WATER RESOURCES RESEARCH, VOL. 49, 2552–2572, doi:10.1002/wrcr.20189, 2013
Evaluating controls on coupled hydrologic and vegetation dynamics
in a humid continental climate watershed using a subsurface-land
surface processes model
Chaopeng Shen,1 Jie Niu,2 and Mantha S. Phanikumar2
Received 20 November 2012; revised 23 February 2013; accepted 8 March 2013; published 28 May 2013.
[1] Understanding key controls on hydrologic dynamics is important for effectively
allocating resources for data collection, reducing model dimensionality, and making
modeling decisions. This work seeks to elucidate the physical factors responsible for the
observed hydrologic patterns of a watershed in Michigan using an integrated hydrologic
model, Process-based Adaptive Watershed Simulator and Community Land Model
(PAWSþCLM). The model is tested using observed data for streamflows, soil temperature,
groundwater table depths, and satellite-based observations of evapotranspiration (ET) and
leaf area index (LAI). Numerical experiments are carried out to lump the effects of key
controls, including land use types, nitrogen levels, groundwater redistribution, and soil
texture, into different process indices. Using analysis of variance (ANOVA), we
quantitatively determine the strengths of these controls on ET, net primary production
(NPP), and other important variables. Groundwater flow is found to be the major control on
runoff and infiltration, with soil texture ranking next, while vegetation type and nitrogen
levels are found to dominate NPP, top soil temperature, and transpiration. Soil texture and
groundwater are found to have comparable influence on soil moisture, which is in
agreement with analysis of field data in the literature. All controls are found to colimit ET,
which serves as the nexus for ecosystem-hydrology interactions. From the simulation
results, we find that nitrogen significantly controls transpiration, through which it influences
other hydrologic fluxes. While there is room for improving descriptions of the nitrogen
cycle in the current version of CLM, these novel results call for an understanding of the
interplay between hydrology and biogeochemistry. Additional analysis shows that the
relative strengths of the controls examined in this work are fairly robust with respect to
changes in parameters and spatial resolution.
Citation: Shen, C., J. Niu, and M. S. Phanikumar (2013), Evaluating controls on coupled hydrologic and vegetation dynamics in a humid
continental climate watershed using a subsurface-land surface processes model, Water Resour. Res., 49, 2552–2572, doi:10.1002/wrcr.20189.
1.
Introduction
[2] Recently, there has been a surge of interest in identifying, understanding, and utilizing hydrologic and ecosystem patterns and controls in hydrologic predictions [Hwang
et al., 2012; McDonnell et al., 2007; Thompson et al.,
2011; Vivoni, 2012; Wagener et al., 2010]. Understanding
the physical processes that control the variability of hydrologic variables is crucial for predictions and management
[Joshi and Mohanty, 2010; Pan and Wang, 2009; Wilson
et al., 2004]. The recognition of key patterns and controls
can help reduce the dimensionality of the prediction problem [Sivapalan et al., 2011], facilitate predictions in unga1
Department of Civil and Environmental Engineering, Pennsylvania
State University, University Park, Pennsylvania, USA.
2
Department of Civil and Environmental Engineering, Michigan State
University, East Lansing, Michigan, USA.
Corresponding author: C. Shen, Department of Civil and Environmental
Engineering, Pennsylvania State University, University Park, PA 16802,
USA. ([email protected])
©2013. American Geophysical Union. All Rights Reserved.
0043-1397/13/10.1002/wrcr.20189
uged basins [Sivapalan, 2003], and promote an in-depth
understanding of the underlying principles. Topographyinduced groundwater flow leads to convergence of subsurface moisture, which crafts spatial patterns in hydrologic,
geomorphologic, and vegetation processes. In some systems, groundwater flow plays a vital role in ecosystem
functioning. However, despite intensive recent studies
[e.g., Ivanov et al., 2008; Maxwell and Kollet, 2008a,
2008b; Miguez-Macho and Fan, 2012], the relative importance of groundwater flow as compared to other controls
such as soil texture, land use, and vegetation characteristics
remains unclear.
[3] The coupling between hydrology and biogeochemistry has recently received growing attention [Chorover et
al., 2011; Lohse et al., 2009; Manzoni and Porporato,
2011; Riveros-Iregui et al., 2012]. Strong controls and
feedbacks have been identified between water, carbon, and
nitrogen biogeochemistry in various environments [Burt
and Pinay, 2005; D’Odorico et al., 2003; Lohse et al.,
2009; Schimel et al., 1997]. The strong relationship
between evapotranspiration (ET) and biological nitrogen
fixation (BNF) [Cleveland et al., 1999] has long been recognized, so has the correlation between nitrogen
2552
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
availability, successional stage, soil moisture, and topography [Robertson et al., 1988]. In ecosystems where vegetation growth is limited by nutrient availability [Galloway et
al., 2004; Wilson et al., 1999], higher inputs of nitrogen
are expected to lead to a relatively more active ecosystem
with higher productivity, thus altering ET and the hydrologic cycle. Presently, human activities are creating reactive N at unprecedented rates [Galloway et al., 2004],
leading to increased nitrogen deposition [Dentener et al.,
2006; Galloway et al., 2008]. The implications of this
excess nitrogen and its interplay with the carbon and water
cycles is worth detailed, mechanistic and integrated investigation. The relative importance of N is difficult to evaluate
from field surveys alone due to the various time scales of
the complex N biogeochemical reactions, its interactions
with the carbon cycle, and its covariance with climatic
input [e.g., see Burke et al., 1997]. On one hand, inclusion
of nitrogen dynamics in global land surface simulations is
relatively recent (see Bonan and Levis [2010], Reich et al.
[2006], and Sokolov et al. [2008] for the effects of the C/N
interactions), and the current hydrologic descriptions in
these climate scale models are very crude, oversimplifying
the subsurface and channel dynamics. On the other hand,
with the exception of very few (TOPOG-IRM [Vertessy et
al., 1996] and RHESSys [Band et al., 2001; Mackay and
Band, 1997; Tague and Band, 2004]), ‘‘traditional’’ watershed-scale models often do not consider any carbon/nitrogen dynamics. Understanding the relative strengths of the
water-carbon-nitrogen coupling helps identify unrecognized yet important linkages.
[4] Besides synthesizing and mining observed data for
correlations, one may also study the influences of separate
processes using the simulations of physically based hydrologic models (PBHMs), which are ‘‘derived deductively
from fundamental physical laws’’ [Beven, 2002]. Although
PBHMs may never match the full complexity of reality,
they have the potential to reveal patterns and pinpoint
cause-effect relations. To reconcile differences and similarities in formulations and to aid scientific understanding of
the complex processes involved, new models, approaches,
and model intercomparison exercises are all expected to
play an important role. Recent advances in PBHMs have
focused on the linkages of terrestrial fluxes, ecohydrology,
and the integral role played by surface and subsurface
water dynamics. For an incomplete list, see the papers by
Fatichi et al. [2012], Ferguson and Maxwell [2010],
Goderniaux et al. [2009], Ivanov et al. [2008], Kollet and
Maxwell [2008], Mackay and Band [1997], Miguez-Macho
et al. [2007], and Niu et al. [2013]. A novel hydrologic
model, Process-based Adaptive Watershed Simulator
(PAWS), was recently introduced by Shen [2009] and Shen
and Phanikumar [2010, hereinafter SP10]. PAWS was created with intermediate complexity, Geographical Information System (GIS) data interface and parameterization
functionalities, attempting to strike a balance between
physics and computational tractability at large scales. The
model efficiently solves the governing equations for major
hydrologic processes using some of the best available algorithms. By reducing the dimensionality of the fully threedimensional (3-D) subsurface problem using a quasi-3-D
saturated groundwater domain and one-dimensional Richards’ equation to approximate soil columns in every grid
cell, the model significantly reduces the computational
demand with little loss of physics. The computational efficiency of the PAWS model allows for long-term, large-scale
simulations and makes parameter estimation and uncertainty
analysis feasible. The PAWS model was tested extensively
(SP10) with laboratory ‘‘recharge’’ experiments, analytical solutions, idealized test cases, and numerical solutions from models
that solve the full 3-D Richards equation. The model achieved
good performance in simulating hydrology in a medium-sized
watershed in the U.S. midwest. Descriptions of land surface
processes, including energy balance and vegetation growth
cycles, were nonetheless simple in the original PAWS model.
[5] The National Center for Atmospheric Research
(NCAR) community land model (CLM), a process-based
land surface model, was developed from grassroots collaboration among climate scientists [Collins et al., 2006;
Dickinson et al., 2006; Lawrence et al., 2011; Niu and
Yang, 2007; Oleson et al., 2010; Sakaguchi and Zeng,
2009; Zeng et al., 1998, 2002]. The model now encompasses a comprehensive suite of land surface processes
including surface heat/momentum/vapor transfer, surface
radiation balance, snow/soil heat transfer and freeze-thaw
phase changes, photosynthesis and plant growth, as well as
carbon and nitrogen fluxes, mostly using process-based
descriptions. However, the flow modules in CLM, e.g., surface runoff and channel flow, subsurface flow, and characterization of geologic and soil properties, all tend to be
overly simplified as compared to other components. CLM
treats each computational element as independent columns,
neglecting the lateral fluxes and hence their spatial interactions. As discussed in Anyah et al. [2008], Maxwell and
Kollet [2008a], and Shen and Phanikumar [2010], the subsurface flow is an integral component of the hydrologic
cycle that significantly influences local water fluxes and the
climate through feedbacks. To further enhance the simulation capabilities of the PAWS model, we recently coupled
the flow processes in PAWS to the land surface processes
in CLM. This suite of modules brings together surface
flow, 3-D subsurface flow, carbon/nitrogen dynamics, and
land surface processes to offer a higher degree of model
realism.
[6] In this paper, we employ the PAWSþCLM model to
study the water-energy-nutrient controls on hydrology in a
humid continental climate watershed in the Great Lakes
region of North America. Although progress has recently
been made in understanding the hydrology of this region
with a focus on the impacts of land cover/land use changes
[Mao and Cherkauer, 2009] and projected climate change
on hydrology [Mishra et al., 2010], the applications of
PBHMs have the potential to offer additional insights into
fundamental processes. Our objectives in this paper are (a)
to quantify the relative importance of physical processes
including land cover, soil texture, land use and groundwater flow on water, energy, and carbon fluxes using a process-based hydrologic model and (b) to examine the
coupling strengths and mutual influences between hydrology and nitrogen biogeochemistry. The paper is organized
as follows. After a brief description of the PAWSþCLM
model, we present the results of extensive model testing
using field and satellite-based observations. Using the welltested model, we show the rich spatiotemporal patterns of
the hydrologic and vegetation dynamics in the basin. Then,
2553
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
we carry out numerical experiments to analyze the relative
importance of the controlling processes on various fluxes
and states and how they together produce the patterns.
2.
Model Description
[7] Since the mathematical and algorithmic details of the
PAWS model have already been presented elsewhere
(SP10), we will only provide a brief outline of the model
here followed by short descriptions of the coupling with
CLM. PAWS uses 3-D structured grids with the top layer
representing the ground surface and the overland flow
domain. Several key processes take place exclusively on
the ground surface (e.g., overland flow, vegetation interception, depression storage) while others (e.g., ET) extend
over the entire vertical depth of the soil column. To avoid
confusion, we define our terminology as follows: the term
‘‘cell’’ is used to refer to the unit of the 2-D horizontal discretization of the ground surface (this corresponds to the
‘‘grid cell’’ concept in CLM 4.0). We use the word ‘‘column’’ to denote the totality of the soil matrix and pore
spaces underneath a cell (under each cell, we only have one
column, with a single set of moisture/thermodynamic
states, as opposed to the possibility of multiple columns in
CLM); the word ‘‘layer’’ refers to the vertical discretization of the soil column.
2.1. Flow Modules
[8] The flow domain is divided into surface overland
flow, channel flow, unsaturated soil water flow, and saturated groundwater. Processes in these domains are linked
by the surface-unsaturated-saturated coupling method
described in SP10. The ground surface is partitioned into
the flow domain and the ponding domain. The ponding
domain is updated together with the soil water states. Runoff rate from the ponding domain to the flow domain is
determined by the Manning’s equation. Only water in
excess of the interception depth can become runoff. Under
flooding conditions, water in the flow domain may also
backfill into the ponding domain. For the overland flow domain, we solve the 2-D diffusive wave equation using an
efficient and stable Runge-Kutta finite volume (RKFV)
scheme. A similar method is used for the channel network,
which exchanges flow with the overland and the groundwater flow domains. The exchange between groundwater
and channel flow is calculated based on the leakance concept [Gunduz and Aral, 2005], after the diffusive wave
equation is solved. Wetlands are an important land cover
type that modifies hydrologic responses. A new lowlandstorage module is developed as part of this work and is
described in Appendix A (Figure 1 illustrates the conceptual model used by the module).
