HESSHydrology and Earth System SciencesHESSHydrol. Earth Syst. Sci.1607-7938Copernicus GmbHGöttingen, Germany10.5194/hess-19-4317-2015Impacts of grid resolution on surface energy fluxes simulated with an
integrated surface-groundwater flow modelShresthaP.pshrestha@uni-bonn.deSulisM.https://orcid.org/0000-0002-3149-4096SimmerC.https://orcid.org/0000-0003-3001-8642KolletS.Meteorological Institute, University of Bonn, Bonn, GermanyForschungszentrum Jülich GmbH, Jülich, GermanyCentre for High-Performance Scientific Computing in Terrestrial
Systems, Geoverbund ABC/J, Jülich, GermanyP. Shrestha (pshrestha@uni-bonn.de)23October201519104317432628April20153July201522September201523September2015This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://hess.copernicus.org/articles/19/4317/2015/hess-19-4317-2015.htmlThe full text article is available as a PDF file from https://hess.copernicus.org/articles/19/4317/2015/hess-19-4317-2015.pdf
The hydrological component of the Terrestrial Systems Modeling Platform
(TerrSysMP), which includes integrated surface-groundwater flow, was used to
investigate the grid resolution dependence of simulated soil moisture, soil
temperature, and surface energy fluxes over a sub-catchment of the Rur,
Germany. The investigation was motivated by the recent developments of new
earth system models, which include 3-D physically based groundwater models
for the coupling of land–atmosphere interaction and subsurface
hydrodynamics. Our findings suggest that for grid resolutions between 100 and
1000 m, the non-local controls of soil moisture are highly grid
resolution dependent. Local vegetation, however, strongly modulates the
scaling behavior, especially for surface fluxes and soil temperature, which
depends on the radiative transfer property of the canopy. This study also
shows that for grid resolutions above a few 100 m, the variation of
spatial and temporal patterns of sensible and latent heat fluxes may
significantly affect the resulting atmospheric mesoscale circulation and
boundary layer evolution in coupled runs.
Introduction
In recent years, a growing number of earth system modeling
platforms attempted to include physically based hydrological models, with
lateral flow and groundwater surface water interactions, to study the
linkages between land–atmosphere and subsurface hydrodynamics (e.g., Anyah
et al., 2008; Maxwell et al., 2011; Shrestha et al., 2014; Butts et
al., 2014; Larsen et al., 2014). These studies show that the inclusion of
groundwater dynamics improves the simulated spatial variability in root zone
soil moisture and groundwater table depth, and shows the potential for
improved forecasts of the whole terrestrial system. However, as soon as one
moves from column-based land surface models to 3-D models with lateral flows,
a new dimension of spatial complexity is added where scaling issues become
highly relevant (Becker and Braun, 1999). This is mainly due to the
introduction of non-local controls on soil moisture patterns (e.g., patterns
of soil moisture dominated by lateral fluxes of surface and subsurface flow)
as earlier identified by Grayson et al. (1997), which also depend on grid
resolution. For spatial extents of 100 km and above atmospheric
models are still run at grid resolutions ≥ 1 km due to
computational limitations or physical parameterizations, and hydrological
models coupled the atmospheric models are usually run at similar grid
resolutions, which may however be inadequate to correctly simulate subsurface
flow. Hyper-resolution models have already been suggested by, e.g., Wood et
al. (2011), while Beven and Cloke (2012) have suggested the need for
spatial-scale-dependent subgrid-scale parameterizations to adequately
simulate soil moisture variability.
In reality, catchments exhibit variability and heterogeneity at a range of
scales (Blöschl and Sivapalan, 1995), while in numerical models, the
variability of soil moisture, soil temperature and surface fluxes can only be
controlled by the heterogeneity at the chosen grid resolution. Previous
studies with offline hydrological models have shown that the aggregation of
topography to coarser grid resolution (e.g., 1 km) has a strong
impact on the water balance (e.g., Zhang and Montgomery, 1994; Kuo et
al., 1999; Vivoni et al., 2005; Bormann, 2006; Herbst et al., 2006; Giertz et
al., 2006; Dixon and Earls, 2009; Sciuto and Diekkrueger, 2010; Sulis et
al., 2011). Many of these studies focused primarily on the
spatial-scale-dependent behavior of catchment discharge, groundwater table
depth and catchment mean soil moisture. A better understanding of the
spatial-scale dependence of the simulated patterns of soil moisture,
temperature and surface fluxes is required, however, when such models are
coupled to atmosphere models. In an idealized setup using a mosaic approach,
Shrestha et al. (2014) demonstrated the importance of subgrid-scale
topography on topographically driven surface–subsurface flow for
land–atmosphere interactions and stressed the importance of accurate
simulation of spatiotemporal variability of surface fluxes for the evolution
of the terrestrial system as a whole.
The aim of this study is to examine the effects of resolution-dependent model
heterogeneity using the Terrestrial System Modeling Platform (TerrSysMP,
Shrestha et al., 2014; Gasper et al., 2014; Sulis et al., 2015) on the
variability of modeled soil moisture, soil temperature and surface fluxes in
a temperate climate when no subgrid-scale parameterizations are used. The
rest of the manuscript is organized as follows: Sect. 2 describes the
modeling tool used for the study; experiment design and setup of the
catchment are discussed in Sect. 3. Topography heterogeneity analysis is
presented in Sect. 4, while results and discussions are presented in Sect. 5,
and conclusions in Sect. 6.
