We assumed that the beginning and end of the

growing season values are the same and correspond to the low value for the season with a high

value in the early half of the season (see Figure 4.10). Seasonality of Vmax has little impact

on the canopy fluxes (See Figure 4.11) and results in a small change in total ET (1.8%) and

specific surface runoff (1.7%) in comparison to the constant Vmax case (Table 4.3). Similarly,

simulations performed with a seasonally high but constant value of Vmax result in only small

changes in the fluxes for maize (See Figure 4.11 and Table 4.5). Data on seasonal variation

of Vmax for miscanthus and switchgrass is not available, but given the lack of any significant

response in maize, it is deemed that the results with constant value (Table 4.1) capture the

tendencies well.

41

−20

0

20

40

60

[µm

ol m

−2s

−1]

Miscanthus

A1

An

−20

0

20

40

60

[µm

ol m

−2s

−1]

Switchgrass

A2

An

−100

100

300

500

[W m

−2]

B1

LE

−100

100

300

500

[W m

−2]

B2

LE

−100

100

300

500

[W m

−2]

C1

H

−100

100

300

500

[W m

−2]

C2

H

10

15

20

25

30

35

[oC

]

D1

Tl

10

15

20

25

30

35

[oC

]

D2

Tl

2 7 12 17 220

0.03

0.06

0.09

0.12

Time [hour]

[mol m

−2s

−1]

E1

gsv

2 7 12 17 220

0.03

0.06

0.09

0.12

Time [hour]

[mol m

−2s

−1]

E2

gsv

Figure 4.9: Diurnally averaged net-canopy fluxes and variables obtained from MLCan model withvertical bars representing ± one SD over one growing season of (A) Photosynthetic rate An; (B)Latent heat LE; (C) Sensible heat H; (D) Leaf temperature Tl; and (E) Stomatal conductance forvapor gsv for miscanthus (in red - A1, B1, C1, D1, and E1) and switchgrass (in blue - A2, B2, C2,D2, and E2) under present climate condition in 2005.

42

160 185 204 228 240 260

35

40

45

50

55

60

65

Day of Year

[µm

ol m

−2 s

−1]

Vm

ax

Vmax

Figure 4.10: To understand the impact of seasonality of leaf photosynthetic capacity on theresults presented earlier, Vmax for maize was varied as shown here. The values of Vmax from day185 to the end of the growing season was obtained and interpolated linearly from the study byMarkelz et al. [48]. We also assumed a linear increase of Vmax at the beginning of the season untilit reached the maximum value at day 185 [49,50].

4.7 Impacts on hydrology

Changes in weekly mean water balance components for the three crops are compared under

the same weather condition of the year 2005 (Figure.4.12). The water use of miscanthus

is the highest while that of maize is the lowest further reflecting the role of ET as a key

determinant of the water balance. Transpiration is the largest component of the water

balance, accounting for more than 80% of total ET (See Table 4.4). A conversion from

maize to miscanthus or switchgrass will lead to a reduction in soil water storage and a

consequent reduction in specific runoff. Under present climate in 2005, the total decrease in

soil water storage are 115 and 63 mm for miscanthus and switchgrass, respectively, relative to

maize. The corresponding decrease in specific surface runoff are 24 and 6 mm, respectively.

43

−1

0

1

2

3

4

5

[µm

ol m

−2s

−1]

SEA − CTL

A1

[%]

∆A

n

−3

0

3

6

9

12

15

18

∆ A

n

−1

0

1

2

3

4

5

[µm

ol m

−2s

−1]

MAX − CTL

A2

[%]

∆A

n

−3

0

3

6

9

12

15

18

∆ A

n

−10

−5

0

5

10

15

20

25

[W m

−2]

B1

[%]

∆L

E

−4

−2

0

2

4

6

8

10

∆ L

E

−10

−5

0

5

10

15

20

25

[W m

−2]

B2

[%]

∆L

E

−4

−2

0

2

4

6

8

10

∆ L

E

−25

−20

−15

−10

−5

0

5

10

[W m

−2]

C1[%

]

∆H

−30

−20

−10

0

10

∆ H

−25

−20

−15

−10

−5

0

5

10

[W m

−2]

C2

[%]

∆H

−30

−20

−10

0

10

∆ H

−0.5

−0.25

0

0.25

[oC

]

D1

[%]∆

Tl

−2

−1

0

1

∆ T

l

−0.5

−0.25

0

0.25

[oC

]

D2

[%]

∆T

l

−2

−1

0

1

∆ T

l

2 7 12 17 22

0

0.02

0.04

0.06

0.08

Time [hour]

[mo

l m

−2s

−1]

E1

[%]

∆g

sv

2 7 12 17 22

0

10

20

30

40

∆ g

sv

2 7 12 17 22

0

0.02

0.04

0.06

0.08

Time [hour]

[mo

l m

−2s

−1]

E2

[%]

∆g

sv

2 7 12 17 22

0

10

20

30

40

∆ g

sv

Figure 4.11: Diurnally averaged change of net-canopy fluxes and variables obtained for maize dur-ing 2005 using the MLCan model with vertical bars representing ± one SD over one growing seasonof (A) Photosynthetic rate An; (B) Latent heat LE; (C) Sensible heat H; (D) Leaf temperature Tl;and (E) Stomatal conductance for vapor gsv. SEA represents the case with seasonal variation ofVmax as shown in Fig. S5, CTL represents the control case presented earlier and MAX representsthe situation when Vmax is set to a constant but at seasonally high value of 60 µmol·m−2·s−1.The differences between SEA and CTL cases are presented in red (A1, B1, C1, D1, and E1) whiledifferences between MAX and CTL are presented in blue (A2, B2, C2, D2, and E2). The rightaxes represent percentage change with respect to the maximum diurnally averaged value in thecorresponding CTL simulation for maize.

44

Table 4.5: Change of total evapotranspiration (ET ) and specific surface runoff (R) for the twocases shown in Figure 4.11. For CTL simulation, Total ET = 380 mm, and R = 42 mm (SeeTable 4.3 and Figure 4.12)

Simulations ∆ET [mm] ∆ET [%] ∆R [mm] ∆R [%]

SEA - CTL 6.8 1.8 - 0.7 1.7MAX - CTL 12.2 3.2 - 1.55 3.7

To capture the uncertainty associated with these estimates, the model was run using an

ensemble of 30 years of weather forcing obtained using a stochastic weather generator [32]. To

understand the possible range of variation for each projected climate scenario the weather

ensemble was modified to represent the conditions summarized in Table 2.1. Figure 4.13

shows the box plots of total ET, soil water storage change, and total specific surface runoff

if maize is replaced by miscanthus or switchgrass. Under projected climate change scenarios

our simulations demonstrate that there will be a decrease in both soil water storage and

specific surface runoff.

In the first scenario (elevated CO2) soil water storage and specific surface runoff decrease

the least, as increased transpiration loss due to denser canopies and longer growing seasons

is somewhat offset by reductions in stomatal conductance associated with ecophysiological

acclimation. The mean soil water storage decreases approximately 110 mm for miscanthus

and 40 mm for switchgrass. For mean total specific surface runoff, the decreases are 25 and

3 mm, respectively.

In the second scenario, as precipitation is decreased 15% in the summer [33], water storage

and surface runoff are further decreased, highlighting the role of reduced water availability.

In scenarios 3-5, air temperature is increased at 3 levels without any change in precip-

itation. We found that the decrease of total water storage and surface runoff is directly

dependent on the increase of air temperature. Mean soil water storage decreases ranged

from 160 to 240 mm and 70 to 120 mm for miscanthus and switchgrass, respectively, for

temperature increases ranging from 1 to 3◦C.