[9] The subsurface is discretized into a series of 1-D soil
columns connected to saturated groundwater layers at the
bottom. The saturated-unsaturated soil water flow is governed by the Richards equation, which is solved using a
modified Picard iteration approach [Celia et al., 1990; van
Dam and Feddes, 2000]. The vadose zone module handles
infiltration into the soil under normal conditions; however,
under heavy rainfall conditions, we employ a modified
form of the generalized Green and Ampt method [Jia,
1998]. We bring the dynamics of regional groundwater
Figure 1. Illustration of the lowland-storage module.
This compartment is conceptualized as a storage compartment of the overland flow domain.
flow into the soil columns by writing a separate mass balance equation for the last grid cell, whose thickness
changes as the water table fluctuates. The bottom of the last
cell extends to the top of the bedrock. Lateral flow
exchange calculated by the groundwater flow equation is
fed into the Richards equation, and the recharge flux
obtained from the solution of the Richards equation is in
turn passed to the groundwater flow equation as a source
term. As described in SP10, this coupling method produces
physically consistent (in that the soil moisture profile is
consistent with the groundwater head) and stable solutions
and compares well with solutions based on the fully 3-D
variably saturated Richards equation models for different
conditions (e.g., conditions corresponding to infiltration
excess, saturation excess, and return flow).
2.2. (PAWS1CLM): Coupled Subsurface and Land
Surface Processes
[10] The processes included in CLM have been detailed
in Oleson et al. [2010]. For the sake of completeness, we
provide a brief overview of CLM processes incorporated in
the coupled model in this section and then discuss linkages.
PAWS is coupled to the latest release of CLM (version
4.0), with a prognostic crop model [Lawrence et al., 2011].
The soil hydrology and river routing routines in CLM are
replaced by the corresponding procedures in the PAWS
model. The coupling details are described in section 2.2.2.
2.2.1. Introduction to CLM Processes
[11] For canopy radiative fluxes, the two-stream approximation of Dickinson [1983] and Sellers [1985] is used to
keep track of incoming, transmitted, reflected and absorbed
direct or indirect radiation, over multiple solar wavebands.
The plants are parameterized by detailed ecophysiological
descriptors to capture the optical variability among plant
species, especially trees. Canopy scaling/integration of sunlight penetration is based on the specific leaf area (SLA,
the ratio of leaf area to leaf mass) concept [Thornton and
Zimmermann, 2007], which is based on the assumption that
2554
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
SLA increases linearly as a function of overlying leaf area
index (LAI), and the two-big-leaf canopy model, which
treats the canopy as a sun-lit leaf and a sun-shaded leaf
[Dai et al., 2004]. CLM employs the resistance concept,
rather than the widely used Penman-Monteith (PM)
approach to calculate momentum/heat/vapor transfer.
Although the PM equation assumes a wet-bulb approximation for vegetation surface temperature, the resistance
approach explicitly solves for the leaf temperature to satisfy the coupled latent and sensible heat transfer equations.
Aerodynamic resistances are calculated based on the
Monin-Obukov similarity theory [Kundu and Cohen, 2010;
Zeng et al., 1998]. Canopy resistances are solved simultaneously with photosynthesis using the Farquhar model for
C3 plants [Farquhar et al., 1980] or the model of Collatz
[Collatz et al., 1992] for C4 plants. As classified in Arora
[2002], this is a biochemical approach that provides a process-based description of CO2 assimilation and may be
more suitable for the assessment of future climate scenarios. The water stress function couples hydrology with ecosystem dynamics and is parameterized as described in
section 2.2.2.
[12] Soil and snow temperatures are updated by solving
the heat conduction equation. The SNICAR (SNow, ICe,
and Aerosol Radiative model) module [Toon et al., 1989]
included in CLM 4.0 keeps track of the mass balance of
aerosols and impurities in the snowpack. Snow aging
impacts effective grain size and consequently snow albedo.
Soil freezing is calculated by a freezing-point depression
formulation of Niu and Yang [2006]. This formulation
allows supercooled water and ice to coexist in a wide range
of below-freezing temperatures. The soil hydraulic conductivities are scaled down by an impermeability factor
depending on the fraction of ice in the soil water content.
[13] The carbon and nitrogen cycle on the land surfaces
are fully incorporated with the biogeochemical vegetation
dynamics and phenology module of Biome-BGC [Thornton
and Rosenbloom, 2005; White et al., 1997]. The photosynthesized carbon is sent into a central carbon pool, which is
first spent on maintenance respiration by live compartments
(leaf, fine root, live coarse root, and live stem if applicable). Surplus carbon is then allocated to growth and phenological carbon pools (leaf, dead steam, root, and storage/
transfer pools) depending on the allometric relationships
prescribed for different plant species. This process is
strongly controlled by available nitrogen. Growth respiration is consumed when allocation occurs. Offsetting, gap
mortality, and fragmentation of coarse wood debris send
carbon and nitrogen into the litter pools. Carbon and nitrogen (C/N) then trickle down a cascade of litter/soil organic
matter (SOM) pools with varying turnover rates. The displayed structure of vegetation (LAI, tree height) is prognostically updated based on the carbon stored in different
pools. Different plant species are parameterized with different phenological traits (e.g., evergreen, stress-deciduous,
seasonal deciduous, etc). Nitrogen transformations are
book kept in plant pools, three litter pools, four SOM pools,
and a soil mineral pool. Simulated nitrogen-specific
dynamics include deposition, fixation, leaching, and uptake.
[14] It is worth mentioning that although CN (Carbon
and Nitrogen) module in CLM is a complex model with
respect to the number of pools as compared to many other
models [Manzoni and Porporato, 2009], there are still large
uncertainties with the biogeochemical cycling in CLM.
The CN module is a compartment model and does not track
chemical species (except for the distinction between cellulose and lignin pools), microbial community, age/residence
time, and vertical distribution of C/N, all of which can play
important roles [e.g., see Berg and Laskowski, 2005; Fang
et al., 2005; Maggi and Porporato, 2007; Maggi et al.,
2008; Wutzler and Reichstein, 2008; Xu-Ri and Prentice,
2008]. The allometric coefficients, base turnover rates
(base rates are then adjusted by temperature and water
functions), nitrogen deposition/fixation rates, and denitrification rates all face uncertainties. There are also uncertainties associated with the appropriate spatiotemporal scales at
which these rates should be applied [Manzoni and Porporato, 2009]. Data sets to validate the internal dynamics of
CN cycling models are generally lacking.
2.2.2. Coupling CLM With PAWS
[15] The coupling between PAWS and CLM is done in a
mass-conservative, theoretically consistent, and tightly
integrated fashion. At initialization, vertical and horizontal
discretization information in PAWS, which depends on topography, soil, and geology, is ported to CLM so that both
have the same definitions of computational units. At each
time step, climate forcing is passed into CLM to compute
interception, throughflow, surface radiation processes, surface energy balance, photosynthesis, soil temperature, and
snow processes. Then, the ET demand (both transpiration
and soil evaporation) as a function of depths is passed into
PAWS as a source term for the unsaturated zone. Also
passed along are the soil temperature, ice content, canopy
storage, ground precipitation, dew, and snowmelt amounts.
The soil hydrology module in CLM is replaced with its
PAWS counterpart, which then solves the Richards equation together with ET, groundwater flow, and runoff based
on the coupling scheme described in SP10. Overland flow
and streamflow are also calculated in PAWS. The resulting
soil moisture states are then converted into the soil water
state variables in CLM and supplied back into CLM for
ecosystem updates and calculations in the next time step.
[16] In this paper, we will focus on the results obtained
by running the (PAWSþCLM) model in the ‘‘offline’’
mode, in which the climate forcings are based on observations from conventional land-based stations. PAWS uses a
structured grid; therefore, the CLM discretization is configured such that there is a one-to-one mapping between a
PAWS cell and a CLM grid cell, and there is only one column in each CLM grid cell. At the same time, subcell heterogeneity on the plant functional types (PFT) level is
supported as PAWS also adopts a similar approach. In
SP10, this subcell heterogeneity of land use/land cover
types was called representative plant types (RPT). To conform to the CLM literature, we adopt the term PFT in this
paper. In the coupled model, PAWS provides the front-end
processing utilities that reclassify raw land use data sets
into CLM PFTs.
[17] Some changes to the CLM processes are necessary
to make the two models consistent in theory. Since the soil
water flow processes are primarily computed in PAWS, the
combined model uses the van Genuchten formulation for
soil water retention relationships as in the original PAWS
model. Field capacity, saturated water content, and wilting
2555
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Table 1. Model Calibration Parameters
Symbol (Unit)
K (m/day)
KS (m/day)
N
A (1/m)
ice
Kr (m/day)
l (m)
Parameter Meaning
Groundwater hydraulic conductivity
Soil saturated hydraulic conductivity
van Genuchten parameter
van Genuchten parameter
Parameter in Lai and Katul [2000] root efficiency
function
Scale-dependent freezing fraction parameter as in
equation (11) in Niu and Yang [2006]
River leakances
Length of flow path for runoff contribution to
overland flow domain
points are set by the PAWS model. The soils characterizations are derived from the SSURGO database (Soil Survey
Geographic database from the Natural Resources Conservation Service [Soil Survey Staff, 2010]), which is a high
resolution, comprehensive data set. The soil resistances to
evaporation are reformulated according to Neitsch et al.
[2005]. The root water uptake efficiency (or root resistance
in CLM) adopted by PAWS, that is, the Lai and Katul
[2000] formulation, is used. This formulation contains a
tunable parameter (Table 1) that modulates the root efficiency-water stress function. When coupled to atmospheric
models, the climate input data is normally forced at 30 m
above canopy. With land-based climatic stations, however,
the observation height is between 1 and 2 m. Therefore, the
aerodynamic resistances to vapor/heat/momentum transfer
are formulated using a neutral stability assumption. When
wind flow velocity is low (<1 m/s), free convection is
assumed, and the formulation in Ivanov [2006] and Kondo
and Ishida [1997] is employed.
[18] Climate-forcing data are passed into CLM from
PAWS for the calculation of energy and water fluxes. The
resulting ET fluxes are passed back to PAWS and converted into source terms for the Richards equation. We also
consider phase change in the subsurface. The freezingpoint depression formula in Niu and Yang [2006] is utilized
in the vadose zone model in PAWS to reduce the hydraulic
conductivities in freezing soils. The urban and dynamic
land cover modules in CLM 4.0 are not currently used.
[19] Some minor adjustments include the following: to
match with literature values in Hessl et al. [2004], the fine
root carbon/nitrogen ratio of broadleaf deciduous forest
(DBF) was changed from 45 to 60 and the leaf carbon/nitrogen ratio of temperate needleleaf evergreen forest was
changed from 25 to 45; both C3 grass and C3 crop have
been changed to seasonal deciduous to avoid a problem with
the CLM parameterization of stress-deciduous PFTs that lead
to spurious growth in brief periods of warmth in winter.
3.
Results
[20] In this section, we apply the PAWS model to a
medium-sized watershed located in the U.S. Laurentian
Great Lakes region. The model is first tested using observations and then utilized to elucidate the essential hydrologic
dynamics of this watershed. Although extensive validations
are provided to demonstrate that the new model can
describe key processes accurately, the focus of this paper is
in evaluating the relative strengths of controls on coupled
hydrologic and vegetation dynamics. After describing the
background hydrology, we use the model to study the relative importance of several controlling processes for surface
fluxes: land use and vegetation type, soil texture, topography-induced groundwater flow, and nitrogen levels.
3.1. Site Description and Input Data
[21] The Clinton River (CR) watershed (1837 km2) is situated in the mid-southern part of the lower peninsula of Michigan, as shown in Figure 2, and drains into Lake St. Clair.
Located in the humid continental climate (hot summer)
region, the basin experiences large seasonal temperature variations with very warm summer (average July temperatures
above 22 C) and cold winter (average January temperatures
around 3 C). Precipitation is evenly distributed in the year
with an annual average of approximately 900 mm, which is
slightly higher compared to nearby regions due to the lake
effect. Daily or subdaily weather data are obtained from the
National Climatic Data Center (NCDC) [2010]. The locations of the weather stations are shown in Figure 2a.
[22] This watershed is selected because of its diversity in
land use and topography. The topography of the watershed
mainly consists of rugged hills on the highlands of the west
and flat, low-lying plains in the east, joined by a steep eastfacing slope. The southwest corner forms semienclosed
subbasins with numerous small lakes. The 30 m resolution
national elevation data set (NED) is preprocessed to generate average cell elevation and lowland-storage bottom elevation in the computational grid. As shown in Figure 2b,
the watershed is urbanized in the southern portion with
varying extent of modification. The northern half is occupied by forest in the western quarter and agriculture in the
eastern quarter. The 30 m resolution IFMAP (Integrated
Forest Monitoring, Assessment, and Prescription) 2001
land use land cover data [Michigan Department of Natural
Resources, MNDR, 2010], which was classified from Landsat Thematic MapperTM imagery is used to provide land
use information. Three dominant land use types (PFTs) are
modeled in each horizontal cell. The soil color data are
extracted from the global data set [Global Soil Data Task,
2000].