Modeling tool
The hydrological component of TerrSysMP consists of the NCAR Community Land
Model CLM3.5 (Oleson et al., 2008) and the 3-D variably saturated groundwater
and surface water flow code ParFlow (Ashby and Falgout, 1996; Jones and
Woodward, 2001; Kollet and Maxwell, 2006; Maxwell, 2013). The two models
(Fig. 1) are coupled using the OASIS3 external coupler (Valcke, 2013). In the
sequential information exchange procedure, ParFlow sends the updated relative
saturation (Sw) and pressure (Ψ) for the top 10 layers to
CLM. In turn, CLM sends the depth-differentiated source and sink terms for
soil moisture (top soil moisture flux (qrain), soil
evapotranspiration (qe)) for the top 10 soil layers to ParFlow
(see Fig. 1). A more detailed description of the coupling can be found in
Shrestha et al. (2014). In this study, the hydrological component of
TerrSysMP is decoupled from its atmospheric component and forced with
spatially distributed atmospheric forcing data at 2.8 km spatial resolution
and hourly temporal resolution (air temperature (T), wind speed (U),
specific humidity (QV), total precipitation (Rain), pressure (P), and
incoming shortwave (SW) and longwave (LWdn)) from COSMO-DE (Baldauf et
al., 2011) analysis data of the German Weather Service (DWD).
Schematic diagram of the hydrological component of the Terrestrial
System Modeling Platform (TerrSysMP). OASIS3 is the driver of the component
models for land surface (CLM) and subsurface (ParFlow). The configuration
file for OASIS3 prescribes the end-point data exchange between the component
models in a sequential manner. The variables exchanged between the two models
are relative saturation (Sw), soil pressure head (Ψ), top
soil moisture flux (qrain), soil evapotranspiration
(qe), air temperature (T), wind speed (U), specific humidity
(QV), total precipitation (Rain), pressure (P), and incoming shortwave (SW)
and longwave (LWdn).
Numerical experiment design
The model was set up for a sub-catchment of the Rur River (TR32 test bed
site; Vereecken et al., 2010; Simmer et al., 2014), on the northern foothills
of the lower Eifel mountain range with an approximated drainage area of
325 km2 (Fig. 2a). The sub-catchment encompasses the tributary
Wehebach, which merges with the River Inde. The elevation in the model domain
reaches from 50 to 600 m from north to south (Fig. 2b), with mostly
agricultural crops (c1n) near the foothills and needleleaf evergreen trees
(nle) and broadleaf deciduous trees (bld) along the sloping terrain
(Fig. 2c). The urban areas are represented in the model as agricultural
canopy (c1f) with a fixed leaf area index (LAI = 0.6). Topography and
land use are based on the 90 m resolution Shuttle Radar Topography Mission
(SRTM) data and 15 m resolution data available from the TR32 database
(Waldhoff, 2012), respectively.
(a) Topography of the Rur catchment, which lies at the
border of Germany, Belgium, Luxembourg and the Netherlands.
(b) Topography of the sub-catchment of the Rur used in this study
(catchment of the River Inde including the Wehebach tributary in the east),
overlaid with the stream networks. (c) Land-use map of the
sub-catchment.
The SRTM data were aggregated to 120, 240, 480 and 960 m horizontal grid
resolution for the model domain setup (Table 1) by first interpolating the
90 m topography to 120 m using bilinear interpolation before
aggregating to the coarser resolutions. Land use was aggregated by specifying
only the dominant plant functional type (PFT) at the coarser resolution. The
chosen grid resolutions are within the limits where a positive spatial
autocorrelation of the topographic index exists (Cai and Wang, 2006), and
roughly cover the range of grid resolutions between large-eddy simulation
(LES) and mesoscale atmospheric modeling. For fully coupled mesoscale
modeling with grid resolutions ≥ 1000 m (atmosphere component
model), the above selected grid resolutions can be used for the hydrological
component in TerrSysMP with the mosaic approach to better resolve the
heterogeneity of topography, land use and geology.
The model setup for all grid resolutions used the same 10 vertically
stretched layers (2–100 cm from top to bottom) followed by
20 constant depth levels (135 cm) extending to 30 m below the
land surface. A uniform soil texture was used for this study by keeping the
soil parameters spatially constant. The subsurface parameters were set as
follows: saturated hydraulic conductivity, Ks=0.00034mh-1; van Genuchten parameters, α=2.1 and n=2.0m-1; and porosity, ϕ=0.4449. This removes any impact
of soil heterogeneity on non-local controls of simulated soil moisture
variability and scaling of soil hydraulic properties. The soil moisture
profile for all setups was initialized with a horizontally homogeneous
hydrostatic pressure head and a water table depth at 5 m from the
surface. The soil temperature was also initialized horizontally homogeneous
with a uniform temperature of 10 ∘C for all levels. A time step of
3600 s was used, and the simulation was integrated for 6 years using
hourly atmospheric forcing from COSMO-DE analysis data. The same atmospheric
forcing data for the year 2009 were used recursively for the 6 years. The
model outputs were averaged over 5 days for the analysis. The NCAR Command
Language (NCL; NCAR, 2013) was used for data analysis.