For scenarios 6-8, air temperature is increased at 3 levels along with the 15% reduction in

precipitation. Water storage decreases are slightly greater than those in scenarios 3-5 due

to the further reduction in water input.

45

The fractions of soil water storage and specific surface runoff change during the overlapping

and longer growing season (with respect to maize) in both bioenergy crops are also different.

For miscanthus, 87% of soil water storage change occurred during the overlapping period of

the growing season, and 13% due to water utilization during the longer growing season. For

switchgrass, these estimates are 83% and 17%, respectively. However, both crops showed a

92% decrease of specific surface runoff during the overlapping period of growing season, and

8% decrease in the longer growing season.

4.8 Discussion

National policies and economic viability is likely to foster a shift in agricultural practices

from maize to bioenergy crops such as miscanthus and switchgrass [51]. The differences of

miscanthus and switchgrass from maize in the density (LAI) and architecture (LAD) of above

ground foliage results in increased transpiration. This also facilitates larger interception of

rainfall and condensation leading to increased direct evaporation from the foliage. Attenua-

tion of radiation through the denser canopy reduces the radiation reaching the soil thereby

increasing the albedo as more light is reflected from the more reflective foliage and reducing

the soil evaporation. All these impact both the energy and water balance. Elevated atmo-

spheric CO2 concentrations lead to an ecophysiological reduction in stomatal conductance

of the C4 plants which results in a suppression of transpiration and a corresponding increase

in soil-water storage and specific runoff. However, this conservation is not sufficient to offset

the reductions induced by the land use change from maize to miscanthus and switchgrass.

When expected increase in air temperature and reductions in rainfall are further included,

the conservative advantage of elevated CO2 is lost resulting in large reductions of both soil

water storage and specific runoff. Miscanthus, by virtue of its significantly larger LAI, has

the larger impact among the two bioenergy crops. Additionally, the longer growing season

adds to the net reduction in storage and runoff but this accounts for roughly 15% of the

change while the remainder is due to the structural differences in the vegetation foliage. As

illustrated in Figure 4.13, the increase in evapotranspiration, and reductions in soil water

storage and specific surface runoff is quite large, but these per unit area estimates need to

46

0

30

60

90

120

Maize

T E S S P+C

−30

0

30

60

90

dS/dt

0

15

30

R

0

30

60

90

120

mm

/week

Miscanthus

−30

0

30

60

90

mm

/week

0

15

30

0

30

60

90

120

Swichgrass

−30

0

30

60

90

21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 400

15

30

week

I

E

A

C

F

D

G

B

H

R E P

Figure 4.12: Weekly mean water balance, soil water storage change, and total specific surfacerunoff for 3 crops in several weeks in growing season 2005 under present CO2 condition (370 ppm).(A, D, G) Weekly mean water balance; (B, E, H) weekly change of soil water storage dS/dt;and (C, F, I) total specific surface runoff for maize, miscanthus, and switchgrass, respectively.Black solid lines in (A, D, G) represents the total weekly precipitation P and condensation C onfoliage (incoming water). Color boxes in (A, D, G) represent outgoing water components include:Transpiration TR (dark blue); Evaporation E (brown); Soil Evaporation SE (green); Seepage SP(cyan). Note that: the first 2 and last 4 weeks are outside of maize growing season.

47

300

400

500

600

700

ET

[m

m]

Total evapotranspirationA

-20

0

20

40

60

80

∆ E

T [%

]

-360

-240

-120

0

120

∆d

S/d

t [m

m]

Total water storage change

B

Maize

Miscanthus

Switchgrass

-40

-30

-20

-10

0

10

∆ d

S/d

t [%]

0

10

20

30

40

50

60

70

R [m

m]

Total specific surface runoffC

-90

-60

-30

0

30

60

∆R

[%]

Present climate

1. CO2

2. CO2 +PPT

3. CO2 +T1

4. CO2 +T2

5. CO2 +T3

6. CO2 +PPT+T1

7. CO2 +PPT+T2

8. CO2 +PPT+T3

Figure 4.13: Ensemble of total evapotranspiration (A), total soil water storage change (B), andtotal specific surface runoff (C) for maize (green), miscanthus (red), and switchgrass (blue) un-der present climate and eight combinations of climate change scenarios involving expected CO2concentration, higher air temperature, and altered precipitation magnitudes [33]. The right axesrepresent percentage change with the median value of maize in the present climate as baseline.Forcing data in each scenario is obtained from stochastically generated weather ensemble of 30independent years. Black dots represent total evapotranspiration, total soil water storage change,and total surface runoff from MLCan model for all crops that forcing data is only based on obser-vations in 2005 associated with climate change projections. Vertical solid and dash lines separatedifferent scenarios.

be factored in with the fraction of land use conversion to get estimates of watershed scale

impacts. For extensive areal alterations, the impact of bioenergy crops on the runoff and the

local environment from increased atmospheric humidity due to transpiration can be quite

significant. These issues should be weighed together with other environmental and energy

and economic benefits.

48

CHAPTER 5

CONCLUSIONS AND FUTURE WORKS

This thesis addresses the potential impacts of large-scale land cover change, possibly arising

from a shift in agricultural practices from maize to bioenergy crops such as miscanthus and

switchgrass, on the hydrologic cycle in Midwestern U.S. under climate change. Several model

modifications and significant improvements in model interfaces were performed to conduct

this research. The goal of this work was to integrate a vertically-resolved model of canopy

biophysical processes [22] with a stochastic Weather Generator [31, 32] and climate change

projections in the Midwest [33] for capturing the eco-physiological acclimations of bioenergy

crops and characterizing the variability range of crop responses to potential meteorological

forcing. The study revealed important insights into the water use of bioenergy crops and

impact on the hydrologic cycle which was characterized through the change in soil-water

storage and specific runoff.

Throughout this study, a mechanistic multi layer canopy-root-soil model (MLCan; [22])

was applied to contrast the eco-hydrological response of maize, the main feedstock for current

starch-based biofuel production, with miscanthus and switchgrass. For each crop, MLCan

model resolves vertically the light and thermal regimes within the canopy and provide pre-

dictions of photosynthetic CO2 assimilation and respiratory losses, as well as latent and

sensible heat fluxes for each canopy layer, through consideration of energy balance for both

sunlit and shaded leaf fraction. Water storage on foliage, as a function of dew or rainfall in-

terception, and soil moisture states are likewise considered to estimate water use, soil-water

storage change, and specific surface runoff.

The major limitation of the MLCan model was the lack of a friendly graphic user interface

(GUI) used for simplifying the modeling steps. This shortcoming was fully resolved by a

GUI developed and linked to the original model in MATLAB that can be executed in many

49

operating systems. The GUI takes advantage of built-in functions and Image Processing

packages in MATLAB to provides a visualization tool with plotting capabilities for the

MLCan model. Our tests for different crops demonstrated that the GUI may contribute

significantly to the MLCan simulations by ways of reducing time for model processing and

avoiding error-prone processes in the text modes.

The stochastic Weather Generator (WG; [31,32]) was used to provide an ensemble forcing

that captures the range of climate variability observed in recent past. The WG was also

included climate change projections in the Midwest [33] to provides Monte Carlo simulation

at hourly scale for: precipitation, total cloud cover, incoming shortwave and longwave radi-

ations, air temperature, humidity, and wind speed. Those variables are then interpolated to

obtain half-hour values and used for driving the MLCan model.