[23] The surficial aquifer systems of the watershed are
mainly composed of Pleistocene glacial drift that overlies
the bedrock. Due to its proximity to Lake St. Clair, the
watershed is covered by less transmissive lacustrine finegrained sediment in the nearshore plain, while the highland
portion of the watershed is characterized by relatively more
stratified and permeable outwash. The glacial drift is
treated as the unconfined aquifer. Spatial fields of the conductivities (lateral) of glacial drift were obtained by interpolating well records from the WELLOGIC database
[Groundwater Inventory and Mapping Project (GWIM),
2006; Oztan, 2011; Simard, 2007] using kriging after noise
was removed. Also interpolated from the WELLOGIC
database are the aquifer thicknesses and static water table
heights to provide discretization information and comparison data for the model. The bedrock is mainly coldwater
shale, which is composed of shale and some limestone.
This is a confining unit layer with very low permeability
that produces little water yield. Lithology-based estimates
of bedrock conductivities (lateral) are rasterized and input
into the model. Lake St. Clair has been found to influence
2556
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Figure 2. (a) Map of the Clinton River watershed showing its location, elevation, weather stations,
USGS gages (labeled by eight digit numbers), and the hillslope transect (A-B). (b) Land use map of the
watershed.
groundwater flow in the basin. To better approximate reality, we enforce a constant head boundary condition at the
boundary face nearest to Lake St. Clair. The average lateral
conductivity of the unconfined aquifer in the basin is 13 m/
day.
[24] The model was spun-up by repeatedly looping
through the climatic forcing data between 2001 and 2009
over a span of 4000 years, as in Thornton and Rosenbloom
[2005]. To reduce the computational time required for
model spin-up, several typical sites in the watershed are
chosen. The spin-up was done in a ‘‘column mode,’’ in
which the phreatic water table depth is kept constant. It has
been verified that the ecosystem has approximately reached
equilibrium. The resulting carbon and nitrogen states and
other ecophysiological parameters are stored to be read in
at run time.
[25] A 880 m 880 m horizontal grid is used for the
computations on the landmass. Twenty vertical layers are
used for the unsaturated soil zone. The vertical discretization of the soil layer at a location depends on the local
2557
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Figure 3. Simulated daily streamflow compared with
USGS measurements at basin outflow gage (a, b) and an
inner gage (c).
thickness of the vadose zone and uses an adaptive approach
that assigns finer grid near the surface and coarser cells
near the bottom. Fourteen major rivers are simulated in the
model, all discretized with spatial grid sizes of approximately 1500 m (A flexible exchange scheme in PAWS
allows for flexible discretization of channel and land). The
land surface and subsurface processes run on a maximum
time step of 1 h. Overland flow and channel flow use a maximum of 10 min time steps. All components can adaptively
change step sizes as necessary. On average, a day of serialized simulation takes 2.5 s on an Intel 2.3 GHz i-7 CPU.
root transformed data (SP10). The rationale for using
MNASH is that the NASH coefficient tends to give too
much importance to the peaks (runoff) at the expense of
important subsurface processes (e.g., baseflow contribution
to streams). Use of the MNASH metric is expected to
address this deficiency of the original NASH coefficient.
Figure 3a shows the comparison between simulated and
observed daily hydrographs at the outflow USGS gage, CR
at Moravian Drive at Mt. Clemens, MI. Figure 3b shows a
close-up of a portion of the hydrograph. In Figure 3c, we
also show the hydrograph comparison at an inner gage, CR
at Sterling Heights, which is not involved in calibration.
The decent daily NASH values (0.61 and 0.65) at both
basin outlet and inner gages are satisfactory. Qualitatively,
the baseflow is captured very well. The model grossly
underestimates the largest peak in May 2004, while later in
June 2004 there is a separate, large simulated peak that
does not correspond to any observed peaks. We suspect
that some error with precipitation data has led to this temporal mismatch, which causes heavy penalty in NASH.
[27] Simulated depths to water table (averaged from
2007 to 2009) are compared with the measured values from
the WELLOGIC database in Figure 4 (the model is started
in 2001). As can be seen from the spatial fields and the
high R2 value (0.66), an overall good agreement is noted
between simulated and observed water table depths.
[28] Figure 5 shows the simulated soil temperature at a
local measurement site in Romeo, Michigan [Enviroweather, 2010]. The main deviation is in the winter of
2008, when the model predicts as low as 5 C while the
temperature sensor reported around 0. This is attributed to
a combination of factors, including potentially imperfect
soil freezing scheme, local conditions, and measurement
errors. Overall, the model was able to reproduce the major
dynamic cycle of soil temperature fluctuations. The formulations for soil freeze and thaw should continue to improve
as this process heavily influences surface water and energy
fluxes [Sinha and Cherkauer, 2008]. However, we need to
keep in mind that the simulated variable represents an average over the model grid cell (880 m 880 m) while the
3.2. Model Testing Against Observations
[26] The majority of the parameters in PAWS are spatially distributed (e.g., hydraulic conductivity or soil properties), and their spatial heterogeneity is honored in our
simulations. Modest adjustments are done to some parameters, such as the van Genuchten soil parameters, by applying a global multiplier to improve model performance.
Parameters adjusted are listed in Table 1. USGS (U.S. Geological Survey) gaging station 04165500 at Mt. Clemens,
which represents basin outflow, is selected to show the
comparison between model simulations and observations
for the watershed. A modified Nash-Sutcliffe coefficient,
MNASH is used as a model performance metric :
MNASH ¼ 1 NASH þ RNASH
;
2
where NASH is the standard Nash-Sutcliffe coefficient, and
RNASH is the Nash-Sutcliffe computed based on square-
Figure 4. Observed versus simulated depth to water
table. The observed data are obtained by kriging data in the
WELLOGIC database.
2558
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
given that the region is filled with inland lakes. The simulated agricultural ET tends to be higher than MODIS products. Considering the uncertainties and errors with the
MODIS product, we argue that this level of match is quite
encouraging. Figure 7b shows the temporal comparison of
basin-average ET between simulated and MOD15A2. The
simulated ET tends to be slightly higher than the MODIS
product for most of the months. In general, the temporal
dynamics agree very well.
3.3. General Background Hydrology of the Watershed
[31] The primary purpose of this section is to use the
model to elucidate the essential hydrologic dynamics in the
Figure 5. Observed versus simulated daily soil temperature at Romeo, Michigan.
observed data are measured at a point, which can be
expected to deviate from the average.
[29] Figure 6a shows the comparison of simulated LAI
with the level 4, 8-day Terra MODIS LAI product
(MOD15A2) [Knyazikhin et al., 1999; NASA Land Processes Distributed Active Archive Center (LP DAAC),
2012]. In Figure 6a, the urban areas (where MOD15A2
gives NaN values) are blanked out and shown in white for
easier visual comparison with the MODIS observations.
Figure 6b shows the time-series comparison of simulated
and observed (MODIS) LAI at two locations with relatively
uniform land use type of DBF and soybean, respectively.
We experience some problems when comparing with
MODIS LAI: (a) MODIS products always show no LAI at
all in winter while the model does predict LAI for evergreen forest (ENF), which ranges from 5.5 to 6.5 (but can
be covered by snow) and (b) peak values of LAI for corn
published in the literature can reach 5–6 [e.g., Howell et
al., 1997; Suyker and Verma, 2009], which matches with
the simulated peak corn LAI, whereas the MOD15A2 LAI
for the agricultural regions in this basin never goes above 3.
Both problems lead to simulated LAI being higher than
MODIS observed LAI. However, the literature values seem
to suggest this may be a problem of the MODIS rather than
the model. We notice that the LAI of the DBF and soybean
are very well simulated, especially the onset and peak periods
(the MODIS product does show some temporal variation of
peak LAI for the DBF, which is more likely due to cloud
interferences and observational errors). The simulated offset
of DBF is much more abrupt than the observed, which is due
to the linear offset function used in the tree phenology in
CLM, suggesting that a more smooth offset formulation
should be considered in future model enhancements.
[30] Figure 7a shows the simulated ET versus the
MODIS-based ET estimates (MOD16A2 and MOD16A3
[Mu et al., 2011, 2007]). The large-scale spatial pattern is
preserved, including relatively lower ET in the eastern agricultural region and higher ET in the central-western forested areas. However, some fine-scale patterns predicted by
the model were not observed in the MOD16A2 data. For
example, the model simulates a high ET pocket at the corner of the southwestern boundary, whereas the MODIS values there are very low. The values cannot be this low,
Figure 6. Comparison between MODIS product
(MOD15A2, 8 day) and the PAWSþCLM simulated LAI.
(a) Spatial map comparison for June 2005 and December
2005. (b) Time series comparison for two locations with
relatively uniform land cover type of deciduous broadleaf
forest (DBF) and soybean, respectively.
2559
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Figure 8. Monthly mean streamflow generating fluxes
and ET (mm) from 8 years of simulation. The black line
with diamond symbol is precipitation.
Figure 7. Comparison between MODIS products and
PAWSþCLM simulated evapotranspiration (ET). (a) Spatial map comparison of annual ET (mm) in 2005. (b) Timeseries comparison of basin-average monthly ET (mm).
watershed, which serves as the context for the analysis of
relative importance in section 3.4. As we will see, the spatiotemporal patterns observed in this watershed are governed by different processes with variable strengths of
controls. These patterns naturally drive us to ask the questions about the relative strengths of controls.
[32] Insight into the hydrologic cycle of this watershed
can be gained from Figures 8 and 9, which show, respectively, mean monthly basin-average fluxes, temporal trend
of the total soil moisture content, TSMC (including water
and ice), and snow water equivalent, SWE. Qgc, Ex, SatE,
and InfE are the four major mechanisms of streamflow generation, namely, groundwater contribution to streamflow,
lowland exfiltration, saturation excess, and infiltration
excess. Their sum can be seen as approximate, but not
equal, to the streamflow, because of interception and
reinfiltration.
[33] Figure 8 vividly describes a system that is energy
controlled most of the year and becomes moisture limited
only in the hottest summer months. Precipitation increases
from March until it reaches its peak in May. Streamflow
normally reaches its peak in March. The first few precipitation events can be very pronounced in terms of streamflow
generation. In April and May, although precipitation
increases more than 80%, ET jumps even more rapidly and
overcomes the added precipitation. Correspondingly, the
TSMC of the basin shifts into the ‘‘net-loss’’ phase (Figure
9), and the soil moisture pool starts to be tapped to compensate for the increasing ET demand. There is still ample
moisture in the subsurface so that streamflow does not drop
noticeably. From June to August, we observe a rapid
decline of soil moisture content in the basin. Soil seldom reaches saturation. ET shifts from energy-limited to
moisture-limited conditions. Streamflow consequently
drops to its annual minimum in August. From September to the end of the year, ET wanes quickly when
TSMC steps into the ‘‘net-gain’’ region. ET almost vanishes in winter and the soil continues to store water
until the next cycle starts.
[34] The energy-driven dynamics is also clear from the
streamflow generation. The overland flow contribution to
streamflow is commonly divided into saturation excess
(SatE) and infiltration excess (InfE) [Dingman, 2008],
which together create the peaks in the hydrograph. We
define SatE as runoff that occurs when the entire soil
Figure 9. Basin average state variables showing interannual variations of (a) soil water content (change from initial
state) and (b) snow water equivalent. The year 2003 is a
drought year.
2560
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Figure 10. Annual average spatial distributions of (a) latent heat flux (W/m2); (b) sensible heat flux
(W/m2); (c) top 10 cm soil temperature ( C); (d) top 10 cm soil moisture (e) infiltration excess (mm/yr)
(runoff generated before soil column is saturated) ; (f) saturation excess (mm/yr) (runoff generated at
time of saturation) ; (g) recharge to groundwater (mm/yr); and (h) NPP (gC/m2).
column underneath the surface area is saturated. Figure 8
demonstrates that streamflow generation is tightly related
to the seasonal energy cycle. Saturation excess evolves in
phase with TSMC in Figure 9 and is the dominant process
in winter and spring. The dominance of SatE again implies
that the system is energy limited in cold seasons. Infiltration excess, on the other hand, is less important than SatE
in cold seasons, which is explained by the high conductivity of glacial drift soils in the region. The relative importance of InfE rises in summer, when streamflow diminishes
and river stages are lowest. In August, storm events barely
stimulate much streamflow, as TSMC is very low and the
basin is far from saturation, and the small flashy peaks that
do occur are mostly composed of infiltration excess, as
shown in Figure 8b. InfE is primarily generated from the
urbanized regions of the southern basin and, when SatE
diminishes, its percentage in the runoff rises.