Model setup indicating the horizontal grid resolution (ΔX=ΔY) and domain discretization.
Model setupΔX=ΔY (m)NX × NY × NZd120120190×220×30d240240100×110×30d48048050×70×30d96096020×30×30
Heterogeneity analysis of topography at different grid resolutions.
Heterogeneity analyses of the topography at different grid resolutions are
summarized in Table 2. The profile and plan curvature represents the flow
acceleration and convergent/divergent flow, respectively. The profile
curvature is parallel to the direction of maximum slope and a
positive/negative value indicates that the surface is upwardly concave/convex
at the grid cell. The plan curvature is perpendicular to the direction of the
flow and the positive/negative value indicates that the surface is
convex/concave to the side at the grid cell (see
https://resources.arcgis.com for pictorial descriptions). The
distributions of plan and profile curvature change with coarsening of grid
resolution. The plan curvature is negatively skewed at small grid resolution,
and the skewness decreases with the coarsening, while skewness changes for
the profile curvature are negligible. However, the kurtosis of both the plan
and profile curvatures decreases exponentially with higher exponential power
for plan curvature. This is also qualitatively visible in the streamline maps
of D4 flow direction and the local slopes for the sub-catchment at the
different grid resolutions (Fig. 3a, b, c, d): the aggregation of topography
results in smoothing of slope magnitudes and the filtering of small-scale
convergence and divergence zones. Without sub-grid-scale parameterization,
this spatial filtering will impact lateral flow and simulated mean grid cell
soil moisture distributions (Shrestha et al., 2014). Similar to the findings
of Quinn et al. (1995), grid coarsening also affects the location of the
water divides and makes it difficult to accurately delineate the catchment
contributing area especially for d480 and d960.
(a–d) Streamlines resulting from D4 flow directions along
with slopes for the four grid resolutions. The black solid outline represents
the catchment boundary at 120 m resolution.
Results and discussions
The simulated unsaturated storage (Sunsat) at different grid
resolutions for the sub-catchment showed different temporal evolutions
(Fig. 4). The unsaturated storage was normalized by the modeled sub-catchment
area to account for differences in catchment size at different grid
resolutions. The increase in Sunsat during the first year reflects
the adjustment of the subsurface storage from the horizontally homogeneous
hydrostatic initial condition, with a ground to water table depth of
5 m. In the first half of this year, the finer grid resolutions
adjust faster due to a more efficient drainage mechanism. In the second year,
Sunsat starts to decrease gradually, reaching a quasi-equilibrium
in all simulations in the fifth year. This steady-state value range is lower
for the coarser grid resolutions, caused by a higher groundwater table, which
results from a less efficient drainage combined with higher infiltration at
the lower resolutions. This is also illustrated in Fig. 4b, which shows that
the average annual unsaturated storage (Sunsat‾t) in
the sixth year is concurrent with the average slope of the sub-catchment. The
decrease in average catchment slope with resolution coarsening is, however,
also accompanied by decreasing plan and profile curvature kurtosis (see
Table 2). These results are consistent with, e.g., Kuo et al. (1999), who
related the increase in average soil moisture contents with grid coarsening
to decreasing slope gradient and curvature variations. Sulis et al. (2011)
also explained catchment wetness in terms of storage and ground to water
table depth via decreasing local slopes and plan curvature variations due to
aggregation effects. Different Sunsat‾t with grid
resolutions reflects different spatial soil moisture variability, which in
turn influences simulated land–atmosphere interactions.
(a) Time series of unsaturated storage per unit catchment
area (Sunsat) for the sub-catchment at the four grid resolutions.
(b) Relationship between the annual average unsaturated storage
(Sunsat‾t) and the average catchment slope for the
four grid resolutions.
Figure 5 shows the distribution of average top 10 cm relative soil moisture
(Sw), average top 10 cm soil temperature (Tsoil),
sensible heat flux (SH) and latent heat flux (LH) for time periods when the
model exhibits strong coupling with the atmospheric forcing. We assume strong
coupling when the 5-day mean incoming solar radiation exceeds
128 Wm-2, which corresponds to the period from April to September.
We further filter for different PFTs to analyze the linkages between local
vegetation and the non-local controls of soil moisture patterns with grid
coarsening. Their respective temporal and catchment-averaged values
(Sw‾x,t, Tsoil‾x,t,
SH‾x,t and LH‾x,t) are indicated
as solid markers (Fig. 5) and summarized in Table 3 for the discussion below.
For relative soil moisture, grid coarsening leads to a sharp decrease in
interquartile range and an increase in the 25 % quartile, median and mean
value, i.e., reduced variability and higher mean simulated soil moisture.
This is true for all PFTs.