Lignocellulosic bioenergy is considered as a potential alternative for fossil fuel in term of

energy, economic, and environmental advantages. However, higher biomass production of

these bioenergy crops also require more water for their physical and biological functioning.

The work contained in this thesis showed that, under present climate in 2005, miscanthus

and switchgrass utilized more water than maize for total seasonal ET by approximately

58% and 36%, respectively. Elevated atmospheric CO2 concentrations (550 ppm) lead to an

eco-physiological reduction in stomatal conductance of the C4 plants which results in a sup-

presion of transpiration and a corresponding increase in soil-water storage and specific runoff.

However, this conservation is not sufficient to offset the reductions induced by the land use

change from maize to miscanthus and switchgrass, resulted from significant differences in

the density (LAI) and architecture(LAD). When expected increase in air temperature and

reductions in rainfall are further included, the conservative advantage of elevated CO2 is lost

resulting in large reduction of both soil water storage and specific runoff, highlighting the

critical role of warming air temperature in climate change.

A view from the multi layers canopy root soil model (MLCan) showed that the increase

in ET and reduction in soil water storage and specific runoff is quite large, but these per

unit area estimate need to be factored with fraction of land use conversion to get estimate

of water scale impacts. For extensive areal alteration, it is important to recognize that we

have to prepare for the impact of bioenergy crop on runoff and the local environment for

50

achieving sustainable development.

In the next step of this study, we will expand the application of MLCan model for bioenergy

crops to larger spatial scales. The integration of MLCan with watershed models will provide

important insights to watershed scale impacts of land use conversion as well as climate

change to hydrologic cycle.

51

CHAPTER 6

REFERENCES

[1] Ragauskas AJ, et al. (2006). The path forward for biofuels and biomaterials. Science311:484–489.

[2] Wall JD, Harwood CS, Demain AL (2008). Bioenergy. (ASM Press, Washington, DC).

[3] Kladivko EJ (2001). Tillage systems and soil ecology. Soil Till Res 61:61–76.

[4] Donner SD (2007). Surf or turf: A shift from feed to food cultivation could reducenutrient flux to the Gulf of Mexico. Global Environ Chang 17:105–113.

[5] Schnoor JL, et al. (2008). Water implications of biofuels production in the United States.(National Academies Press, Washington, DC).

[6] Costello C, Griffin WM, Landis AE, Matthews HS (2009). Impact of biofuel cropproduction on the formation of hypoxia in the Gulf of Mexico. Environ Sci Technol43:7985–7991.

[7] Tilman D, Hill J, Lehman C (2006). Carbon-negative biofuels from low-input high-diversity grassland biomass. Science 314:1598–1600.

[8] Cai X, Zhang X, Wang D (2011). Land Availability for Biofuel Production. EnvironSci Technol 45:334–339.

[9] Schmer M, Vogel K, Mitchell R, Perrin R (2008). Net energy of cellulosic ethanol fromswitchgrass. Proc Natl Acad Sci USA 105:464–469.

[10] Hess JR, Thomas DF, Hoskinson R, Thompson D (2003). Roadmap for agriculturebiomass feedstock supply in the United States. DOE/NE-ID-11129. (U.S. Departmentof Energy, Washington, DC).

[11] Heaton EA, Dohleman FG, Long SP (2008). Meeting US biofuel goals with less land:the potential of Miscanthus. Glob Change Biol 14:2000–2014.

[12] Heaton EA, Voigt T, Long SP (2004). A quantitative review comparing the yields of twocandidate C4 perennial biomass crops in relation to nitrogen, temperature and water.Biomass Bioenergy 27:21–30.

52

[13] Dondini M, Groenigen KJV, Galdo ID, Jones MB (2009). Carbon sequestration undermiscanthus: a study of 13c distribution in soil aggregates. Glob Change Biol Bioenergy1:321–330.

[14] Liebig MA, Schmer MR, Vogel KP, Mitchell RB (2008). Soil carbon storage by switch-grass grown for bioenergy. Bioenerg Res 1:215–222.

[15] Clifton-Brown JC, Breuer J, Jones MB (2007). Carbon mitigation by the energy crop,miscanthus. Glob Change Biol 13:2296–2307.

[16] Ma Z, Wood CW, Bransby DI (2000). Soil management impacts on soil carbon seques-tration by switchgrass. Biomass Bioenergy 18:469–477.

[17] Hickman GC, Vanloocke A, Dohleman FG, Bernacchi CJ (2010). A comparison ofcanopy evapotranspiration for maize and two perennial grasses identified as potentialbioenergy crops. Glob Change Biol Bioenergy 2:157–168.

[18] McIsaac GF, David MB, Mitchell CA (2010). Miscanthus and Switchgrass Productionin Central Illinois: Impacts on Hydrology and Inorganic Nitrogen Leaching. J EnvironQual 39:1790–1799.

[19] Vanloocke A, Bernacchi CJ, Twine TE (2010). The impacts of Miscanthus × giganteusproduction on the Midwest US hydrologic cycle. Glob Change Biol Bioenergy 2:180–191.

[20] Pieruschka R, Huber G, Berry JA (2010). Control of transpiration by radiation. ProcNatl Acad Sci USA 107:13372–13377.

[21] Sauer TJ, Singer JW, Prueger JH, DeSutter TM, Hatfield JL (2007). Radiation balanceand evaporation partitioning in a narrow-row soybean canopy. Agric Forest Meteorol145: 206-214.

[22] Drewry DT, et al. (2010a). Ecohydrological responses of dense canopies to environmen-tal variability: 1. Interplay between vertical structure and photosynthetic pathway. J.Geophys. Res. 115, G4, G04022.

[23] Drewry DT, et al. (2010b). Ecohydrological responses of dense canopies to environmen-tal variability: 2. Role of acclimation under elevated CO2. J. Geophys. Res. 115, G4,G04023.

[24] Leakey ADB, et al. (2006). Photosynthesis, productivity, and yield of maize are notaffected by open-air elevation of CO2 concentration in the absence of drought. PlantPhysiol 140:779–790.

[25] Long SP, Ainsworth EA, Rogers A, Ort DR (2004). Rising atmospheric carbon dioxide:Plants FACE the Future*. Annu. Rev. Plant Biol. 55: 591-628.

[26] Leakey ADB, et al. (2009). Elevated CO2 effects on plant carbon, nitrogen, and waterrelations: six important lessons from FACE. J Exp Bot 60: 2859-2876.

53

[27] Leakey ADB, Bernacchi CJ, Dohleman FG, Ort DR, Long SP (2004). Will photosynthe-sis of maize (Zea mays) in the US Corn Belt increase in future [CO2] rich atmosphere?An analysis of diurnal courses of CO2 uptake under free-air concentration enrichment(FACE). Glob Change Biol 10:95–962.

[28] Ghannoum O, von Caemmerer S, Ziska LH, Conroy JP (2000). The growth responseof C-4 plants to rising atmospheric CO2 partial pressure: a reassessment. Plant CellEnviron 23:931–942.

[29] Knapp AK, Hamerlynck EP, Owensby CE (1993). Photosynthetic and Water RelationsResponses to Elevated CO2 in the C4 Grass Andropogon gerardii. Int J Plant Sci154:459-466.

[30] Nobel PST (1980). Water Vapor Conductance and CO2 Uptake for Leaves of a C4Desert Grass, Hilaria Rigida. Ecology 6:252-258.

[31] Curtis DC, Eagleson PS (1982). Constrained stochastic climate simulation. Tech. Rep.274, Mass. Inst. of Technol., Dep. of Civ. and Environ. Eng., Ralph M. Parson Lab.,Cambridge.