[35] Figure 10 illustrates the spatial patterns of long-term
average annual energy/water states and fluxes. The water
variables seem dominated by topography, whereas the
2561
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
energy variables are strongly influenced by land use/land
cover types. Figure 10 shows that SatE is commonly generated from the lowland plains on the east, only sporadically
on the local depressions on the high hills and around the
bending corner in the southwest, which as mentioned previously is a semienclosed subbasin that hosts a chain of small
lakes. The pattern of SatE is very similar to the top 10 cm
soil moisture map, which shows the lowland areas being
generally wetter than upland. This pattern indicates that
groundwater lateral flow may have transported a large
quantity of water from the western hills to the eastern low
plains, over the entire spatial scale of the watershed. This
hypothesis is supported by the groundwater recharge map,
which shows the hills as mainly recharging areas and the
low plains as discharging areas (Recharge < 0). On the
other hand, sensible heat, temperature, and net primary production (NPP) maps are primarily controlled by land use
types. Areas with high sensible heat correspond to the
urban centers. This pattern is also obvious from the soil
temperature. Ground surfaces of urban areas are, on annual
average, 2 to 4 hotter than forested areas due to the
‘‘urban heat island’’ effect [Landsberg, 1981]. These values
are similar to reported values (e.g., 4.3 C in Vukovich
[1983] and 2.1 C in Turkoglu [2010]). Unsurprisingly, the
regions with the lowest annual temperature are the same
areas with the highest latent heat flux. The agricultural
land, dominated by soybean, has quite high soil temperature, but its sensible heat flux is not higher than the urban
area. Compared to the forested areas, soybean provides less
shading and dissipates less latent heat due to the rather low
LAI and low canopy height of soybean (maximum LAI is
around 2). Compared to the urban areas, soybean has a
larger resistance to sensible heat. The highest values of
latent heat are found in the forested or perennially ponded
areas near streams in the high hills. These areas have ample
moisture supply due to local groundwater convergence.
Following the same rationale, we would expect higher
latent heat to be found on the low plains. However, such an
expected pattern is not visually identifiable from the latent
heat flux map. The latent heat figure shows generally higher
latent heat in the northern half of the basin with vegetated
land covers. The southern tip of the basin is urbanized and,
therefore, does not have enough vegetation to transpire
water. Apparently, the influence of groundwater flow on ET
is overshadowed by the land use and other spatial trends.
[36] Strong influences of landform and topography on
runoff, energy, and vegetation processes have also been
noted [Band et al., 1993; Ivanov et al., 2008; Jencso and
McGlynn, 2011; Kim et al., 1999; Rihani et al., 2010] and
modeled [Agnese et al., 2007; Bracken and Croke, 2007;
Laio et al., 2009; Troch et al., 2003]. To get a better understanding of the spatiotemporal distribution of runoff generation and groundwater influences on vegetation in this
watershed, we show in Figure 11 the fluxes and states along
a hillslope transect. The straight transect line begins on the
central hill, nearly follows the elevation gradient and ends in
the lowland plain, as highlighted by the line segment A-B in
Figure 2. To reduce noise and facilitate our interpretation,
the vegetation types along the profile are all set to DBF in
simulation S5 (Table 2), and the soil texture is set uniformly
to the most typical values. In Figure 11, time is on the x axis,
and distance along the transect (abbreviated as distance) is
on the y axis, such that viewing from top of a panel to the
bottom corresponds to going from uphill to downhill. We
show weekly averaged plots of surface runoff, top 10 cm
soil moisture, groundwater head (H), and NPP. Since the
runoff values vary across several orders of magnitude, we
show the data after applying a cube-root transformation. As
we can see, the generation of runoff is highly concentrated,
both in space and in time. Unsurprisingly, the pattern of runoff correlates closely with the soil moisture pattern. There is
a thin horizontal line (near distance ¼ 4 km) of runoff generation that occurs sporadically in the year. At this location,
transect A-B intersects a level 3 stream (the Stony Creek).
Apart from this line, runoff occurs primarily on the lowland,
which corroborates the findings from Figure 10. The divide
at a distance of 15 km, where the hillslope descends to the
flat plains, is clearly identifiable. As discussed by Kim et al.
[1999] and Salvucci and Entekhabi [1995], this divide distinguishes two distinct zones, the upper of which is dominated
by infiltration and moisture-limited evaporation while the
lower is characterized by discharge and runoff production.
The peaks of runoff generation occur only in two short time
windows, one in early May and another in December, when
there is little vegetation activity. Like in other years, several
spring and winter storm events generate most of the runoff
volume. Very low volumes of runoff are generated in
summer even in the lowland, despite a large storm event in
July, which corresponds to a relatively ‘‘quiet’’ summer period in the hydrograph. Compared to runoff, groundwater
head (H) varies much more smoothly, demonstrating spatiotemporal continuity. In summer, the contour lines of H bend
upward, corresponding to the drydown of groundwater levels. From Figure 10h, it is clear that the availability of water
plays a lesser role than the seasonal energy input. NPP, both
uphill and downhill, responds primarily to seasonality. The
plants achieve maximum productivity in June and July, following the maturity of the LAI and the high energy input.
However, during these 2 months, the influence of moisture is
also visible. The lowland trees have approximately 15%
higher NPP than uphill trees. Besides NPP magnitude,
downhill trees also have longer period of high productivity.
In June, there is a noticeable gap between the productivity
peaks. Reading from the precipitation and runoff plots, we
see that the timing of this gap corresponds to the hiatus
between two large storm events. This period of lowered productivity is clearly due to moisture limitation. The gap is
very brief and shallow for the lowland trees, owing to
groundwater subsidy.
[37] The rich patterns of different fluxes and states are
apparently dominated by different processes. These interesting observations force us to ask the questions: How do
these processes produce, either in isolation or in concert,
the observed spatiotemporal patterns? Since both groundwater and vegetation type seem to influence ET and NPP,
which has a stronger influence ? How relatively important
is each process on different hydrologic and vegetation
processes? What is the role of soil water retention and unsaturated conductivity? We use hypothetical simulations to
answer these questions in the next section.
3.4. Relative Strengths of Controls
[38] In order to answer the questions raised at the end of
section 3.3, we design a series of simulations to quantify
2562
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Figure 11. (a) Precipitation (mm/day). (b-e) are spatiotemporal plots of fluxes and states along the hillslope transect A-B. The x axis is time and the y axis is distance from point A along the transect. Reading
from top of the panel to the bottom is going from upland to lowland. (b) Runoff (mm/yr); (c) soil moisture content (m3/m3); (d) groundwater head (m); and (e) NPP (gC/m2/yr). The runoff values in Figure
11b are cube-root transformed because the original values span several orders of magnitude and two
peaks overwhelm all other events. Elevation plot is placed to the left of the panels to show the topographic descent of the hillslope.
the impacts of different factors on long-term annual ET and
NPP as well as other variables (infiltration, runoff, transpiration, soil temperature, and soil moisture). For simplicity,
we generically refer to the outcome variable as u. Details
of these simulations and the objectives of the numerical
experiments are summarized in Table 2 and will be discussed in the following sections. In this paper, we put our
focus on four factors: land use types, topography-induced
groundwater flow, soil water retention, and soil nitrogen
levels. Other factors such as slope aspect and micrometeorology may also be important for some sites [see Ivanov
et al., 2008], but they are outside the scope of this paper.
We first characterize the effect of each factor and search
for an appropriate surrogate index for the factor. Realizing
that each process is a complex combination of different
controls (e.g., for soil properties, soil conductivity, and
water retention characteristics all affect plant growth), we
attempt to simplify our analysis by considering both soil
characteristics and groundwater convergence as lumped
‘‘factors.’’ Then, we evaluate the relative strength of each
control in multivariate simulations using analysis of variance (ANOVA).
2563
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Table 2. List of Hypothetical Simulations
Simulation
S0
S1
S2
Procedures
Purpose
Use the baseline parameters
Based on S0, land uses are randomly distributed in the
watershed
Based on S0, land use uniformly set to DBF, lateral
groundwater disabled
To serve as the baseline scenario
To show the intraclass and interclass variability of ET
S3
Based on S0, land use uniformly set to DBF
S4
Based on S0, land use randomly selected from vegetated
classes: temperate evergreen tree, broadleaf
deciduous tree, C3 grass, and corn
Same simulation as S0, however, along the topographic
gradient (line A-B), land use is set to D BF, and soil is
set to the dominant soil type along the paths.
S5
3.4.1. Single Factor Analysis and Process Indices
[39] We refer to the baseline simulation as S0, which
most closely represents the realistic setting of the basin and
is used to generate all results in the previous sections. To
evaluate the impact of land use types, we carry out a hypothetical simulation, referred to as S1, with randomly distributed land uses, while all other factors are kept identical
to S0. In S1, five land use types (namely, temperate evergreen forest, temperate deciduous forest, grass, corn, and
medium-intensity urban), each assigned equal areas, are
randomly placed in the watershed to eliminate any spatial
correlation with soil types, topography, and convergence
levels. The resulting long-term annual average ET is summarized by land use types and presented in the boxplot in
Figure 12. The simulated ET are in general agreement with
literature reported values for DBF [e.g., Oishi et al., 2010],
corn, and soybean [e.g., Suyker and Verma, 2009]. The ET
for corn is somewhat in the lower range of the reported irrigated plants [e.g., Howell et al., 1997], because irrigation is
not simulated. As we can see, the interspecies variation of
ET can be significant. Mean forest ET is higher than that of
corn, which is higher than ET of grass. However, all vegetated lands significantly have greater ET values than imper-
Figure 12. Boxplot of land use influences on ET from
simulation S1, where land use types are randomly placed in
the watershed and the correlation with spatial locations are
removed. The land use types are temperate needleleaf evergreen forest (ENF), temperate deciduous broadleaf forest
(DBF), C3 grass (C3g), C3 general crop (C3c), corn, soybean (soybn), and impervious (imp.)
To show the influence of soil characteristics on plant
growth (in terms of water supply) and also to serve as a
basis for comparison with S3
Together with S2, to show the influence of lateral
groundwater flow
To assess the relative significance of the controlling
factors with effect of urbanization silenced.
To demonstrate the influence of spatiotemporal
distribution of runoff, soil moisture, groundwater head,
and NPP.
vious areas, supporting the spatial pattern observed in
Figure 10. The results suggest that urbanization has a significant impact on ET fluxes and explains why its spatial
pattern overwhelms that of topography and groundwater
convergence. To further study the impact of other factors,
we carry out the ensuing analysis in this section after
removing the masking effect of urbanization.
[40] In simulation S2, the land use types are uniformly
set to deciduous broadleaf forest (DBF). We disable
groundwater lateral flow by setting the lateral conductivity
of the aquifers to 1014 (m/day). As a result, the cells are
almost isolated from each other, as in the original CLM
model. Because other variables, including vegetation type
and lateral flow, are homogenized, we argue that the NPP
obtained from S2 also measures the overall ability of soil
(including the effect of soil water retention and thickness)
to retain and supply water for tree growth. This effect is
influenced not only by soil water retention characteristics
but also by conductive properties and thickness. Since both
soil retention and conductivities are heavily dependent on
texture, we may roughly refer to this physical factor as soil
texture. Therefore, the NPP values obtained from S2 can be
viewed as a ‘‘soil water supply index’’ for NPP, or
SWSI(NPP). More generally, we define SWSI(u,x) as the
long-term annual average values of variables u from S2 at
spatial location x in the basin. For simplicity, x is omitted
in the ensuing notations. u can be either ET, NPP, runoff,
infiltration, or any other variable we choose. In the present
simulation, SWSI(u) then represents how spatial variability
of soil texture impact u. SWSI(NPP) are shown as a spatial
map in Figure 13a.
[41] An index for the groundwater flow process can be
extracted from simulation S3, which has all settings identical to S2 except that lateral groundwater flow is enabled.
Therefore, the difference uS3 uS2 represents the net effect
of lateral groundwater flow on variable u. We therefore
define a ‘‘lateral flow index’’ for the variable u, or LatI(u),
as the difference of long-term (5 years, 2005–2009) annual
average u between S3 and S2. ðNPP S3 NPP S2 Þ=NPP s2 100% is shown in the map in Figure 13b. As we can see,
lateral groundwater flow subsidizes vegetation in the
lowland plains under drought conditions and increases their
annual productivity by 5%. However, this subsidy comes
at the cost of a 10–15%, even up to 30% in some cases,
reduction of NPP on the highland hill areas. Earlier studies
2564
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
fore, for the purpose of comparing their relative influences,
we choose SWSI and LatI to represent, respectively, soil
properties and groundwater flow dynamics.
[42] Although the representation of the N cycle in CLM
is still too crude compared to reality, it does allow us to
make some initial assessment of the order of magnitude of
the influence as compared to other factors. Over a long
term, a dynamic balance is established between nitrogen in
the plant, litter, SOM and mineral pools, BNF, deposition,
leaching, and denitrification. Random perturbation of the N
pools will lead to a rebalancing of the system and a gradual
regression to the equilibrium. In order to express the N variability while approximately maintaining such a balance,
we set the BNF rates using equation (1) [Cleveland et al.,
1999; Luo et al., 2006; Oleson et al., 2010]:
BNF ¼ 1:8 1 e0:003NPP ;
Figure 13. Spatial distributions of (a) NPP from S2, representing the soil’s ability to provide water to plants and
(b) NPP percentage change from S2 to S3 (the difference is
termed LatI), showing the net effect of lateral groundwater
flow in subsidizing lowland trees while reducing available
moisture for upland trees.