While no significant loss in interquartile range with reduced resolution is
observed for temperature and the turbulent fluxes, a clear PFT-dependent
scaling with grid coarsening exists, especially for crop PFTs and the crop
PFT with a fixed low LAI. Average relative soil moisture increases by
30 % for trees and 23 % for crops when coarsening the grid resolution
from 120 to 960 m. The difference between the different PFTs is
partly related to their spatial location: trees are mostly located in steeper
terrain compared to crops (according to the distribution of slopes for the
different PFTs, not shown here). For trees, grid coarsening leads to lower
Tsoil‾x,t by 0.6 and 0.3 ∘C for needleleaf
evergreen trees and broadleaf deciduous trees, respectively, while for crops,
Tsoil‾x,t is lowered by almost 1 ∘C. Thus
forested grid cells exhibit higher grid resolution sensitivity for soil
moisture and lower sensitivity for soil temperature, while grid cells with
crops show the inverse.
Scaling behavior of relative soil moisture (Sw), soil
temperature (Tsoil), sensible heat flux (SH) and latent heat flux
(LH) for the four different canopy covers (nle, bld, c1n and c1f). The solid
markers indicate the mean value of the distribution for each grid
resolution.
Mean sub-catchment relative soil moisture (Sw), soil
temperature Tsoil, sensible SHtavg and latent fluxes
LHtavg for grid columns with land-use classes nle (needleleaf
evergreen tree), bld (broadleaf deciduous tree), c1n (crops with seasonal
LAI), and c1f (crops with fixed LAI).
These findings can be attributed to the PFT-specific transmissivity for solar
radiation, or the partitioning of absorbed solar radiation by vegetation and
the ground. Needleleaf evergreen trees (nle) absorb more and transmit less
solar radiation to the ground compared to broadleaf deciduous trees (bld),
while the crop type with variable LAI (c1n) absorbs more solar radiation and
transmits less compared to the constant LAI type (c1f). The PFT-specific
optical parameters, including albedo and amplitudes of seasonal LAI, control
the variability in solar radiation partitioning. This directly modulates the
partitioning of sensible and latent heat flux and constitutes a local
vegetation control.
For all grid resolutions, the inter-PFT flux differences are much higher than
the intra-PFT differences due to different grid resolutions. In general, grid
cells with crops are more sensitive to grid resolution changes. A 60 %
decrease in SH‾x,t and a 35 % increase in
LH‾x,t are observed for c1f, with the grid coarsening
from 120 to 960 m. SH‾x,t decreases by 30 %
and LH‾x,t increases by 16 % for c1n. For forest,
SH‾x,t decreases only by about 20 and 11 %, while
LH‾x,t increases only by 8 and 11 % for bld and
nle, respectively. In general, the latent heat flux increases and sensible
heat flux decreases for all PFTs for coarser grid resolutions, while
amplitudes of change are PFT-dependent, which suggests a local control by the
PFT. This also suggests that the non-local control of soil moisture, which is
affected by change in grid resolution, also controls the partitioning of
surface energy fluxes, especially for PFTs transmitting more solar radiation
to the ground. The average relative soil moisture for the sub-catchment in
this study was relatively wet; for drier regimes, the effects of model grid
resolution on surface fluxes may be stronger.
The above findings bear important consequences in terms of spatial
variability of surface fluxes, which may affect the evolution of the
atmospheric boundary layer, induce mesoscale circulations, and even lead to
the formation of clouds and precipitation (Avissar and Schmidt, 1998; Baidya
Roy and Avissar, 2002). This spatial variability is illustrated using a cross
section along the sub-catchment. Figure 6a shows the spatial heterogeneity of
topography and land use along cross-section AA′ indicated in Fig. 2 for
different grid resolutions. The difference in variability of topography is
not visible along the cross section due to the plot scale, but the effect of
grid resolution on the simulated average top 10 cm soil moisture
Sw‾t is obvious in Fig. 6b. With grid coarsening,
the finer scale plan and profile curvatures are filtered, which reduces
drainage efficiency, increases the simulated grid cell soil moisture, and
damps the spatial variability. This effect is more pronounced from d240 to
d480 and d960 than from d120 to d240. Quantitatively, the spatial average and
standard deviation for Sw‾t along the cross-section
AA′ are 0.70±0.16, 0.75±0.17, 0.90±0.10 and 0.93±0.04
for d120, d240, d480 and d960, respectively, which clearly shows the increase
in soil wetness and decrease in variability with coarsening with largest
changes between d240 and d480.
Horizontal variations over cross-section AA′ in Fig. 2:
(a) topography and land use, (b) annual average top 10 cm
relative soil moisture and (c) annual average Bowen ratio at
different model grid resolutions.
The strong scaling behavior of soil moisture also modulates the partitioning
of surface energy fluxes. Figure 6c shows the annual average Bowen ratio
(ratio of sensible to latent heat flux) along cross-section AA′; its
profile matches well with the PFT profile, indicating local vegetation
control on the spatial pattern of surface flux partitioning. This could be
partly enhanced due to the wet soil condition in the sub-catchment for the
simulated period. However, the change in the non-local control of soil
moisture with grid resolution also contributes to the profile of surface flux
partitioning visible as a perturbation in the amplitudes of the Bowen ratio
for d960. Some perturbations are also caused by different PFTs for finer grid
resolution. Table 4 summarizes the statistics of the Bowen ratio profile
along the cross section and shows that its dominant pattern is strongly
controlled by the PFT pattern. For trees, nle has a higher Bowen ratio
(> 1) than bld (< 1), which is mainly due to differences in plant
physiological properties, consistent with observations (Baldocchi et
al., 1997). For crops, c1f has a higher Bowen ratio compared to c1n, which is
due to the LAI difference. For both trees and crops, the Bowen ratio in
general decreases with grid coarsening. Coarsening from d120 to d960
decreases the Bowen ratio by 20, 28, 39, and 73 % for nle, bld, c1n and
c1f, respectively. The most significant change is found for crops with low
LAI. Radiation absorbed by the ground plays a significant role in
amplifying/attenuating the grid resolution dependence of surface flux
partitioning. Again, it has to be mentioned that this statement may be valid
only for wet regimes.