[32] Ivanov VY, Bras RL, Curtis DC (2007). A weather generator for hydrological, ecological,and agricultural applications. Water Resour Res 43:W10406.

[33] Wuebbles DJ, Hayhoe K (2004). Climate change projections for the United StatesMidwest. Mitig. Adapt. Strategies Glob. Chang. 9:335–363.

[34] Campbell GS, Norman JM (1998). An introduction to environmental biophysics(Springer-Verlag, New York).

[35] Goudriaan J (1977). Crop micrometeorology: A simulation study (Cent. for Agric. Publ.and Doc., Wageningen, Netherlands).

[36] Spitters C (1986). Separating the diffuse and direct component of global radiation andits implications for modeling canopy photosynthesis. II: Calculation of canopy photo-synthesis. Agric. For. Meteorol. 38:231–242.

[37] Amenu GG, Kumar P (2008). A model for hydraulic redistribution incorporating cou-pled soil-root moisture transport. Hydrol Earth Syst Sc 12: 55–74.

[38] Dohleman FG, Heaton EA, Leakey ADB, Long SP (2009). Does greater leaf-level pho-tosynthesis explain the larger solar energy conversion efficiency of miscanthus relativeto switchgrass? Plant Cell Environ 32:1525–1537.

[39] Houghton JT, et al. (2001). Climate Change 2001: The Scientific Basis (CambridgeUniversity Press, Cambridge, U.K).

[40] Kromdijk J, et al. (2008). Bundle sheath leakiness and light limitation during C4 leafand canopy CO2 uptake. Plant Physiol. 148:2144–2155.

54

[41] Madakadze IC, et al. (1998). Leaf area development, light interception, and yield amongswitchgrass populations in a short-season area. Crop Sci. 38:827–834.

[42] Monti A, Zatta A (2009). Root distribution and soil moisture retrieval in perennial andannual energy crops in northern Italy. Agr Ecosyst Environ 132:252–259.

[43] Monteith JL, Unsworth MH (1990). Principles of environmental physics. (EdwardArnold Publishers, London).

[44] Schenk H, Jackson R (2002). The global biogeography of roots. Ecol. Monogr. 72:311–328.

[45] Durran DR (1999). Numerical Methods for Wave Equations in Geophysical Fluid Dy-namics. (Springer-Verlag, New York).

[46] Ivanov VY (2006). Effects of dynamic vegetation and topography on hydrological pro-cesses in semi-arid areas. Ph.D. Thesis, Mass. Inst. of Technol., Cambridge.

[47] Hopkins WG, Huner NPA (2009). Introduction to plant physiology (John Wiley & Son,Inc.).

[48] Markelz RJC, Strellner RS, Leakey ADB (2008). Impairment of C4 photosynthesis bydrought is exacerbated by limiting nitrogen and ameliorated by elevated [CO2] in maize.J Exp Bot 62:3235–3246.

[49] Wilson KB, Baldocchi DD, Hanson PJ (2000). Spatial and seasonal variability of photo-synthetic parameters and their relationship to leaf nitrogen in a deciduous forest. TreePhysiol 20:565–578.

[50] Xu L, Baldocchi DD (2003). Seasonal trends in photosynthetic parameters and stomatalconductance of blue oak (Quercus douglasii) under prolonged summer drought and hightemperature. Tree Physiol 23:865–877.

[51] Chen X, Huang H, Khanna M, Onal H (2011). Meeting the Mandate for Biofuels:Implications for Land Use, Food and Fuel Prices. National Bureau of Economic ResearchWorking Paper No. 16697 (Natl. Bureau Econ. Res., Cambridge, MA).

[52] Dermody O, Long S, DeLucia E (2006). How does elevated CO2 or ozone affect the leafarea index of soybean when applied independently? New Phytol 169:145–155.

[53] Dermody O, O’Neill B, Zangerl A, Berenbaum M, DeLucia E (2008). Effects of ele-vated CO2 and O3 on leaf damage and insect abundance in a soybean agroecosystem.Arthropod-Plant Inte 2:125–135.

[54] Grayman WM, Eagleson PS (1969). Streamflow record length for modeling catchmentdynamics. Tech. Rep. 114, Mass. Inst. of Technol., Dep. of Civ. and Environ. Eng.,Ralph M. Parson Lab., Cambridge.

55

[55] Eagleson PS (2002). Ecohydrology: Darwinian Expression of Vegetation and FormFunction. (Cambridge University Press, Cambridge, U.K).

[56] Gueymard C (1989). A 2-band model for the calculation of clear sky solar irradiance,illuminance, and photosynthetically active radiation at the earth’s surface. Sol. Energy43, 253–265.

[57] Becker S (2001). Calculation of direct solar and diffuse radiation in Israel. Int. J. Clim.21, 1561–1576.

[58] Thomas GE, Stamnes K (1999). Radiative Transfer in the Atmosphere and Ocean.(Cambridge University Press, Cambridge, U.K).

[59] Slingo A, Schrecker HM (1982). On the shortwave radiative properties of stratiformwater clouds. Q. J.R. Meteorol. Soc 108, 407–426.

[60] Stephens GL (1978). Radiation profiles in extended water clouds. 2. Parameterizationschemes. J. Atmosph. Sci. 35, 2123–2132.

[61] Slingo A (1989). A GCM parameterization for the shortwave radiative properties ofwater clouds. J. Atm. Sci. 46, 1419–1427.

[62] Han Q, Rossow WB, Lacis AA (1994). Near-global survey of effective droplet radii inliquid water clouds using ISCCP data. J. Clim. 7, 465–497.

[63] Kimball JS, Running SW, Nemani R (1997). An improved method for estimating surfacehumidity from daily minimum temperature. Agric. For. Meteorol. 85, 87–98.

[64] Priestley CHB, Taylor RJ (1972). On the assessment of surface heat flux and evaporationusing large-scale parameters. Mon. Weather Rev. 100, 81–92.

[65] Brutsaert W (1982). Evaporation into the Atmosphere : Theory, History, and Applica-tions. (D. Reidel, London).

[66] Farage PK, Blowers D, Long SP, Baker NR (2006). Low growth temperatures modifythe efficiency of light use by photosystem II for CO2 assimilation in leaves of twochilling-tolerant C4 species, Cyperus Longus l. and Miscanthus × Giganteus. PlantCell Environ, 29:720–728.

[67] Clifton-Brown JC, Neilson B, Lewandowski I, Jones MB (2000). The modelled produc-tivity of Miscanthus × Giganteus (Greef et deu) in Ireland. Ind Crop Prod, 12:97–109.

[68] Maass A, Hufschmidt MM, Dorfman R, Thomas Jr, HA, Marglin SA, Fair GM (1962).Design of water-resources systems. (Harvard Univ. Press, Cambridge, Mass).

56

APPENDIX A

WEATHER GENERATOR

The Weather Generator (WG) parameter values estimated from observed meteorological

data in 10 years (1997-2006) at Bondville, Illinois are presented briefly in this appendix.

The precipitation model is independent, and parameters used for this model are presented

in Table A.1. Parameters for precipitation model is in seasonal time-scale. The parameters

used for wind speed model is presented in Table A.2. The parameters used for cloud cover,

air temperature, and humidity model in monthly time-scale is presented in Table A.3.