[Ivanov et al., 2008, 2010] that focused on arid, moisturelimited environments showed stronger influence of lateral
groundwater redistribution on vegetation growth. Our
results indicate that, even for an energy-limited region as
the present watershed, groundwater may still play a noticeable role in regulating plant growth, by subsidizing lowland
plants while sacrificing upland productivity. The groundwater influence, in our case, also seems to impact plants on
the highland much more significantly than those on the
lowland plains. Interestingly, the subsidy seems higher immediately adjacent to the foot of the hills and reduces as
we go deeper into the plains. This is because, on the flat
plains, the trees at the foot of the hill have the best access
to the lateral groundwater discharge that comes from uphill.
On a side note, one would expect LatI to be correlated to
topographic index (TopoI) [Beven and Kirkby, 1979], a
‘‘wetness’’ index that aims to provide a simple metric to
estimate the likelihood of saturation. However, as shown in
the (LatI-TopoI) plot in Figure 14b, such a correlation is
not obvious (plotting TopoI against ET yields a similar
plot). Roughly, a rectangle could be drawn to enclose the
scattered dots. However, TopoI is unable to explain a large
portion of the variability, which might include the heterogeneity of aquifer properties, foothill dynamics, and the
particular topographic and watershed construct (concave or
diverging, etc.), among other factors. It is known that the
topographic index may not work well in flat terrains. There-
(1)
where BNF is in gN/m2/yr, NPP is the annual net primary
production, and a is an adjustment factor whose values are
kept at these hypothetical levels : 2 0:2; 0:6; 0:8;
1; 1:2; 1:4; 1:8. We choose this range as it resembles the
range of the uncertainty with natural BNF rate estimates
[Galloway et al., 2004]. In the original CLM, a is always
equal to 1.0. In our relative control analysis, the model is
spun up over 4000 years by looping through the climatic
forcing data between 2001 and 2009, as in Thornton and
Rosenbloom [2005], but with different a values as given
above. The final C and N states at steady state differ greatly
between different assumed BNF rates. Table 3 shows some
typical maximum LAI values, end-of-year total SOM carbon/nitrogen states as a result of the different BNF rates.
While the alteration of the BNF rates changes not only N
pools but also C pools, we refer to this factor as nitrogen
levels, to reflect the fact that nitrogen is the limiting factor.
We note that although we perturbed the BNF rates function,
we do not posit any causal or ordinal relationship between
BNF and NPP change. This is just a way for us to express
the variability of nitrogen availability and propagate it
through the ecosystem, while maintaining the balance
between different N pools. Of course, we may also choose
Figure 14. LatI(NPP) (which is NPP of S3 minus NPP of
S2) as a function of topographic index. As we can see, the
topographic index is not a very good predictor for NPP in
this watershed.
2565
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Table 3. Different Ecosystem States (Maximum Value in a 9 Year Period) for DBF at Steady State as a Result of Different BNF Ratesa
a
0.2
0.6
0.8
1.0
1.2
1.4
1.8
LAI
Canopy top height (m)
Total soil organic matter carbon (gC/m2)
Total soil organic matter Nitrogen (gN/m2)
Total ecosystem carbon (gC/m2)
Total ecosystem nitrogen (gN/m2)
4.0
18.4
5,930.2
591.3
13,770.9
623.5
4.8
20.5
7440.3
741.9
18,275.9
783.0
5.1
21.5
8,195.2
817.1
20,562.3
862.7
5.5
22.3
8,913.5
888.8
22,731.2
938.6
5.8
23.0
9,535.1
950.8
24,645.0
1,004.3
6.2
23.6
10,154.2
1,012.5
26,519.8
1,069.7
6.8
24.7
11,234.2
1,120.2
29,805.8
1,183.9
a
BNF ¼1:8að1 e0:003NPP an Þ where NPPan is the annual NPP and a is the adjustment factor.
nitrogen deposition rates as the knob to perturb, which
would generate similar effects. The C/N pools at the equilibrium corresponding to different a values are recorded to
be used as initial states in the following combined
simulations.
3.4.2. Combined Analysis Using ANOVA
[43] Using ANOVA, we can quantitatively assess the relative strengths of the controlling processes by approximately estimating the variance of vegetation growth
attributable to each one of the factors. In simulation S4,
baseline soil water retention properties and aquifer conductivities are used. To mask the dominant role of urbanization, the land cover in S4 are limited to evergreen forest
(ENF), DBF, C3 grass, and soybean, and these classes are
randomly placed in the watershed. The nitrogen level of
each cell is randomly coded between 1 and 7. For each
coded level, the initial C/N states are looked up from the
long term spin-up results with the corresponding BNF rate
described in the last paragraph.
[44] A four-way ANOVA test (assessed at the 1% confidence level) with interactions is used to partition the total
variance of u in S4 into the individual contributions. The
four predictors are land use types (PFT), soil (SWSI(u)),
lateral GW flow (LatI(u)), and nitrogen levels (N level).
The ANOVA results are presented in Tables 4 and 5 and
graphically in Figure 15 for different u. We examine seven
different u’s, which are the long-term annual averages of
ET, NPP, runoff (Rf), infiltration (Inf), top 10 cm soil temperature (t10cm), and top 10 cm soil moisture (10cm). The
cross terms are generally small compared to the main factors. For simpler presentation, they are combined as an
‘‘interactions’’ term in Figure 15. The ‘‘error’’ term, which
represents the variance that cannot be explained by the
main factors and their interactions, is also shown. The partitioned variances sum up to 1 for all u’s.
[45] It is obvious from the table that all factors are statistically significant for all variables, but their relative importance shifts for different u, revealing the controlling
processes. Groundwater lateral flow dynamics clearly dominate hydrologic fluxes like runoff and infiltration (but not
ET), with LatI weighing more than 70% and 76%, respectively. The soil textural properties (SWSI) rank next to
LatI, explaining 16% and 11% of the variance, respectively. LatI’s dominance over SWSI for Rf and Inf can be
related to the fact that saturation excess is the primary
streamflow generation mechanism in the basin, and the
state of saturation in space and time is strongly controlled
by groundwater dynamics. NPP is primarily controlled by
nitrogen levels, PFT, and their interactions. The cross term
(N level PFT) occupies more than 97% of the interactions. The results highlight N as a limiting factor of vegetation growth in the system. The top 10 cm soil temperature
is almost completely driven by land use types, even after
urban land use is excluded from S4. The surface soil temperature t10cm reflects the optical and thermal properties of
the canopy structure. Although CLM4.0 does consider the
difference in thermal properties between water and soil
(PAWSþCLM does not track heat transfer by water), neither LatI nor SWSI seem to be have any noticeable impact.
[46] The most interesting variable is ET, for which all of
LatI, SWSI, PFT, and N levels and their interactions are
important. This is attributed to the colimitation by soil
water, energy and vegetation, which is unique to this humid
continental climate where no single factor dominates. The
ecosystem controls (N level : 23%, PFT: 38%, their interactions: 8%) seem to be stronger than water controls (SWSI :
Table 4. Four-Way ANOVA of N Level/Land Use Type/Soil
Texture/Groundwater Flow on u ¼ ET
Table 5. Four-Way ANOVA of N Level/Land Use Type/Soil
Texture/Groundwater Flow on u ¼ NPP
Source
Sum of
Squares
Percent of
Variance
Explained (%)
N level
PFT
SWSI
LatI
N level PFT
N level SWSI
N level LatI
PFT SWSI
PFT LatI
SWSI LatI
Error
Total
4,835,447
7,831,229
1,988,126
3,242,024
1,100,465
131,682
100,415
95,569
158,611
10,647
1,056,450
20,550,664
23.53
38.11
9.67
15.78
5.35
0.64
0.49
0.47
0.77
0.05
5.14
100.00
F
Prob > F
Source
Sum of
Squares
1,783
5,775
4,398
7,172
135
49
37
70
117
24
0
0
0
0
0
0
0
0
0
1.30E-06
N level
PFT
SWSI
LatI
N level PFT
N level SWSI
N level LatI
PFT SWSI
PFT LatI
SWSI LatI
Error
Total
84,307,345
2,594,772
742,608
253,061
24,127,980
2,326,108
632,842
62,980
10,551
5
1,281,601
116,339,853
2566
Percent of
Variance
Explained (%)
72.47
2.23
0.64
0.22
20.74
2.00
0.54
0.05
0.01
0.00
1.10
100.00
F
Prob > F
25,590
1,575
1,352
461
2,441
706
192
38
6
0
0
0
0
0
0
0
0
0
2.56E-04
9.27E-01
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Figure 15. ANOVA analysis results show the relative strengths controls of physical processes on
hydrologic and vegetation dynamics (Rf, runoff; Inf, infiltration; ET, evapotranspiration; Tp, transpiration; NPP, net primary productivity; t10cm, top 10 cm soil temperature ; 10cm, top 10 cm soil moisture;
Interactions, for simplicity, the cross terms between different factors are combined into one interaction
term). ‘‘Original’’ stands for ANOVA test carried out without global parameter changes described in section 3.4.3. In ‘‘K 0.1’’ ANOVA test, all simulations (S2, S3, and S4) have their K scaled by 0.1. Similarly, 0.1 is added to N in all simulations in ‘‘N þ 0.1’’ ANOVA test. ‘‘Double resolution’’ means the
spatial step size is halved (number of cells quadrapled). (a) Original, (b) K 0.1, (c) double resolution,
(d) N þ 0.1.
10% and LatI: 15%). Soil nitrogen level (as a result of ecosystem spin-up to equilibrium using different BNF rates)
plays an important role in ET, confirming previous observations of positive correlation between BNF and ET. As is
clear from Figure 15, this influence is mainly exerted
through its control on transpiration (Tp). It can then be
inferred that groundwater and soil texture are much more
dominant controls on soil evaporation. The influence of N
levels also propagates to hydrologic fluxes and temperature, although not in a strong fashion. The findings suggest
that in order to accurately predict spatiotemporal distributions of ET (or latent heat flux), the ecosystem dynamics
including nitrogen availability must be carefully considered.
[47] In contrast, LatI and SWSI have comparable significance for 10cm and each explain about 40% of the variance. As a central variable for hydrologic and
biogeochemical processes, soil moisture has been the most
intensively studied in the literature among the variables analyzed in this paper. Interestingly, our results have been
corroborated by analysis of field data [Famiglietti et al.,
1998; Joshi and Mohanty, 2010; Pan and Wang, 2009;
Wilson et al., 2004], where soil and topography are found
to be of similar importance (although soil/vegetation
impacts are lumped in Wilson et al. [2004], and it is not
clear how much variance they each explain). Therefore, for
hydrologic and biogeochemical processes, which depend
on soil moisture in a strongly nonlinear fashion, it is important to consider both local soil textural properties and
regional groundwater flow. Qiu et al. [2001] report a larger
control of mean soil moisture by land use. The actual
controls of vegetation on soil moisture are expected to be
higher than explained in the model. Further study should
consider more diversified vegetation and the effects of
hydraulic lift by vegetation or macropores created by vegetation roots. Our study intentionally breaks the correlation
that naturally exists between topography and vegetation
types [e.g., Brown, 1994]. We also note a large error for
10cm, indicating that the complex dynamics of moisture
cannot be explained by simple linear models of the variables. The complex controls of soil moisture have been
reported by Zhu and Lin [2011], who conclude that terrain
attributes are more important in wetter places among other
dynamics.
[48] It must be noted that, for NPP, although LatI and
SWSI appear unimportant compared to vegetation types
and soil nitrogen content, both have significant influence
once we hold PFT and N levels constant across the basin,
as we have shown in S3.
3.4.3. Robustness of the Analysis
[49] A reasonable question to ask is whether the relative
importance of different processes will stay the same when
we perform the same exercise with a different model, a different spatial resolution or in another region where climatic
input and physical parameters (e.g., aquifer transmissivities) are different. Intuitively, the results should change
when physical parameters are altered since, if we imagine a
basin with very low aquifer transmissivities, the influence
of groundwater flow dynamics should be very low.
Although the effect of different model formulations should
also be examined in the future, in this paper, we assess the
effect of the parameters and spatial resolution. To quantitatively assess this impact, we redo the ANOVA analysis in
section 3.4.2 with exactly one of the following changes
applied: (a) the conductivity of the unconfined aquifer
2567
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
everywhere in the basin is reduced by a factor of 10. Therefore, the basin average unconfined aquifer K becomes
1.3 m/day; (b) the van Genuchten parameter n is increased
by 0.1, thereby increasing the nonlinearity of the soil water
retention and unsaturated conductivities ; (c) the horizontal
cell size is reduced in half (resolution is doubled), thereby
quadrupling the number of cells in the basin.