Annual average Bowen ratio and standard deviation along the
cross-section AA′ for the land-use classes nle (needleleaf evergreen tree),
bld (broadleaf deciduous tree), c1n (crops with seasonal LAI), and c1f (crops
with fixed LAI).
Figure 7 shows the scatterplot between the Bowen ratio and average top 10 cm
relative soil moisture along cross-section AA′ over the averaged time
period filtered for different PFTs to illustrate the PFT-related dependence
of the Bowen ratio on relative soil moisture. Large scatter for d480 and
d960 m is found for Sw<0.7, while for d120 and d240,
large scatter is only found for Sw≤0.4. Thus, the Bowen ratio
distribution shifts with grid resolution. A linear regression between Bowen
ratio and relative soil moisture gives a first-order estimate of the Bowen
ratio dependence on relative soil moisture. Tree canopies exhibit more
variability in the Bowen ratio than crops. For trees, nle exhibits stronger
scaling behavior with relative soil moisture than bld. For crops, c1f
exhibits stronger scaling behavior with relative soil moisture than c1n. The
results again show that crops with low LAI have a stronger influence on flux
partitioning with grid resolution.
We also evaluated the grid resolution effects on fluxes using the mosaic
approach by aggregating the simulated surface sensible and latent heat fluxes
for d120, d240, and d480 to 960 m resolution. Figure 8a shows the time
series of the 5-day average of sensible heat flux along cross-section AA′
for d120. It shows the strong seasonal cycle of sensible heat flux that
correlates with the seasonal cycle of net radiation (not shown here), and
also the strong gradient along AA′, owing to the different canopy cover.
The differential heating along AA′ potentially generates mesoscale boundary
layer circulations embedded in the local topographic circulation, whose
strength would also depend on the mean heating rates of the catchment and the
synoptic wind strength as indicated in many previous studies (Lemone et
al., 2002; Baidya Roy and Avissar, 2002; Grossman et al., 2005). Figure 8b–d
shows the difference in the time series of SH along cross-section AA′
between the coarser grid resolutions and d120. With grid coarsening the
amplitude of the SH difference increases, suggesting an overall decrease in
SH as also observed from the bulk quantities. Some d960 grid cells also
exhibit an increase in SH mainly due to the changing dominant PFT, when the
subgrid cells consist of trees and crops. Similarly, Fig. 9a shows the time
series of 5-day average latent heat flux along cross-section AA′ for d120,
which also exhibits a strong seasonal cycle and a strong gradient along
AA′. The difference in time series of LH along AA′ for coarser
resolutions with respect to d120 shows the sharp increase in latent heat
flux. The decrease and increase in SH and LH, respectively, are particularly
high between 480 and 960 m resolution. Thus, when using coarser grid
resolutions for the hydrological model, coupled to the atmospheric model, the
simulated boundary layer would be relatively moister and cooler.
Distribution of the Bowen ratio as a function of relative soil
moisture along cross-section AA′. The distribution is plotted separately
for different land-use types in the sub-catchment.
(a) Time evolution of 5-day mean sensible heat flux (SH)
along cross-section AA′ for the d120 domain. (b) Difference in the
time evolution of SH between the d240 and d120. (c) Difference in
the time evolution of SH between the d480 and d120. (d) Difference
in the time evolution of SH between the d960 and d120. For d120, d240 and
d480, the SH fluxes were spatially aggregated to 960 m resolution.
(a) Time evolution of 5-day mean sensible heat flux (LH)
along cross-section AA′ for the d120 domain. (b) Difference in the
time evolution of LH between the d240 and d120. (c) Difference in
the time evolution of LH between the d480 and d120. (d) Difference
in the time evolution of LH between the d960 and d120. For d120, d240 and
d480, the LH fluxes were spatially aggregated to 960 m resolution.
Summary and conclusions
This study was motivated by recent
efforts in including physically based hydrological models in earth system
models both for seasonal-scale and climate studies that would allow for
examination of the linkages between land–atmosphere and subsurface
hydrodynamics. The hydrological component of the newly developed TerrSysMP
was used over a sub-catchment of the Rur at grid resolutions encompassing
roughly the spatial scales between LES and mesoscale atmospheric models to
quantify the effect of grid resolution on simulated soil moisture, soil
temperature and surface fluxes.
The terrain analysis of the sub-catchment showed the expected smoothing of
slopes and filtering of the profile and plan curvatures with grid coarsening.