The formulation of the WG shown in this appendix is based on published sources [31,32],

and is also presented here for completion and a quick look at the important formulation. For

details, the reader is referred to [31,32]. The WG simulates precipitation, total cloud cover,

shortwave radiation, air temperature, atmospheric humidity, and wind speed. With these

variables, it is possible to drive detailed models of the water and energy balances in multilayer

canopy-root-soil model (MLCan) used in this study. The diurnal cycle is important for all

these quantities, so hourly simulations are required. In a consistent climate most of these

variables are physically and thermodynamically related. A sound weather generator must

preserve most of such interactions to maintain consistency and realism. The dependence of

variables in the WG is presented in Figure A.1.

Table A.1: Parameters for precipitation model. See Table 4.2 for more information.

ParametersSeason 1 Season 2 Season 3 Season 4 Season 5 Season 6 Season 7

(Dec-Feb) (Mar) (Apr) (May-Jun) (Jul-Aug) (Sep) (Oct-Nov)

µr [hour] 6.081 5.173 4.493 3.815 3.529 4.529 5.339

µb [hour] 75.300 50.193 46.161 33.633 52.698 70.979 52.354

µd [mm] 5.010 4.482 4.926 4.782 6.900 6.026 4.812

57

Precipitation (PPT )

(μb, μr, μd)

Cloud cover (N )

(PPT, M0 , ρm , σm , , a, b)

Shortwave (SW )

(lat, long, Z, u0 , β, ωAi , ρg , LWP0 , N )

Temperature (Ta )

(ρg , SW, δT , ρδT , σδT )

Longwave (LW )

(Ta , N )

Dew Point Temperature (Td )

(Ta , P*ann

)

Wind speed (U )

(Ws, σs2, ρs(1), ϒs )

Figure A.1: Flowchart of the dependence of variables in the Weather Generator. The precipitation(PPT) model is generated independently. The cloud cover (N) is considered as a function ofprecipitation. The attenuation of shortwave radiation (SW ) is strongly affected by the cloud cover.The air temperature (Ta) is then generated as a function of incoming shortwave radiation. Thelongwave radiation (LW ) and dew point temperature (Td)are functions of the air temperature. Thewind speed (U) model is independent.

Table A.2: Parameters of the wind speed model. See Table 4.2 for more information.

Parameters Ws[m·s−1] σs[m·s−2] ρs(1)[-] γs[-]

Values 4.447 2.570 0.940 1.016

58

Table

A.3:

Par

amet

ers

for

clou

dco

ver

,ai

rte

mp

erat

ure

,an

dhu

mid

ity

mod

el.

See

Tab

le4.

2fo

rm

ore

info

rmat

ion.

Mon

th

Clo

ud

cove

r(N

)A

irte

mp

erat

ure

(Ta)

-D

eter

min

isti

cTa

-R

and

omva

riat

e

M0

σ2 m

ρm

(1)

b 0b 1

b 2b 3

b 4δT

σ2 δT

ρδT

(1)

P∗ ann

[-]

[-]

[-]

[-]

[-]

[-]

[-]

[-]

[◦C

][◦

C]

[-]

[mm

]

Jan

uar

y0.4

100.

212

0.81

0-1

.432

0.02

22.

004

8.01

03.

291

-0.1

560.

989

-0.1

5611

0

Feb

ruar

y0.

331

0.1

940.7

74-1

.625

0.03

31.

865

7.75

33.

718

-0.1

220.

983

-0.1

2280

Mar

ch0.2

550.1

520.8

00

-2.6

250.

054

1.85

210

.064

6.45

4-0

.100

1.00

0-0

.100

80

Ap

ril

0.1

510.

096

0.71

9-3

.220

0.07

41.

700

12.7

628.

453

-0.1

130.

962

-0.1

1360

May

0.22

00.

115

0.6

95-2

.420

0.07

81.

463

11.4

617.

072

-0.0

740.

957

-0.0

7465

Ju

ne

0.14

90.

074

0.6

00-1

.614

0.08

01.

402

10.8

455.

740

-0.0

680.

950

-0.0

6875

Ju

ly0.1

700.0

810.6

31-1

.886

0.11

71.

491

11.0

718.

400

-0.0

060.

930

-0.0

0695

Au

gust

0.1

710.0

860.

650

-1.4

820.

120

1.72

711

.216

7.44

90.

008

0.94

40.

008

60

Sep

tem

ber

0.12

30.0

700.6

84-2

.591

0.11

92.

244

16.5

389.

120

0.03

50.

947

0.03

510

0

Oct

ob

er0.1

780.

113

0.76

2-2

.591

0.11

92.

244

16.5

389.

120

-0.0

210.

966

-0.0

2110

0

Nov

emb

er0.2

220.1

480.8

16-2

.405

0.07

41.

918

18.0

536.

689

0.22

30.

976

0.22

313

0

Dec

emb

er0.

366

0.20

40.

809

-1.9

140.

048

2.08

914

.370

4.78

50.

291

0.98

80.

291

150

59

A.1 Precipitation

Precipitation is the most crucial meteorological variable for many applications. Its pres-

ence or absence also affects statistics of many other hydrometeorological variables. Storm

durations and interstorm times could be treated as independent events [54]. Thus, the pre-

cipitation model can be expressed by successive sampling from the fitted probability density

functions. For example, time between two successive storms tb [hour] follows the exponential

distribution (by definition of the Poisson arrival process):

f(tb) =1

µbe− 1µbtb (A.1)

where µb [hour] is the mean time between storms. The storm duration tr [hour] is also

simulated using the exponential distribution [54]:

f(tr) =1

µre−

1µrtr (A.2)

where µr [hour] is the mean storm duration. Grayman and Eagleson [54] showed that storm

depth were highly dependent on storm durations. Storm depth h [mm] were found to follow

a gamma distribution conditioned by storm duration:

f

(h| trµr

)=

1

µd

hµd

trµr−1e− h

µd

Γ( trµr

)(A.3)

where µr [mm] is the mean storm depth. The two-paramter gamma distribution has been

shown to satisfactorily preserve the rainfall statistical characteristics [31].

The following procedure is used to simulate rainfall. At some initial time t0, an interstorm

duration tb is generated from the fitted exponential distribution. The period [t0, t0 + tb] is

considered dry. When the time reaches [t0 + tb], the storm duration tr is generated. Using

the value set for tr, a storm depth h is generated from the gamma distribution. The period

[t0 + tb, t0 + tb + tr] is then considered wet. When time reaches [t0 + tb + tr] the process is

repeated to determine the next storm-interstorm sequence. The adopted method assumes

the model of rectangular pulses that considers a uniform rainfall intensity throughout the

60

entire period tr.

A.2 Cloud cover

Cloud cover is defined as the fraction of the celestial dome covered by clouds. The cloudi-

ness process, N(t) [dimensionless], is therefore bounded by 0 (clear sky) and 1 (overcast).