[50] The results are shown in Figure 15. Indeed, the variances attributed to each factor do shift after making the
changes (Figure 15a). The influence of LatI reduces for all
variables, implying a less dominant groundwater system.
The relative roles of other factors, especially SWSI,
increased accordingly to make up for the drop in LatI. This
means the relative importance of groundwater dynamics
will shift when we move to a geologic configuration that
extends beyond the current range of variability. With
change (b), however, the changes are relatively minor, with
SWSI explaining slightly more variance in Rf and Inf. The
shifts in the relative importance due to a change in the
range of variability have also been previously noted by
field data analysis [Zhu and Lin, 2011]. With the help of a
numerical model, however, the relative importance of controls at each site can be analyzed prior to extensive data
collection.
[51] The spatial resolution has a relatively weak influence on the results of the analysis. For ET, Rf, and Inf, LatI
now plays a more important role when resolution is
doubled, scoring 20%, 74%, and 87%, respectively. This is
due to the better resolution of the effects of topography relative to the original simulation. For t10cm and 10cm, however, SWSI increased slightly, presumably due to a larger
range of SWSI as higher resolution allows more distinct
soil classes to be simulated. However, the general trend
stays the same as in the lower resolution simulations.
4.
Discussion and Conclusions
[52] The knowledge and the mechanistic understanding
obtained in this study may help guide future modeling and
data-gathering decisions. For example, many previous
efforts focused on downscaling satellite soil moisture
observations [e.g., Busch et al., 2012; Mascaro et al.,
2010; Pellenq et al., 2003], and our results indicate that
both topography and soil texture should be environmental
predictors in such downscaling methods as they are important controls for soil moisture. For another example, for
gross primary production (GPP/NPP) estimation, at least
for the humid continental climatic region examined here,
plant type and nutrient availability far outweigh groundwater dynamics and soil water supply properties in terms of
their relative importance. It is therefore important to obtain
soils data sets with better carbon/nitrogen storage estimations. For quantification of terrestrial-atmospheric vapor/
heat exchange, groundwater dynamics, soil properties, and
ecosystem processes are all important. Because both shortranged and long-ranged patterns of groundwater flow exist,
it is important to consider regional groundwater dynamics
and subsurface processes in land surface models. This
means that for large-scale simulations that aim at studying
terrestrial-atmospheric vapor/energy exchange, many simplified cell-based hydrologic formulations (such as the
TOPMODEL-based formulations in original CLM and
Noah LSM) are likely to produce unreliable estimates.
Although similar conclusions have been drawn in the literature [Anyah et al., 2008; Fan et al., 2007; Maxwell and
Kollet, 2008a], our study is unique in this climatic region.
[53] In spite of the limitations of the current nitrogen
cycling schemes in CLM4.0, the results clearly indicate that
the nitrogen availability heavily influences not only vegetation growth but also ET, through which other hydrologic
fluxes are affected. Conventional hydrologic models often
do not consider the nutrient dynamics mechanistically. For
assessment of climate change on regional hydrology and
vegetation, nitrogen dynamics should be considered. The
results in this paper suggest that we need to take an integrative view of the ecosystem-hydrology interactions in order
to make better predictions. It is important for hydrologists
to continue expanding the domain of sciences in hydrologic
simulations to identify previously undiscovered but potentially important linkages, echoing the call for more crossinterdisciplinary integration [Wagener et al., 2010] and
encouraging grass-roots community collaborations.
[54] For many variables studied in Figure 15 (e.g., ET,
Tp, NPP, and t10cm), the error terms account for less than
10% of the total variance. Consequently, it is possible to
predict these variables (at least their annual averages) with
a fair degree of accuracy by just using simple linear models
of the main processes (land use, nitrogen, groundwater
flow, and soil water retention) and their interactions. However, the variables with higher percentage of error (e.g.,
10cm) will need to be addressed by more complex models
detailing the space-time interactions of factors.
[55] As with other similar studies, the conclusions in the
paper are somewhat tied to the model employed. The ability of a numerical model to represent linkages and interactions depends on its formulations. Considering the large
uncertainties associated with descriptions of the nitrogen
cycle in CLM, the numerical value of its relative importance will likely shift when improved N cycling schemes
are employed. Given that a large range of BNF rates
(altered by a) are used to get carbon nitrogen states, the
true influence of nitrogen is likely smaller than estimated.
Future simulations supported by novel observations will be
crucial for confirming the results. PAWSþCLM considers
water to move only vertically in the unsaturated zone. As a
result, the model does not produce accurate results when
significant interflow exists. The model currently does not
track nutrient transport within groundwater and surface
water. Also, the analysis in the paper does not include the
effects of microtopography, micrometeorological conditions, and aspect of the slope. In the future, it would be
interesting to compare results from other models using the
approach described here. The effect of very deep soils on
the coupling algorithm in PAWS seems to be mild given
the good match with observed groundwater heads and wellcaptured baseflow in the hydrograph. In the future, very
high resolution simulations will likely provide more insight
on the effect of spatial resolution. Scale-aware or multiscale simulation approaches are called for to help with the
issue of spatial scaling.
[56] To summarize, we have described the application of
a novel process-based hydrologic model PAWSþCLM to a
watershed in the humid continental climate region of the
U.S. Midwest to understand the relative importance of
2568
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
vegetation type, nitrogen availability, groundwater dynamics, and soil texture. The model shows good comparison
with various observation data sets. The background hydrology of the basin is revealed as one limited by energy and
water at different times during the annual cycle, with saturation excess being the dominant runoff generating mechanism. Groundwater plays an important role in the
watershed, controlling saturation excess and lowland exfiltration. Through a series of hypothetical simulations based
on the model, we use ANOVA to quantify the influence
and relative importance of the controlling processes. The
results confirm the basin, which has an average unconfined
aquifer conductivity of 13 m/day, as a groundwater-dominated system. Groundwater flow has been found to be the
major control on runoff and infiltration, explaining more
than 70% of the variance. Soil texture ranks next. Vegetation type and nitrogen levels are found to determine NPP,
top 10 cm soil temperature, and transpiration. Interestingly,
all factors are found to significantly control ET, which
serves as the nexus for ecosystem-hydrology interactions.
Nitrogen is shown to be a major control of ecosystem variables, but it also significantly controls transpiration, through
which it influences other hydrologic fluxes. While keeping
in mind the limitations with the CN cycling in CLM, the
results indicate that hydrologists need to continue expanding the domains of sciences in hydrologic simulations to
discover important linkages. The analysis method used in
the paper is shown to be relatively robust with respect to
changes in parameter values and spatial resolution, but
when we move to a region whose parameters extend
beyond the current range of variability, the results are
expected to shift.
Appendix A: Lowland-Storage Module
[57] Here we describe a new enhancement to the model.
Wetlands are an important land cover type that exerts
strong influence on hydrology, ecosystem functioning, and
water quality. A flexible lowland-storage module is added
to the original PAWS model to keep track of the exchanges
between wetlands, overland flow, and groundwater. The
wetland is conceptualized as a lowland-storage compartment of the overland flow domain, as shown in Figure 1.
To accommodate the wetland compartment, the mass balance equation for the flow domain is modified as follows:
H zw þ hf =fw
@hf
¼ fw Kw
Ew þ Fg rðhof uÞ;
@t
zw
(A1)
in which hf is the flow depth in the flow domain, H is the
groundwater head, Ew is the evaporation from lowland storage, fw is the areal fraction of lowland storage, Fg is runoff
from the ponding domain, u is overland flow velocity vector,
Kw is the bottom conductivity, zw is the bottom elevation of
the lowland storage, zw is the maximum of zwH or thickness of bed material layer, and hof is the overland flow depth
in excess of the lowland-storage capacity. Unit of the equation is m/day. We solve equation (A1) in two fractional steps.
The first one is an implicit step that calculates the groundwater exchange and obtains intermediate states hf . The second step takes hof ¼ max 0; hf dw , where dw is the
bucket depth of the depression storage, and sends result to
the 2-D diffusive wave overland flow solver as described in
SP10 to solve for the convective term, rhof u. The exchange
between groundwater and lowland storage becomes either
exfiltration (Ex) or infiltration (Inf), depending on the direction. Equation (A1) can also be applied to describe transient
shallow concentrated flows, paddy fields, or potholes. In
addition, the lowland storages in different locations are connected via overland flow paths. When flooding occurs, river
water may flood the overland flow domain, and in turn, fill
the lowland-storage compartments. The threshold for backfill
to happen is termed as hback, fw, and dw, all of which can be
inferred from analyzing land cover maps and fine-resolution
digital elevation model (DEM).
[58] Acknowledgments. J.N. and M.S.P. are supported by the NOAA
Center of Excellence for Great Lakes and Human Health. We thank Sam
Levis, NCAR, for valuable advice related to CLM and Ben Ong, Institute
for Cyber-Enabled Research (iCER) at MSU for help with high-performance computing. We thank three anonymous reviewers whose constructive
and detailed comments have greatly helped in improving the manuscript.
References
Agnese, C., G. Baiamonte, and C. Corrao (2007), Overland flow generation
on hillslopes of complex topography: Analytical solutions, Hydrol. Process., 21(10), 1308–1317, doi:10.1002/hyp.6354.
Anyah, R. O., C. P. Weaver, G. Miguez-Macho, Y. Fan, and A. Robock
(2008), Incorporating water table dynamics in climate modeling: 3.
Simulated groundwater influence on coupled land-atmosphere variability, J. Geophys. Res., 113, D07103, doi:10.1029/2007JD009087.
Arora, V. (2002), Modeling vegetation as a dynamic component in soilvegetation-atmosphere transfer schemes and hydrological models, Rev.
Geophys., 40(2), 3-1–3-24, doi:10.1029/2001RG000103.
Band, L. E., P. Patterson, R. Nemani, and S. W. Running (1993), Forest
ecosystem processes at the watershed scale: Incorporating hillslope hydrology, Agric. For. Meteorol., 63(1-2), 93–126, doi:10.1016/0168–
1923(93)90024-c.
Band, L. E., C. L. Tague, P. Groffman, and K. Belt (2001), Forest ecosystem
processes at the watershed scale: Hydrological and ecological controls of
nitrogen export, Hydrol. Process., 15, 2013–2028, doi:10.1002/hyp.253.
Berg, B., and R. Laskowski (2005), Litter Decomposition: A Guide to Carbon and Nutrient Turnover, Adv. Ecol. Res. 38, p. 448, Academic Press,
Waltham, MA.
Beven, K. (2002), Towards an alternative blueprint for a physically based
digitally simulated hydrologic response modelling system, Hydrol. Process., 16(2), 189–206.
Beven, K., and M. Kirkby (1979), A physically based, variable contributing
area model of basin hydrology, Hydrol. Sci. Bull., 24(11), 43–69.
Bonan, G. B., and S. Levis (2010), Quantifying carbon-nitrogen feedbacks
in the Community Land Model (CLM4), Geophys. Res. Lett., 37,
L07401, doi:10.1029/2010GL042430.
Bracken, L. J., and J. Croke (2007), The concept of hydrological connectivity and its contribution to understanding runoff-dominated geomorphic
systems, Hydrol. Process., 21(13), 1749–1763, doi:10.1002/hyp.6313.
Brown, D. G. (1994), Predicting vegetation types at treeline using topography and biophysical disturbance variables, J. Veg. Sci., 5(5), 641–656.
Burke, I. C., W. K. Lauenroth, and W. J. Parton (1997), Regional and temporal variation in net primary production and nitrogen mineralization in
grasslands, Ecology, 78(5), 1330–1340, doi:10.1890/0012–9658.
Burt, T. P., and G. Pinay (2005), Linking hydrology and biogeochemistry in
complex landscapes, Prog. Phys. Geogr., 29(3), 297–316, doi:10.1191/
0309133305pp450ra.
Busch, F. a., J. D. Niemann, and M. Coleman (2012), Evaluation of an empirical orthogonal function-based method to downscale soil moisture patterns based on topographical attributes, Hydrol. Process., 26, 2696–
2709, doi:10.1002/hyp.8363.
Celia, M. A., E. T. Bouloutas, and R. L. Zarba (1990), A general nass-conservative numerical-solution for the unsaturated flow equation, Water
Resour. Res., 26(7), 1483–1496.
Chorover, J., et al. (2011), How water, carbon, and energy drive critical
zone evolution: The Jemez–Santa Catalina Critical Zone Observatory,
Vadose Zone J., 10, 884, doi:10.2136/vzj2010.0132.
2569
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Cleveland, C. C., et al. (1999), Global patterns of terrestrial biological nitrogen (N-2) fixation in natural ecosystems, Global Biogeochem. Cycles,
13(2), 623–645, doi:10.1029/1999GB900014.
Collatz, G. J., M. Ribas-Carbo, and J. A. Berry (1992), Coupled photosynthesis-stomatal conductance model for leaves of C4 plants, Aust. J. Plant
Physiol., 19(5), 519–538.
Collins, W. D., et al. (2006), The Community Climate System Model version 3 (CCSM3), J. Clim., 19(11), 2122–2143.