This grid resolution has a strong effect on the non-local controls of soil
moisture simulated by the model, while the local vegetation exerts a strong
modulation on the transfer of the grid-resolution-dependent soil moisture
variability to soil temperature and surface fluxes. In this study, soil
moisture beneath forests was found to decrease more than beneath crops due to
the location of forests over steeper slopes. However, due to the plant
physiological properties affecting the transmissivity of solar radiation,
crops lead to a higher grid resolution dependence than trees in terms of soil
temperature and surface fluxes. For crops, the magnitude of the LAI was also
found to have a strong effect on the scaling behavior of surface fluxes. This
non-linear scaling behavior of the energy balance with respect to grid
resolution can alter the spatial and temporal pattern of simulated surface
fluxes. Larger differences were especially observed when moving from d480 to
d960. These dependencies can induce or weaken mesoscale circulations and the
ensuing boundary layer evolution when using coupled simulations. Using an
idealized setup, such atmospheric feedback effects for a rainfall–runoff
process (runoff due to excess infiltration, creating wet and dry patches),
with varying grid resolution, were shown earlier by Shrestha et al. (2014).
Here the heterogeneity at different resolutions was introduced in terms of
topographic slopes while keeping a homogeneous land cover (crop). In this
study, the inclusion of convergent zones at finer grid resolutions compared
to coarser grid resolutions, where they are usually filtered out by
topography smoothing, enhanced the overland flow, thereby reducing the
infiltration and the mean soil moisture content while extending the downslope
moist area. This extended downslope moist area increased the extent of the
moist patch compared to the coarser run, thereby lowering the Bowen ratio
along the extended patch, and also increasing the extent of the downdraft
region in the atmospheric boundary layer. Thus, the dependence of the
simulated surface–subsurface physical processes on grid resolution
potentially also affects the local circulation and atmospheric boundary layer
evolution. However, fully coupled real data simulations including the
atmosphere at different grid resolutions are required to further improve our
understanding of the land–atmosphere feedbacks.
The study was limited to grid resolutions from 120 to 960 m. One
could argue that even the 120 m resolution is not sufficient for
hydrological models, and much finer resolutions (≤ 30 m) are
needed (e.g., Kuo et al., 1999). We acknowledge the limitation in this study,
but finer resolutions challenge currently available computation resources and
also the convergence of 3-D integrated surface groundwater models. We realize
that sub-grid-scale parameterizations along with resolutions of approximately
100–200 m would be sufficient for coupled simulations. The
interaction between surface and groundwater in ParFlow is simulated using a
2-D shallow overland flow equation as an upper boundary condition (Kollet and
Maxwell, 2006), instead of the commonly applied conductance concept.
Particularly for the surface–groundwater interaction, this upper boundary
condition and the Darcy flux in the horizontal direction are affected by the
spatial filtering of the terrain curvature in the coarsening of the grid
resolution. This results in an increase in infiltration, a reduction of
lateral flow and a shallower groundwater table. So, there is a need for a
scale-dependent subgrid parameterization for the upper boundary condition and
for the lateral flow in the groundwater model used in this study. For the
upper boundary condition, e.g., the concept of a fractional saturated area
(parameterized by the subgrid-scale distributions of topographic index and
groundwater table depth) is widely used to control surface runoff and hence
infiltration in many 1-D land surface models (e.g., Niu et al., 2005). For
the subsurface, e.g., Niedda (2004) proposed an amplification/upscaling of
hydraulic conductivity to compensate for the reduction in the hydraulic
gradients using the information content of terrain curvature to produce
similar lateral flow. In general, the topography heterogeneity analysis and
the soil moisture data presented in this study do provide the necessary data
to investigate such parameterizations for the upper boundary condition and
lateral subsurface flow. Future studies will involve developing such robust
scale-dependent subgrid parameterization for the 3-D physically based
groundwater model.
Acknowledgements
This study was conducted with support from SFB/TR32 (www.tr32.de),
“Patterns in Soil-Vegetation-Atmosphere Systems: Monitoring, Modelling, and
Data Assimilation”, funded by the Deutsche Forschungsgemeinschaft (DFG). We
thank the Deutscher Wetterdienst (DWD) for the COSMO analysis data. The data
analysis including the pre-processing and post-processing of input data for
TerrSysMP was done using the NCAR Command language (version 6.2.0).
Edited by: I. Neuweiler
ReferencesAnyah, R. O., Weaver, C. P., Miguez-Macho, G., Fan, Y., and Robock, A.:
Incorporating water table dynamics in climate modeling: 3. Simulated
groundwater influence on coupled land–atmosphere variability, J. Geophys.
Res., 113, D07103, 10.1029/2007JD009087, 2008.
Ashby, S. F. and Falgout, R. D.: A parallel multigrid preconditioned
conjugate gradient algorithm for groundwater flow simulations, Nucl. Sci.
Eng., 124, 145–159, 1996.
Avissar, R. and Schmidt, T.: An evaluation of the scale at which
ground-surface heat flux patchiness affects the convective boundary layer
using large-eddy simulations, J. Atmos. Sci., 55, 2666–2689, 1998.Baidya Roy, S. and Avissar, R.: Impact of land use/land cover change on
regional hydrometeorology in Amazonia, J. Geophys. Res., 107,
LBA 4-1–LBA 4-12, 10.1029/2000JD000266, 2002.Baldauf, M., Seifert, A., Förstner, J., Majewski, D., Raschendorfer, M.,
and Reinhardt, T.: Operational convective-scale numerical weather prediction
with the COSMO model: description and sensitivities, Mon. Weather Rev., 139,
3887–3905, 10.1175/MWR-D-10-05013.1, 2011.