Intermediate values can define a variety of hydrometeorological conditions, e.g., 0.2 - scat-

tered,..., 0.7 - broken, etc. Curtis and Eagleson [31] consider N(t) as a random nonstationary

process composed of intrastorm and interstorm periods. During the intrastorm period, the

expected value of the mean of the process is close to 1.0, while if N(t) is examined near the

middle of a sufficiently long interstorm period, the expected value is usually quite different

from 1.0. The central assumption made in this model is that there is a loosely centered

subregion, R0, around the midpoint of the interstorm period in which the process N(t) can

be assumed stationary [31]. By examining the first and second moment properties of the pro-

cess, it is concluded that the fair weather cloud cover process in this subregion is unaffected

by approaching or receding precipitation systems:

E(N(t)|tb)t∈R0 = E(N(t))t∈R0 = M0,

V ar(N(t)|tb)t∈R0 = V ar(N(t))t∈R0 = σ2m, (A.4)

where t is time, t ∈ R0, M0 [dimensionless] is the mean and σ2m [dimensionless] is the variance

of the fair weather N(t). Another major assumption is that there is a smooth transition of

mean and variance from the boundaries of fair weather period:

N(t) = M0 + (1−M0)(1− J(t)) +m(t)J(t) (A.5)

where J(t) is an assumed transition function and m(t) is the stationary sequence of correlated

deviations with E(m(t)) = 0 and V ar(m(t)) = σ2m, and autocorrelation function ρN(τ),

where t [hour] is the lag. The time varying conditional expectation and variance of cloud

61

cover under this assumption are obtained as [31]:

E(N(t)|tb) = M0 + (1−M0)(1− J(t))

V ar(N(t)|tb) = σ2mP

2(t) (A.6)

where t ∈ tb. The autocorrelation structure of the cloud cover process (Eq. A.5) is

not affected by J(t) and is identical to the autocorrelation function of the process m(t):

ρN(τ) = ρm(τ) [31]. The transition function J(t) is assumed to be of an exponential form:

J(t) = (1− e−ς(t−t0))(1− e−γ(t0+tb−t)) (A.7)

where ς and γ [hour−1] are decay coefficients controlling the transition rates from the bound-

aries (end/beginning of precipitation events) to/from the region R0. These rates are assumed

to be equal in the current model implementation. As follows from Eq. (A.7), for t ∈ R0:

limtb→∞

N(t) = M0 +m(t) (A.8)

J(t) reaches a value close to 1.0 for all reasonable values of the decay coefficients and

therefore the fair weather cloudiness is essentially simulated as:

limtb→∞

N(t) = M0 +m(t) (A.9)

The stationary deviations process, m(t), is taken to be a firstorder Markov process:

m(t) = ρm(1)m(t− 1) + ε(t)σm√

1− ρ2m(1) (A.10)

where ε(t) is a random deviate. In the model of Curtis and Eagleson [31], ε(t) is assumed

to be a normally distributed variable, N(0, 12). However, a modification is introduced here

and ε(t) is simulated as a random deviate from the Beta probability distribution.

Beta(y) =Γ(a+ b)

Γ(a)Γ(b)

1

(y2 − y1)a+b−1(y − y1)a−1(y2 − y)b−1 (A.11)

62

where y1 and y2 are the lower and upper bounds of the Beta distribution for independent

variable y and a and b are the distribution shape parameters (a > 0, b > 0). The lower and

upper bounds (y1, y2) are found from Eq. (A.5) and (A.10) at every simulation step, i.e., the

knowledge of m(t− 1) and J(t) at every time t allows one to derive y1 and y2 by imposing

the requirement for the cloud cover N(t) to be in the range [0, 1]. The Beta distribution

shape parameters a and b are estimated based on conditioning by the cloudiness value at

(t − 1) (see Ivanov et al. [32] for more detail). Sampling of deviate ε(t), performed in this

manner, allows one to avoid the truncation of values of N(t) and preserve the moments and

shape of the cloud cover distribution.

A.3 Shortwave radiation

For all practical purposes, the Sun radiates its energy at a constant rate. Outside the

atmosphere, at the mean solar distance, the beam irradiance, also known as the solar constant

So, is 1367 [W·m−2]. The Earth’s orbit is slightly eccentric and the Sun-Earth distance varies

throughout the year. To allow for the varying solar distance, a correction, a ratio of the

actual Earth-Sun to the mean Earth-Sun Distance, r [-], is introduced so that:

S ′o =Sor2

(A.12)

r = 1.0 + 0.017 cos

[2π

365(186− JDay)

](A.13)

where JDay is the Julian day. Several other variables defining the Sun’s position with

respect to a location on Earth are used in the following. See Ivanov et al. [32] for further

information.

The model considers two wide bands of solar spectrum: the ultraviolet (UV) and vis-

ible (VIS) band, BΛ1 [0.29 µm ÷ 0.70 µm], where ozone absorption and molecular scat-

tering are concentrated, and the infrared in near- and short-wavelength range (NIR), BΛ2,

[0.70 µm ÷ 4.0 µm], where water and mixed gases absorptions are concentrated. The spec-

trum separation into two bands facilitates the transmittance modeling of beam and diffuse

63

clear sky irradiances because overlap between scattering and selective absorption is lim-

ited [56]. In the presence of clouds, the chosen spectral limits are also convenient due to the

above differences in absorption properties by water droplets. Moreover, separate treatment

of these two bands allows one to explicitly compute PAR, which is used in the process of

leaf photosynthesis. According to Slingo [61], the energy contained in the two bands, BΛ1

and BΛ2, is 46.6% and 53.4% of So, respectively. These fractions are applied to obtain the

extraterrestrial flux densities in the two bands: SoΛ1 and BoΛ2 [W·m−2].

A.3.1 Clear Sky

Direct Beam Irradiance

It is assumed that direct rays entering the atmosphere encounter extinction processes, which

are limited to the effects of ozone (’O’), uniformly mixed gases (’G’), water vapor (’W’), and

aerosol absorption (’A’), and Rayleigh and aerosol scattering (’R’) [56]. For the transmit-

tances TX due to an extinction process X (X = ’O’, ’G’, ’W’, ’A’, ’R’), see Ivanov et al. [32]

for more detail. In this model, separate extinction layers are considered, so that the total

direct beam flux at the ground at normal incidence is

SµbΛi = SoΛi∏X

TX,i (A.14)

Sµb = SµbΛ1 + SµbΛ2 (A.15)

where T [-] is transmittances and i (i = 1, 2) is used to denoted the considered bands.

Diffuse Irradiance

The diffuse irradiance at the ground level is modeled as a combination of three components

IdX corresponding to (the X index) the two scattering layers (molecules and aerosols) and

to a backscattering process between ground and sky (See Ivanov et al. [32] for more detail).

64

The total diffuse flux at the ground is therefore:

SµdΛi =∑X

IdX,i (A.16)

Sd = SdΛ1 + SdΛ2 (A.17)

A.3.2 Cloudy Sky

Clouds alter transmission and reflection properties of the atmosphere. Therefore the effects

of cloud cover need to be carefully accounted for. Cloud parameterizations provided in

hydrological literature typically use empirical functions of the fractional cloud cover [57] or

specify fixed bulk properties. However, cloud radiative characteristics strongly depend on

cloud type, structure, and density [58]. Also, different spectral intervals exhibit distinct

absorption and scattering characteristics [59].

The framework used here relies on parameterizations developed by Stephens [60] and

Slingo [61]. On the basis of both observational and theoretical evidence, these studies infer

that the cloud radiative properties are mainly determined by the cloud total vertical liquid

water path, LWP [g·m−2]. Stephens [60] showed that the broadband optical thickness is

essentially the same for clouds of different types that have the same LWP . Slingo [61]

introduced an additional independent relationship between cloud radiative properties and

the effective radius of drop size distribution. The advantage of using LWP is that this

quantity can be obtained by satellite microwave radiometry [62] and allows the introduction

of seasonality effects in the cloud properties.

The approach developed by Stephens [60] is used here to derive the effective radius of drop

size distribution based on LWP. The four-band model of Slingo [61] is then used to derive

the cloud transmittances and reflectances for the incident clear sky direct SµbΛi and diffuse

SµdΛi fluxes (assumed to be incident on top of clouds).