D’Odorico, P., F. Laio, A. Porporato, and I. Rodriguez Iturbe (2003),
Hydrologic controls on soil carbon and nitrogen cycles. II. A case study,
Adv. Water Resour., 26, 59–70.
Dai, Y. J., R. E. Dickinson, and Y. P. Wang (2004), A two-big-leaf model
for canopy temperature, photosynthesis, and stomatal conductance,
J. Clim., 17(12), 2281–2299.
Dentener, F., et al. (2006), Nitrogen and sulfur deposition on regional and
global scales: A multimodel evaluation, Global Biogeochem. Cycles, 20,
GB4003, doi:10.1029/2005GB002672.
Dickinson, R. E. (1983), Land surface processes and climate surface albedos and energy-balance, Adv. Geophys., 25, 305–353.
Dickinson, R. E., K. W. Oleson, G. Bonan, F. Hoffman, P. Thornton, M.
Vertenstein, Z. L. Yang, and X. B. Zeng (2006), The Community Land
Model and its climate statistics as a component of the Community Climate System Model, J. Clim., 19(11), 2302–2324.
Dingman, S. L. (2008), Physical Hydrology, 2nd ed., Waveland Press,
Long Grove, IL.
Enviro-weather (2010), Enviro-weather Automated Weather Station Network (formerly known as MAWN). [Available at http://www.agweather.geo.msu.edu/mawn/.].
Famiglietti, J. S., J. W. Rudnicki, and M. Rodell (1998), Variability in surface moisture content along a hillslope transect: Rattlesnake Hill, Texas,
J. Hydrol., 210(1-4), 259–281, doi:10.1016/s0022-1694(98)00187-5.
Fan, Y., G. Miguez-Macho, C. P. Weaver, R. Walko, and A. Robock
(2007), Incorporating water table dynamics in climate modeling: 1.
Water table observations and equilibrium water table simulations,
J. Geophys. Res., 112, D10125, doi:10.1029/2006JD008111.
Fang, C., P. Smith, J. U. Smith, and J. B. Moncrieff (2005), Incorporating
microorganisms as decomposers into models to simulate soil organic
matter decomposition, Geoderma, 129, 139–146, doi:10.1016/
j.geoderma.2004.12.038.
Farquhar, G. D., S. V. Caemmerer, and J. A. Berry (1980), A biochemicalmodel of photosynthetic CO2 assimilation in leaves of C-3 species,
Planta, 149(1), 78–90.
Fatichi, S., V. Y. Ivanov, and E. Caporali (2012), A mechanistic ecohydrological model to investigate complex interactions in cold and warm
water-controlled environments: 1. Theoretical framework and plot-scale
analysis, J. Adv. Model. Earth Syst., 4, M05002, doi:10.1029/
2011MS000086.
Ferguson, I. M., and R. M. Maxwell (2010), Role of groundwater in watershed response and land surface feedbacks under climate change, Water
Resour. Res., 46, W00F02, doi:10.1029/2009WR008616.
Galloway, J. N., et al. (2004), Nitrogen cycles: Past, present, and future,
Biogeochemistry, 70(2), 153–226, doi:10.1007/s10533-004-0370-0.
Galloway, J. N., A. R. Townsend, J. W. Erisman, M. Bekunda, Z. C. Cai, J.
R. Freney, L. A. Martinelli, S. P. Seitzinger, and M. A. Sutton (2008),
Transformation of the nitrogen cycle: Recent trends, questions, and potential solutions, Science, 320(5878), 889–892, doi:10.1126/science.1136674.
Global Soil Data Task (2000), Global soil data products CD-ROM (IGBPDIS). International Geosphere-Biosphere Programme-Data and Information System. [Available at http://daac.ornl.gov/SOILS/guides/igbp.html,
accessed 22 Nov. 2012.].
Goderniaux, P., S. Brouyere, H. J. Fowler, S. Blenkinsop, R. Therrien, P.
Orban, and A. Dassargues (2009), Large scale surface-subsurface hydrological model to assess climate change impacts on groundwater reserves,
J. Hydrol., 373(1-2), 122–138, doi:10.1016/j.jhydrol.2009.04.017.
Gunduz, O., and M. M. Aral (2005), River networks and groundwater flow:
A simultaneous solution of a coupled system, J. Hydrol., 301(1-4), 216–
234, doi:10.1016/j.jhydrol.2004.06.034.
Groundwater Inventory and Mapping Project (GWIM) (2006), State of
Michigan Public Act 148 Groundwater Inventory and Mapping Project
(GWIM), IWR, Michigan State University, Technical Report. Available
at: http://gwmap.rsgis.msu.edu/ accessed on 22 November, 2012.
Hessl, A. E., C. Milesi, M. A. White, D. L. Peterson, and R. E. Keane
(2004), Ecophysiological parameters for Pacific Northwest trees. Rep.
PNW-GTR-618., U.S. Dep. of Agric., For. Serv., Pacific Northwest Res.
Stn., 14 pp., Portland, Oreg.
Howell, T. A., J. L. Steiner, A. D. Schneider, S. R. Evett, and J. A. Tolk
(1997), Seasonal and maximum daily evapotranspiration of irrigated
winter wheat, sorghum, and corn—Southern High Plains, Trans. ASAE,
40(3), 623–634.
Hwang, T., L. E. Band, J. M. Vose, and C. Tague (2012), Ecosystem processes at the watershed scale: Hydrologic vegetation gradient as an indicator for lateral hydrologic connectivity of headwater catchments, Water
Resour. Res., 48, W06514, doi:10.1029/2011WR011301.
Ivanov, V. Y. (2006), Effects of dynamic vegetation and topography on
hydrological processes in semi-arid areas, Ph.D. dissertation, 453 pp.,
Mass. Inst. of Technol., Boston, Mass.
Ivanov, V. Y., R. L. Bras, and E. R. Vivoni (2008), Vegetation-hydrology dynamics in complex terrain of semiarid areas: 2. Energy-water
controls of vegetation spatiotemporal dynamics and topographic
niches of favorability, Water Resour. Res., 44, W03430, doi:10.
1029/2006WR005595.
Ivanov, V. Y., S. Fatichi, G. D. Jenerette, J. F. Espeleta, P. A. Troch, and T.
E. Huxman (2010), Hysteresis of soil moisture spatial heterogeneity and
the ‘‘homogenizing’’ effect of vegetation, Water Resour. Res., 46,
W09521, doi:10.1029/2009WR008611.
Jencso, K. G., and B. L. McGlynn (2011), Hierarchical controls on runoff
generation: Topographically driven hydrologic connectivity, geology,
and vegetation, Water Resour. Res., 47, W11527, doi:10.1029/
2011WR010666.
Jia, Y. W. (1998), Modeling infiltration into a multi-layered soil during an
unsteady rain, J. Hydrosci. Hydraul. Eng., 16(2), 1–10.
Joshi, C., and B. P. Mohanty (2010), Physical controls of near-surface soil
moisture across varying spatial scales in an agricultural landscape during
SMEX02, Water Resour. Res., 46, W12503, doi:10.1029/2010W
R009152.
Kim, C. P., G. D. Salvucci, and D. Entekhabi (1999), Groundwater-surface
water interaction and the climatic spatial patterns of hillslope hydrological response, Hydrol. Earth Syst. Sci., 3(3), 375–384.
Knyazikhin, Y., et al. (1999), MODIS leaf area index (LAI) and fraction of
photosynthetically active radiation absorbed by vegetation (FPAR) product (MOD15) algorithm theoretical basis document [Available at http://
eospso.gsfc.nasa.gov/atbd/modistables.html.].
Kollet, S. J., and R. M. Maxwell (2008), Capturing the influence of groundwater dynamics on land surface processes using an integrated, distributed
watershed model, Water Resour. Res., 44, W02402, doi:10.1029/
2007WR006004.
Kondo, J., and S. Ishida (1997), Sensible heat flux from the earth’s surface
under natural convective conditions, J. Atmos. Sci., 54(4), 498–509.
Kundu, P. K., and I. M. Cohen (2010), Fluid Mechanics, 4th ed., Academic
Press, Waltham, MA.
Lai, C. T., and G. Katul (2000), The dynamic role of root-water uptake in coupling potential to actual transpiration, Adv. Water Resour., 23(4), 427–439.
Laio, F., S. Tamea, L. Ridolfi, P. D’Odorico, and I. Rodriguez-Iturbe
(2009), Ecohydrology of groundwater-dependent ecosystems: 1. Stochastic water table dynamics, Water Resour. Res., 45, W05419,
doi:10.1029/2008WR007292.
Landsberg, H. E. (1981), The Urban Climate, Academic Press, Waltham,
MA.
Lawrence, D. M., et al. (2011), Parameterization improvements and functional and structural advances in version 4 of the Community Land
Model, J. Adv. Model. Earth Syst., 3, M03001, doi:10.1029/
2011MS000045, 27 pp.
Lohse, K. a., P. D. Brooks, J. C. McIntosh, T. Meixner, and T. E. Huxman
(2009), Interactions between biogeochemistry and hydrologic systems,
Annu. Rev. Environ. Resour., 34, 65–96, doi:10.1146/
annurev.environ.33.031207.111141.
Luo, Y. Q., D. F. Hui, and D. Q. Zhang (2006), Elevated CO2 stimulates net
accumulations of carbon and nitrogen in land ecosystems: A meta-analysis, Ecology, 87(1), 53–63, doi:10.1890/04–1724.
Mackay, D. S., and L. E. Band (1997), Forest ecosystem processes at the
watershed scale: dynamic coupling of distributed hydrology and canopy
growth, Hydrol. Process., 11, 1197–1217.
Maggi, F., and A. Porporato (2007), Coupled moisture and microbial dynamics in unsaturated soils, Water Resour. Res., 43, W07444,
doi:10.1029/2006WR005367.
Maggi, F., C. Gu, W. J. Riley, G. M. Hornberger, R. T. Venterea, T. Xu, N.
Spycher, C. Steefel, N. L. Miller, and C. M. Oldenburg (2008), A mechanistic treatment of the dominant soil nitrogen cycling processes: Model
development, testing, and application, J. Geophys. Res., 113, G02016,
doi:10.1029/2007JG000578.
2570
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
Manzoni, S., and A. Porporato (2009), Soil carbon and nitrogen mineralization: Theory and models across scales, Soil Biol. Biochem., 41, 1355–
1379, doi:10.1016/j.soilbio.2009.02.031.
Manzoni, S., and A. Porporato (2011), Common hydrologic and biogeochemical controls along the soil-stream continuum, Hydrol. Process., 25,
1355–1360, doi:10.1002/hyp.7938.
Mao, D. Z., and K. A. Cherkauer (2009), Impacts of land-use change on
hydrologic responses in the Great Lakes region, J. Hydrol., 374(1-2), 71–
82, doi:10.1016/j.jhydrol.2009.06.016.
Mascaro, G., E. R. Vivoni, and R. Deidda (2010), Downscaling soil moisture in the southern Great Plains through a calibrated multifractal model
for land surface modeling applications, Water Resour. Res., 46, W08546,
doi:10.1029/2009WR008855.
Maxwell, R. M., and S. J. Kollet (2008a), Interdependence of groundwater
dynamics and land-energy feedbacks under climate change, Nat. Geosci.,
1(10), 665–669, doi:10.1038/Ngeo315.
Maxwell, R. M., and S. J. Kollet (2008b), Quantifying the effects of threedimensional subsurface heterogeneity on Hortonian runoff processes
using a coupled numerical, stochastic approach, Adv. Water Resour.,
31(5), 807–817, doi:10.1016/j.advwatres.2008.01.020.
McDonnell, J. J., et al. (2007), Moving beyond heterogeneity and process
complexity: A new vision for watershed hydrology, Water Resour. Res.,
43, W07301, doi:10.1029/2006WR005467.
Michigan Department of Natural Resources, MNDR (2010), 2001 IFMAP/
GAP lower Peninsula Land Cover. [Available at http://www.mcgi.state.mi.us/mgdl/?rel¼thext&action¼thmname&cid¼5& cat¼LandþCoverþ
2001, accessed 28 Nov. 2009.].
Miguez-Macho, G., and Y. Fan (2012), The role of groundwater in the Amazon water cycle: 2. Influence on seasonal soil moisture and evapotranspiration, J. Geophys. Res., 117, D15114, doi:10.1029/2012JD017540.
Miguez-Macho, G., Y. Fan, C. P. Weaver, R. Walko, and A. Robock
(2007), Incorporating water table dynamics in climate modeling: 2. Formulation, validation, and soil moisture simulation, J. Geophys. Res., 112,
D13108, doi:10.1029/2006JD008112.
Mishra, V., K. A. Cherkauer, and S. Shukla (2010), Assessment of drought
due to historic climate variability and projected future climate change in
the Midwestern United States, J. Hydrometeorol., 11(1), 46–68,
doi:10.1175/2009jhm1156.1.