Baldocchi, D. D., Vogel, C. A., and Hall, B.: Seasonal variation of carbon
dioxide exchange rates above and below a boreal jack pine forest, Agr. Forest
Meteorol., 83, 147–170, 1997.
Becker, A. and Braun, P.: Disaggregation, aggregation and spatial scaling in
hydrological modelling, J. Hydrol., 217, 239–252, 1999.Beven, K. J. and Cloke, H. L.: Comment on “Hyperresolution global land
surface modeling: meeting a grand challenge for monitoring Earth's
terrestrial water” by Eric F. Wood et al., Water Resour. Res., 48, W01801,
10.1029/2011WR010982, 2012.
Blöschl, G. and Sivapalan, M.: Scale issues in hydrological modelling: a
review, Hydrol. Process., 9, 251–290, 1995.Bormann, H.: Impact of spatial data resolution on simulated catchment water
balances and model performance of the multi-scale TOPLATS model, Hydrol.
Earth Syst. Sci., 10, 165–179, 10.5194/hess-10-165-2006, 2006.
Butts, M., Drews, M., Larsen, M. A., Lerer, S., Rasmussen, S. H., Grooss, J.,
Overgaard, J., Refsgaard, J. C., Christensen, O. B., and Christensen, J. H.:
Embedding complex hydrology in the regional climate system-dynamic coupling
across different modelling domains, Adv. Water Resour., 74, 166–184, 2014.
Cai, X. and Wang, D.: Spatial autocorrelation of topographic index in
catchments, J. Hydrol., 328, 581–591, 2006.
Dixon, B. and Earls, J.: Resample or not?! Effects of resolution of DEMs in
watershed modeling, Hydrol. Process., 23, 1714–1724, 2009.Gasper, F., Goergen, K., Shrestha, P., Sulis, M., Rihani, J., Geimer, M., and
Kollet, S.: Implementation and scaling of the fully coupled Terrestrial
Systems Modeling Platform (TerrSysMP v1.0) in a massively parallel
supercomputing environment – a case study on JUQUEEN (IBM Blue Gene/Q),
Geosci. Model Dev., 7, 2531–2543, 10.5194/gmd-7-2531-2014, 2014.Giertz, S., Diekkrüger, B., and Steup, G.: Physically-based modelling of
hydrological processes in a tropical headwater catchment (West Africa) –
process representation and multi-criteria validation, Hydrol. Earth Syst.
Sci., 10, 829–847, 10.5194/hess-10-829-2006, 2006.
Grayson, R. B., Western, A. W., Chiew, F. H., and Blöschl, G.: Preferred
states in spatial soil moisture patterns: local and nonlocal controls, Water
Resour. Res., 33, 2897–2908, 1997.Grossman, R. L., Yates, D., LeMone, M. A., Wesely, M. L., and Song, J.:
Observed effects of horizontal radiative surface temperature variations on
the atmosphere over a Midwest watershed during CASES 97, J. Geophys. Res.,
110, D06117, 10.1029/2004JD004542, 2005.
Herbst, M., Diekkrueger, B., and Vanderborght, J.: Numerical experiments on
the sensitivity of runoff generation to the spatial variation of soil
hydraulic properties, J. Hydrol., 326, 43–58, 2006.
Jones, J. E. and Woodward, C. S.: Newton-krylov-multigrid solvers for
large-scale, highly heterogeneous, variably saturated flow problems, Adv.
Water Resour., 24, 763–774, 2001.
Kollet, S. J. and Maxwell, R. M.: Integrated surface–groundwater flow
modeling: a free-surface overland flow boundary condition in a parallel
groundwater flow model, Adv. Water Ressour., 29, 945–958, 2006.
Kuo, W. L., Steenhuis, T. S., McCulloch, C. E., Mohler, C. L., Weinstein,
D. A., DeGloria, S. D., and Swaney, D. P.: Effect of grid size on runoff and
soil moisture for a variable-source-area hydrology model, Water Resour. Res.,
35, 3419–3428, 1999.Larsen, M. A. D., Refsgaard, J. C., Drews, M., Butts, M. B., Jensen, K. H.,
Christensen, J. H., and Christensen, O. B.: Results from a full coupling of
the HIRHAM regional climate model and the MIKE SHE hydrological model for a
Danish catchment, Hydrol. Earth Syst. Sci., 18, 4733–4749,
10.5194/hess-18-4733-2014, 2014.
LeMone, M. A., Grossman, R. L., Mcmillen, R. T., Liou, K. N., Ou, S. C.,
Mckeen, S., Angevine, W., Ikeda, K., and Chen, F.: CASES-97: late-morning
warming and moistening of the convective boundary layer over the Walnut River
watershed, Bound.-Lay. Meteorol., 104, 1–52, 2002.