65

Direct Beam Irradiance

The model uses the simulated cloud cover, N [0.0 ÷ 1.0], to differentiate between the cloud-

less, (1 - N), and cloudy, N , sky fractions. The direct beam flux from the clear sky fraction

is assumed to reach the ground surface unaltered. The cloudy sky fraction is assumed to

contain a homogeneous layer of clouds characterized by LWP [g·m−2]. The total direct beam

normal irradiance SµB′Λj in each of the bands j, j = 1,...,4 of Slingo’s [61] parameterization

is estimated as a linear combination of the fluxes from the clear sky and cloudy fractions of

the sky dome:

SµB′Λj = SµbΛi [(1−N) + TDBjN ]kjK

(A.18)

where TDBj [dimensionless] is the cloud transmissivity for the direct beam flux in band j,

kj is the respective fraction of solar irradiance at the top of the atmosphere in each band

[0.466, 0.320, 0.181, 0.033] (See Ivanov et al. [32]), and K = 0.466 if i = 1, K = 0.534 if

i = 2. The fluxes in the two bands BΛ1 and BΛ2 are then obtained as

SµBΛ1 = SµB′Λ1 (A.19)

SµBΛ2 =4∑j=2

SµB′Λj (A.20)

Diffuse Irradiance

As above, diffuse flux from the clear sky fraction is assumed to reach the ground surface

unaltered. The diffuse radiative flux at the cloud bottom may result from both the diffuse

and beam radiation incident at the cloud top. The total diffuse irradiance SµD′Λj in each of

the bands j = 1,...,4 is estimated as a linear combination of the fluxes from the clear sky

and cloudy fractions [32]:

SµD′Λj = (1−N)SdΛi +N [TDIRjSµbΛi + TDIFjSdΛi]

kjK

(A.21)

66

where TDIRj [dimensionless] is the diffuse transmissivity for direct incident radiation and

TDIFj [dimensionless] is the diffuse transmissivity for diffuse incident radiation. The fluxes

in the two bands BΛ1 and BΛ2 are then obtained as:

SDΛ1 = SB′Λ1 (A.22)

SDΛ2 =4∑j=2

SD′Λj (A.23)

A.4 Air temperature

The model of Curtis and Eagleson [31] is used to simulate the air temperature. It assumes

that the hourly air temperature, T (t) [◦C], is a sum of two variables: a deterministic air

temperature component T (t) and a random variate δT (t):

T (t) = T (t) + δT (t) (A.24)

A.4.1 Deterministic component T (t)

The deterministic component is built on an empirical method of Bryan [1964] that attributes

temporal variation of the air temperature to the divergence of radiative heat flux and the

divergence of eddy heat flux. In essence, an assumption is made that hourly temperature

increments can be regressed on several hydrometeorological variables:

dT (t)

dt= b0 + b1T (t) + b2K(t)s(t) + b3K(t)r(t) + b4q(t) (A.25)

where bi (i = 0, 1,, 4) are the regression coefficients, s(t) and r(t) are the variables of the Sun

position and geographic location, K(t) = 1− 0.65N2(t) is the radiation attenuation factor,

and q(t) is the estimate of incoming longwave radiation. Equation A.25 excludes several

terms of the original model of Curtis and Eagleson [31]: the ground temperature (not a

standard measurement variable and not readily available), and wind speed and direction

67

(assumed to have generally minimal contribution in modifying the air temperature). This

first-order differential equation can be solved if the initial temperature is provided. A solution

method for this equation is summarized in Curtis and Eagleson [31] and Ivanov et al. [32].

A.4.2 Random variate component δT (t)

The deterministic component in the model of Curtis and Eagleson [31] essentially represents

the expected temperature value. It cannot explain all of the temperature variability and

therefore the random deviate δT (t) [◦C] is introduced, defined as: δTo(t) = To(t) − T (t),

where To(t) [◦C] is the observed air temperature and T (t) [◦C] is the deviation component.

The deviations are assumed to be approximated by a first-order Markov process:

δT (t) = δT + ρδT (1)(δT (t− 1)− δT ) + εT (t)σδT

√1− ρ2

δT (1) (A.26)

where δT , σ2δT , and ρδT (1) are the mean, variance, and lag-1 value of autocorrelation of

random temperature deviates, respectively, and εT (t) is the standard normal deviate.

A.5 Humidity

On the basis of long-term records for a number of climatic regions, Kimball et al. [63]

proposed an empirical model that allows for the adjustment of daily dew point temperature

with respect to minimum air temperature using information on daily potential evaporation

and a degree of region aridity:

Tdew = Tmind[−0.127+1.121(1.003−1.444EF+12.312EF 2−32.766EF 3)+0.0006∆Td]−273.15

(A.27)

where ∆Td [◦C] is the amplitude of daily air temperature and EF [dimensionless] is the

evaporative factor, 0 ≤ EF ≤ 1 that is defined as [63]:

EF =1

Pann

EpρwDLH (A.28)

68

where Ep [kg·m−2·s−1] is the daily potential evapotranspiration, ρw [kg·m−3] is the water

density, DLH [s] is the day length (See Ivanov et al. [32]), and Pann [m] is the annual

precipitation. When estimating Ep following Priestley and Taylor [64], Kimball et al. [63]

assume that the ground surface has a seasonally constant albedo, daily net radiation can

be approximated using air temperature, and that ground heat flux is 10% of net radiation.

Kimball et al. [63] show that the model (A.27) improves estimates of Tdew based on Tmind.

The major difficulty in using Eq. (A.27) is to compute the daily values of Ep and ∆Td, since

both quantities have to be available at the beginning of each day when Tdew is estimated.

Both Ep and ∆Td, however, are determined by a number of variables simulated at the

hourly scale that cannot be readily predicted for the entire day. For instance, cloudiness

affects temperature estimation and since both models use random deviates, there is no exact

method to predict daily air temperature amplitude in advance. It is assumed that adjusted

values of Ep and ∆Td from the previous day, (d− 1), can be used for estimation of Tdew on

the current day d. Since cloud cover affects net radiation and therefore energy available for

ET, an adjustment factor based on the radiation attenuation factor, K(t), is introduced:

A(d) =K(d)

K(d− 1)=

1− 0.652N(d)

1− 0.652N(d− 1)(A.29)

where K(d−1) is the average factor value for the previous day and K(d) is the mean expected

value for the day of estimation. The use of “renewal” model of rainfall arrival allows for

the estimation of K(d): at any time between successive storms, both the end time of the

previous and beginning time of the next storm are known. The cloud cover model (A.5) is

thus used to estimate an expected value of cloudiness for the following day, from which the

factor A(d) is then estimated. Once A(d) is known, the expected values of Ep and ∆Td that

appear in equations (A.27) - (A.28) are approximated as

Ep(d) = A(d)Ep(d− 1)

∆Td(d) = A(d)∆Td(d− 1) (A.30)

Two other variables are required for the estimation of Tdew: Tmind and Pann. The problem

69

of computing Tmind is similar to the one described above. Air temperature at the hour

preceding sunrise can be usually associated with Tmind. Hence the air temperature from

(A.24) at the hour preceding sunrise is taken as Tmind and used in the estimation of Tdew.

Kimball et al. [63] used the mean annual precipitation Pann for all days throughout the year.

Here, it is assumed that a monthly basis is more appropriate since different months/seasons

have different degree of dryness. Therefore Pann is considered as a precipitation parameter

for each month, Pann∗.

A.6 Wind speed

To simulate the wind speed, Curtis and Eagleson [31] used the AR(1) model, where the

random term forces skewness on the results of the autoregressive model, leading to an ap-

proximately Gamma distribution of wind speed [68].