Mu, Q., F. A. Heinsch, M. Zhao, and S. W. Running (2007), Development
of a global evapotranspiration algorithm based on MODIS and global
meteorology data, Remote Sens. Environ., 111(4), 519–536,
doi:10.1016/j.rse.2007.04.015.
Mu, Q., M. Zhao, and S. W. Running (2011), Improvements to a MODIS
global terrestrial evapotranspiration algorithm, Remote Sens. Environ.,
115(8), 1781–1800, doi:10.1016/j.rse.2011.02.019.
NASA Land Processes Distributed Active Archive Center (LP DAAC)
(2012), Land Processes Distributed Active Archive Center, MOD15A2,
USGS/Earth Resources Observation and Science (EROS) Center, Sioux
Falls, S. D., available at (https://lpdaac.usgs.gov/get_data)
National Climatic Data Center (NCDC) (2010), National Climatic Data
Center [Available at http://www.ncdc.noaa.gov/oa/climate/climatedata.
html#daily.].
Neitsch, S. L., J. G. Arnold, J. R. Kiniry, and J. R. Williams (2005), Soil
and Water Assessment Tool Theoretical Documentation Version 2005,
U.S. Dep. of Agric.–Agric. Res Service, Temple, Tex.
Niu, G.-Y., and Z. L. Yang (2006), Effects of frozen soil on snowmelt runoff and
soil water storage at a continental scale, J. Hydrometeorol., 7(5), 937–952.
Niu, G.-Y., and Z.-L. Yang (2007), An observation-based formulation of
snow cover fraction and its evaluation over large North American river
basins, J. Geophys. Res., 112, D21101, doi:10.1029/2007JD008674.
Niu, G.-Y., C. Paniconi, P. A. Troch, R. L. Scott, M. Durcik, X. Zeng, T.
Huxman, and D. C. Goodrich (2013), An integrated modelling framework of catchment-scale ecohydrological processes: 1. Model description and tests over an energy-limited watershed, Ecohydrology, n/a-n/a,
doi:10.1002/eco.1362.
Oishi, A. C., R. Oren, K. A. Novick, S. Palmroth, and G. G. Katul (2010),
Interannual invariability of forest evapotranspiration and its consequence
to water flow downstream, Ecosystems, 13(3), 421–436, doi:10.1007/
s10021-010-9328-3.
Oleson, K. W., et al. (2010), Technical Description of version 4.0 of the
Community Land Model (CLM). NCAR Technical Note, Rep. NCAR/
TN-478þSTR, National Center for Atmospheric Research, Boulder, Colo.
Oztan, M. (2011), GIS-enabled modeling of Michigan’s groundwater systems, Ph.D. dissertation, Department of Civil & Environmental Engineering, Mich. State Univ., East Lansing.
Pan, Y.-X., and X.-P. Wang (2009), Factors controlling the spatial variability of surface soil moisture within revegetated-stabilized desert ecosystems of the Tengger Desert, Northern China, Hydrol. Process., 23(11),
1591–1601, doi:10.1002/hyp.7287.
Pellenq, J., J. Kalma, G. Boulet, G.-M. Saulnier, S. Wooldridge, Y. Kerr,
and A. Chehbouni (2003), A disaggregation scheme for soil moisture
based on topography and soil depth, J. Hydrol., 276, 112–127,
doi:10.1016/S0022–1694(03)00066-0.
Qiu, Y., B. J. Fu, J. Wang, and L. D. Chen (2001), Soil moisture variation
in relation to topography and land use in a hillslope catchment of the
Loess Plateau, China, J. Hydrol., 240(3-4), 243–263, doi:10.1016/
s0022–1694(00)00362-0.
Reich, P. B., S. E. Hobbie, T. Lee, D. S. Ellsworth, J. B. West, D. Tilman, J.
M. H. Knops, S. Naeem, and J. Trost (2006), Nitrogen limitation constrains sustainability of ecosystem response to CO2, Nature, 440(7086),
922–925, doi:10.1038/nature04486.
Rihani, J. F., R. M. Maxwell, and F. K. Chow (2010), Coupling groundwater and land surface processes: Idealized simulations to identify
effects of terrain and subsurface heterogeneity on land surface energy
fluxes, Water Resour. Res., 46, W12523, doi:10.1029/2010WR009111.
Riveros-Iregui, D. A., B. L. McGlynn, R. E. Emanuel, and H. E. Epstein
(2012), Complex terrain leads to bidirectional responses of soil respiration to inter-annual water availability, Global Change Biol., 18, 749–
756, doi:10.1111/j.1365–2486.2011.02556.x.
Robertson, G. P., M. A. Huston, F. C. Evans, and J. M. Tiedje (1988), Spatial variability in a successional plant community: Patterns of nitrogen
availability, Ecology, 69(5), 1517–1524, doi:10.2307/1941649.
Sakaguchi, K., and X. Zeng (2009), Effects of soil wetness, plant litter, and
under-canopy atmospheric stability on ground evaporation in the Community Land Model (CLM3.5), J. Geophys. Res., 114, D01107,
doi:10.1029/2008JD010834.
Salvucci, G. D., and D. Entekhabi (1995), Hillslope and climatic controls
on hydrologic fluxes, Water Resour. Res., 31(7), 1725–1739,
doi:10.1029/95WR00057.
Schimel, D. S., B. H. Braswell, and W. J. Parton (1997), Equilibration of
the terrestrial water, nitrogen, and carbon cycles, Proc. Natl. Acad. Sci.
U. S. A., 94(16), 8280–8283, doi:10.1073/pnas.94.16.8280.
Sellers, P. J. (1985), Canopy reflectance, photosynthesis and transpiration,
Int. J. Remote Sens., 6(8), 1335–1372.
Shen, C. P. (2009), A process-based distributed hydrologic model and its
application to a Michigan watershed, Ph.D. dissertation, 270 pp., Department of Civil & Environmental Engineering, Mich. State Univ., East
Lansing.
Shen, C. P., and M. S. Phanikumar (SP10) (2010), A process-based, distributed hydrologic model based on a large-scale method for surface-subsurface coupling, Adv. Water Resour., 33(12), 1524–1541, doi:10.1016/
j.advwatres.2010.09.002.
Simard, A. (2007), Predicting groundwater flow and transport using Michigan’s statewide wellogic database, Ph.D. dissertation, 109 pp., Department of Civil & Environmental Engineering, Mich. State Univ., East
Lansing.
Sinha, T., and K. A. Cherkauer (2008), Time series analysis of soil freeze
and thaw processes in Indiana, J. Hydrometeorol., 9(5), 936–950,
doi:10.1175/2008jhm934.1.
Sivapalan, M. (2003), Prediction in ungauged basins: A grand challenge
for theoretical hydrology, Hydrol. Process., 17(15), 3163–3170,
doi:10.1002/hyp.5155.
Sivapalan, M., S. E. Thompson, C. J. Harman, N. B. Basu, and P. Kumar
(2011), Water cycle dynamics in a changing environment: Improving
predictability through synthesis, Water Resour. Res., 47, W00J01,
doi:10.1029/2011WR011377.
Soil Survey Staff (2010), Soil Survey Geographic Database, Natural
Resources Conservation Service, United States Dep. of Agric. [Available
at http://soils.usda.gov/survey/geography/ssurgo/, accessed 1 June
2010.].
Sokolov, A. P., D. W. Kicklighter, J. M. Melillo, B. S. Felzer, C. A.
Schlosser, and T. W. Cronin (2008), Consequences of considering carbon-nitrogen interactions on the feedbacks between climate and the terrestrial carbon cycle, J. Clim., 21(15), 3776–3796, doi:10.1175/
2008jcli2038.1.
Suyker, A. E., and S. B. Verma (2009), Evapotranspiration of irrigated and
rainfed maize-soybean cropping systems, Agric. For. Meteorol., 149(34), 443–452, doi:10.1016/j.agrformet.2008.09.010.
Tague, C. L., and L. E. Band (2004), RHESSys: Regional hydro-ecologic
simulation system—An Object-Oriented approach to spatially distributed
2571
SHEN ET AL.: EVALUATING CONTROLS ON HYDROLOGIC AND VEGETATION DYNAMICS
modeling of carbon, water, and nutrient cycling, Earth Interact., 8, 1–42,
doi:10.1175/1087–3562(2004)8<1:RRHSSO>2.0.CO;2.
Thompson, S. E., et al. (2011), Patterns, puzzles and people: implementing
hydrologic synthesis, Hydrol. Process., 25(20), 3256–3266, doi:10.
1002/Hyp.8234.
Thornton, P. E., and N. A. Rosenbloom (2005), Ecosystem model spin-up:
Estimating steady state conditions in a coupled terrestrial carbon and
nitrogen cycle model, Ecol. Modell., 189(1-2), 25–48, doi:10.1016/
j.ecolmodel.2005.04.008.
Thornton, P. E., and N. E. Zimmermann (2007), An improved canopy integration scheme for a land surface model with prognostic canopy structure, J. Clim., 20(15), 3902–3923, doi:10.1175/Jcli4222.1.
Toon, O. B., C. P. Mckay, T. P. Ackerman, and K. Santhanam (1989),
Rapid calculation of radiative heating rates and photodissociation rates
in inhomogeneous multiple-scattering atmospheres, J. Geophys. Res.,
94(D13), 16,287–16,301.
Troch, P. A., C. Paniconi, and E. E. van Loon (2003), Hillslope-storage Boussinesq model for subsurface flow and variable source areas
along complex hillslopes: 1. Formulation and characteristic
response, Water Resour. Res., 39(11), 3-1–3-10, doi:10.1029/2002W
R001728.
Turkoglu, N. (2010), Analysis of urban effects on soil temperature in
Ankara, Environ. Monit. Assess., 169(1-4), 439–450, doi:10.1007/
s10661-009-1187-z.
van Dam, J. C., and R. A. Feddes (2000), Numerical simulation of infiltration, evaporation and shallow groundwater levels with the Richards
equation, J. Hydrol., 233(1-4), 72–85.
Vertessy, R. A., T. J. Hatton, R. G. Benyon, and W. R. Dawes (1996),
Long-term growth and water balance predictions for a mountain ash (Eucalyptus regnans) forest catchment subject to clear-felling and regeneration., Tree Physiol., 16, 221–232.
Vivoni, E. R. (2012), Spatial patterns, processes and predictions in ecohydrology: Integrating technologies to meet the challenge, Ecohydrology,
5(3), 235–241, doi:10.1002/eco.1248.
Vukovich, F. M. (1983), An analysis of the ground temperature and reflectivity pattern about St. Louis, Missouri, using HCMM satellite data,
J. Clim. Appl. Meteorol., 22(4), 560–571.
Wagener, T., M. Sivapalan, P. A. Troch, B. L. McGlynn, C. J. Harman, H.
V. Gupta, P. Kumar, P. S. C. Rao, N. B. Basu, and J. S. Wilson (2010),
The future of hydrology: An evolving science for a changing world,
Water Resour. Res., 46, W05301, doi:10.1029/2009WR008906.
White, M. A., P. E. Thornton, and S. W. Running (1997), A continental
phenology model for monitoring vegetation responses to interannual climatic variability, Global Biogeochem. Cycles, 11(2), 217–234.
Wilson, C. A., R. J. Mitchell, J. J. Hendricks, and L. R. Boring (1999), Patterns and controls of ecosystem function in longleaf pine-wiregrass savannas. II. Nitrogen dynamics, Canad. J. For. Res., 29(6), 752–760,
doi:10.1139/cjfr-29-6-752.
Wilson, D. J., A. W. Western, and R. B. Grayson (2004), Identifying and quantifying sources of variability in temporal and spatial soil moisture observations, Water Resour. Res., 40, W02507, doi:10.1029/2003WR002306.
Wutzler, T., and M. Reichstein (2008), Colimitation of decomposition by
substrate and decomposers—A comparison of model formulations, Biogeosciences, 5, 749–759.
Xu, R. and I. C. Prentice (2008), Terrestrial nitrogen cycle simulation with
a dynamic global vegetation model, Global Change Biol., 14, 1745–
1764, doi:10.1111/j.1365–2486.2008.01625.x.
Zeng, X., M. Zhao, and R. E. Dickinson (1998), Intercomparison of bulk
aerodynamic algorithms for the computation of sea surface fluxes using
TOGA COARE and TAO Data, J. Clim., 11, 2628–2644, doi:10.1175/
1520-0442(1998)011<2628:IOBAAF>2.0.CO;2.
Zeng, X., M. Shaikh, Y. Dai, R. E. Dickinson, and R. Myneni (2002), Coupling of the Common Land Model to the NCAR Community Climate
Model, J. Clim., 15, 1832–1854, doi:10.1175/1520-0442(2002)015<1832:
COTCLM>2.0.CO;2.
Zhu, Q., and H. Lin (2011), Influences of soil, terrain, and crop growth on
soil moisture variation from transect to farm scales, Geoderma, 163(1-2),
45–54, doi:10.1016/j.geoderma.2011.03.015.
2572