Maxwell, R. M.: A terrain-following grid transform and preconditioner for
parallel, large-scale, integrated hydrologic modeling, Adv. Water Resour.,
53, 109–117, 2013.Maxwell, R. M., Lundquist, J. K., Mirocha, J. D., Smith, S. G., Woodward,
C. S., and Tompson, A. F.: Development of a Coupled Groundwater–Atmosphere
Model, Mon. Weather Rev., 139, 96–116, 10.1175/2010MWR3392.1, 2011.Niedda, M.: Upscaling hydraulic conductivity by means of entropy of terrain
curvature representation, Water Resour. Res., 40, W04206,
10.1029/2003WR002721, 2004.Niu, G. Y., Yang, Z. L., Dickinson, R. E., and Gulden, L. E.: A simple
TOPMODEL-based runoff parameterization (SIMTOP) for use in global climate
models, J. Geophys. Res., 110, D21106, 10.1029/2005JD006111, 2005.Oleson, K. W., Niu, G. Y., Yang, Z. L., Lawrence, D. M., Thornton, P. E.,
Lawrence, P. J., Stöckli, R., Dickinson, R. E., Bonan, G. B., Levis, S.,
Dai, A., and Qian, T.: Improvements to the Community Land Model and
theirimpact on the hydrological cycle, J. Geophys. Res., 113, G01021,
10.1029/2007JG000563, 2008.Quinn, P. F., Beven, K. J., and Lamb, R.: The in (a/tan/β) index: how
to calculate it and how to use it within the topmodel framework, Hydrol.
Process., 9, 161–182, 1995.
Sciuto, G. and Diekkrueger, B.: Influence of soil heterogeneity and spatial
discretization on catchment water balance modeling, Vadose Zone J., 9,
955–969, 2010.
Shrestha, P., Sulis, M., Masbou, M., Kollet, S., and Simmer, C.:
A scale-consistent terrestrial systems modeling platform based on COSMO, CLM,
and Parflow, Mon. Weather Rev., 142, 3466–3483, 2014.Simmer, C., Thiele-Eich, I., Masbou, M., Amelung, W., Crewell, S.,
Diekkrueger, B., Ewert, F., Hendricks Franssen, H.-J., Huisman, A. J., Kemna,
A., Klitzsch, S., Kollet, N., Langensiepen, M., Loehnert, U., Rahman, M.,
Rascher, U., Schneider, K., Schween, J., Shao, Y., Shrestha, P., Stiebler,
M., Sulis, M., Vanderborght, J., Vereecken, H., van der Kruk, J., Zerenner,
T., and Waldhoff, G.: Monitoring and modeling the terrestrial system from
pores to catchments – the transregional collaborative research center on
Patterns in the soil-vegetation-atmosphere system, B. Am. Meteorol. Soc.,
10.1175/BAMS-D-13-00134.1, online first, 2014.
Sulis, M., Paniconi, C., and Camporese, M.: Impact of grid resolution on the
integrated and distributed response of a coupled surface–subsurface
hydrological model for the des Anglais catchment, Quebec, Hydrol. Process.,
25, 1853–1865, 2011.Sulis, M., Langensiepen, M., Shrestha, P., Schickling, A., Simmer, C., and
Kollet, S. J.: Evaluating the influence of plant-specific physiological
parameterizations on the partitioning of land surface energy fluxes,
J. Hydrometeorol., 16, 517–533, 10.1175/JHM-D-14-0153.1, 2015.The NCAR Command Language: Version 6.1.2 (Software), UCAR/NCAR/CISL/VETS,
Boulder, Colorado, 10.5065/D6WD3XH5, 2013.Valcke, S.: The OASIS3 coupler: a European climate modelling community
software, Geosci. Model Dev., 6, 373–388, 10.5194/gmd-6-373-2013,
2013.
Vereecken, H., Kollet, S., and Simmer, C.: Patterns in
soil-vegetation-atmosphere systems: monitoring, modeling, and data
assimilation, Vadose Zone J., 9, 821–827, 10.2136/vzj2010.0122, 2010.
Vivoni, E. R., Ivanov, V. Y., Bras, R. L., and Entekhabi, D.: On the effects
of triangulated terrain resolution on distributed hydrologic model response,
Hydrol. Process., 19, 2101–2122, 2005.Waldhoff, G.: Enhanced Land Use Classification of 2009 for the Rur catchment,
CRC/TR32 Database (TR32DB), 10.5880/TR32DB.2, 2012.Wood, E. F., Roundy, J. K., Troy, T. J., van Beek, L. P. H., Bierkens,
M. F. P., Blyth, E., de Roo, A., Döll, P., Ek, M., Famiglietti, J.,
Gochis, D., van de Giesen, N., Houser, P., Jaffé, P. R., Kollet, S.,
Lehner, B., Lettenmaier, D. P., Peters-Lidard, C., Sivapalan, M., Sheffield,
J., Wade, A., and Whitehead, P.: Hyperresolution global land surface
modeling: meeting a grand challenge for monitoring Earth's terrestrial water,
Water Resour. Res., 47, W05301, 10.1029/2010WR010090, 2011.
Zhang, W. and Montgomery, D. R.: Digital elevation model grid size, landscape
representation, and hydrologic simulations, Water Resour. Res., 30,
1019–1028, 1994.