Ws(t) = Ws + ρs(1)(Ws(t− 1)− Ws) + εtσs√

1− ρ2s(1) (A.31)

where Ws [m·s−1], σ2s [m2·s−2], and ρs(1) [dimensionless] are the mean, variance, and lag-1

value of the autocorrelation of wind speed, respectively. And the term εt(t) is:

εt =2

γε

[1 +

γεψt6− γ2

ε

36

]3

− 2

γε(A.32)

where ψt is a standard normal deviate and γs(t) is the skewness of εt(t):

γε =(1− ρ3

s)γs(1− ρ2

s)1.5

(A.33)

where γs(t) is the skewness determined from the wind speed data.

70

APPENDIX B

SIMULATIONS FOR OTHER LOCATIONS INMIDWESTERN UNITED STATES

MLCan simulations for maize, miscanthus and switchgrass for growing season in the year

2005 at two other locations in the Midwest area: (i) Ozark, Missouri (38.74 ◦N, 92.20 ◦E)

and (ii) Brookings, South Dakota (44.34 ◦N, 96.84 ◦E) are presented in this appendix for

reference. First, results obtained from MLCan simulations for these two additional sites

and for Bondville, Illinois under present climate conditions are compared to understand the

role of meteorological forcing on the water use of energy crops in Midwestern U.S. In the

next sections, further details and descriptions of the two additional sites, input data, and

simulation results are presented.

A map showing the locations of Ameriflux towers located at Bondville, Ozark, and Brook-

ings is presented in Figure B.1. The Ozark site is located in the University of Missouri Baskett

Wildlife Research area, situated in the region of central Missouri. The site is uniquely located

in the ecologically important transitional zone between the central hardwood region and the

central grassland region of the U.S. (See official website at http://public.ornl.gov/ameriflux/

for details). The Brookings site is located in a private pasture, consisting of a mixture of C3

and C4 species actively used for grazing. Belonging to the Northern Great Plains Rangelands,

the grassland is representative of many in the north central United States, with seasonal win-

ter conditions and a wet growing season (See official website at http://public.ornl.gov/ameriflux/

for further details). For these two additional sites (Ozark and Brookings), the meteorological

forcing data is also collected at the nearby eddy-covariance flux towers in Ameriflux network

(See section 2.3.1 for further information). However, there is no observation of LAI data for

maize, miscanthus, and switchgrass in these locations, the biological data (LAI) is therefore

obtained from the study by Heaton et al. [11] and Drewry et al. [22] in central Illinois as

shown in Chapter 2.

71

All additional MLCan simulations for Ozark and Brookings sites are only conducted under

“present climate” (CO2 concentration is set to 370 ppm) in 2005 using validated parameters

for three crops presented in Table 4.1. Comparison of total ET and its component contribu-

tion maize, miscanthus, and switchgrass under present climate condition in 2005 for three

locations in the Midwest is presented in Table B.1. The results for Bondville are obtained

from Table 4.3.

IL IN

IA

KS

MI

MN

MO

NE

ND

OH

SDWI

0 150 300 450 km150

Bondville

Ozark

Brookings

105° W

100° W 95

° W 90

° W 85

° W 80

° W

35° N

40° N

45° N

50° N

Legend

States

Ameriflux sites

Figure B.1: Map of Ameriflux towers located Bondville, Illinois (red point); Ozark, Missouri(blue point); and Brooking, South Dakota (green point) in Midwestern U.S.

72

Table B.1: Comparison of total evapotranspiration (ET) and its component contribution maize,miscanthus, and switchgrass under present climate condition in 2005 at three locations. Validatedparameters shown in Table 4.1 are used to conduct simulations.

Crops Locations Maize Miscanthus Switchgrass

Bondville 380 588 498

Total ET [mm] Ozark 389 607 533

Brookings 378 541 462

Bondville 302 (82.4%) 473 (80.5%) 402 (80.7%)

Transpiration [mm] Ozark 306 (78.7%) 511 (84.2%) 435 (81.6%)

Brookings 292 (77.2%) 438 (81.0%) 369 (79.9%)

Bondville 36 (9.4%) 98 (16.6%) 66 (13.3%)

Canopy Evaporation [mm] Ozark 32 (8.2%) 82 (13.5%) 62 (11.6%)

Brookings 23 [6.1%] 77 (14.2%) 53 (11.5%)

Bondville 42 (8.2%) 17 (2.9%) 30 (6.0%)

Soil Evaporation [mm] Ozark 51 (13.1%) 14 (2.3%) 36 (6.8%)

Brookings 63 [16.7%] 26 (4.8%) 40 (8.6%)

73

B.1 Ozark Forest, Missouri

0

20

40

60

PP

T [m

m]

0

200

400

600

800

PP

T [m

m]

0

10

20

30

Rg [M

J m

−2 d

ay−

1]

0

10

20

30

40

Tem

pera

ture

[ oC]

100 140 180 220 260 300

0

2

4

6

8

10

Miscanthus

Switchgrass

Maize

LA

I [m

2 m

−2]

Day of year

A

B

D

C

E

Figure B.2: Key meteorological forcing data observed in 2005 at Ameriflux tower located in Ozark,Missouri. Meteorological data includes (A) Daily precipitation (PPT, black bars); (B) Cumulativeprecipitation (PPT, red line); (C) Daily global radiation (Rg, blue line); and (D) Mean day time airtemperature (Ta, magenta line). (E) Leaf Area Index (LAI) for the maize (green circles), miscanthus(green triangles), and switchgrass (green squares) canopies were not changed and obtained from [11,22] as shown in Chapter 2.

74

Figure B.3: Diurnally averaged profiles obtained from MLCan model simulation under presentclimate condition in Ozark, Missouri in August 2005 for photosynthetic rate An; latent heat LE;sensible heat H; total absorbed shortwave radiation included photosynthetically active and near-infrared bands Qabs; and stomatal conductance for vapor gsv for maize (left column- A1, B1, C1,D1, and E1), miscanthus (center column - A2, B2, C2, D2, and E2), and switchgrass (right column- A3, B3, C3, D3, and E3).

75

B.2 Brookings, South Dakota

0

20

40

60

PP

T [

mm

]

0

200

400

600

800

PP

T [m

m]

0

10

20

30

Rg [M

J m

−2 d

ay−

1]

0

10

20

30

40

Tem

pera

ture

[ oC]

100 140 180 220 260 300

0

2

4

6

8

10

Miscanthus

Switchgrass

Maize

LA

I [m

2 m

−2]

Day of year

E

C

A

B

D

Figure B.4: Key meteorological forcing data observed in 2005 at Ameriflux tower located inBrookings, South Dakota. Meteorological data includes (A) Daily precipitation (PPT, black bars);(B) Cumulative precipitation (PPT, red line); (C) Daily global radiation (Rg, blue line); and(D) Mean day time air temperature (Ta, magenta line). (E) Leaf Area Index (LAI) for the maize(green circles), miscanthus (green triangles), and switchgrass (green squares) canopies were notchanged and obtained from [11,22] as shown in Chapter 2.

76

Figure B.5: Diurnally averaged profiles obtained from MLCan model simulation under presentclimate condition in Brookings, South Dakota in August 2005 for photosynthetic rate An; latentheat LE; sensible heat H; total absorbed shortwave radiation included photosynthetically activeand near-infrared bands Qabs; and stomatal conductance for vapor gsv for maize (left column- A1,B1, C1, D1, and E1), miscanthus (center column - A2, B2, C2, D2, and E2), and switchgrass (rightcolumn - A3, B3, C3, D3, and E3).

77