Clemens Simmer, Meteorological Institute, University Bonn, Auf dem Hügel 20, D-. 36 ...... Schomburg, A., V. Venema, R. Lindau, F. Ament, and C. Simmer, 2012:.
1
Monitoring and Modeling the Terrestrial System from Pores to Catchments – the
2
Transregional Collaborative Research Center on Patterns in the Soil-Vegetation-
3
Atmosphere System
4 5
by
6
Clemens Simmer, Matthieu Masbou, Insa Thiele-Eich, Wulf Amelung, Bernhard
7
Blümich, Georg Bareth, Heye Bogena, Andreas Bott, Carsten Burstedde, Christoph
8
Clauser, Susanne Crewell, Bernd Diekkrüger, Hendrik Elbern, Frank Ewert, Peter
9
Fiener, Harrie-Jan Hendricks Franssen, Petra Friederichs, Alexander Graf, Michael
10
Griebel, Michael Herbst, Alexander J. Huisman, Andreas Kemna, Norbert Klitzsch,
11
Stefan Kollet, Manfred Krafczyk, Ulrich Lang, Matthias Langensiepen, Ulrich Löhnert,
12
Andreas Lücke, Oliver Mohnke, Andreas Pohlmeier, A. S. M. Mostaquimur Rahman,
13
Uwe Rascher, Karl Schneider, Jan Schween, Yaping Shao, Prabhakar Shrestha,
14
Maik Stiebler, Mauro Sulis, Jan Vanderborght, Harry Vereecken, Jan van der Kruk,
15
Björn Waske, Lutz Weihermüller, Mark Van Wijk, Andreas Wahner, Gerd Welp, and
16
Tanja Zerenner
17 18
Capsule Summary:
19
Observing and modeling the water and energy flow from soil pores to clouds and
20
from the groundwater to the atmosphere via a strong interdisciplinary effort
21 22
Affiliations:
23
Amelung, Bott, Burstedde, Diekkrüger, Ewert, Friederichs, Griebel, Kemna,
24
Langensiepen, Masbou, Rahman, Shrestha, Sulis, Simmer, Waske, Welp, Van Wijk,
25
Zerenner - University Bonn; Bareth, Crewell, Elbern, Fiener, Lang, Löhnert,
26
Schneider, Schween, Shao - University Cologne; Blümich, Clauser, Klitzsch - RWTH 1
27
Aachen University; Bogena, Hendricks Franssen, Graf, Herbst, Huisman, Kollet,
28
Lücke, Pohlmeier, Rascher, Vanderborght, Van der Kruk, Van Wijk, Vereecken,
29
Weihermüller, Wahner – Forschungszentrum Jülich GmbH; Hendricks Franssen,
30
Kollet - Centre for High-Performance Scientific Computing in Terrestrial Systems,
31
Masbou – now at Deutscher Wetterdienst, Offenbach; HPSC TerrSys, Krafczyk,
32
Stiebler – Technical University Braunschweig, Mohnke – now Baker Huges, Celle,
33
Germany
34 35
Corresponding author:
36
Clemens Simmer, Meteorological Institute, University Bonn, Auf dem Hügel 20, D-
37
53121 Bonn, Germany
38
2
39
Abstract:
40
By far the most activities of mankind take place in the transition zone connecting
41
groundwater, soil, vegetation and atmosphere. Mass, momentum, and heat energy
42
fluxes within and between the neighboring compartments drive their mutual state
43
evolution. Improved understanding of the processes that drive these fluxes is
44
important for climate projections and weather prediction, flood forecasting, water and
45
soil resources management, agriculture, and water quality control. The vastly
46
different flow behavior within the different compartments leads to complex patterns
47
on different time and spatial scales, which make quantitative predictions of the
48
terrestrial system behavior a major challenge to both scientists and policymakers. In
49
2007 the Transregional Collaborative Research Centre No. 32 (TR32) set out to
50
investigate the groundwater-soil-vegetation-atmosphere continuum by integrating
51
monitoring of system parameters, states and fluxes with modeling and data
52
assimilation in order to reach a holistic view of the terrestrial system. The TR32
53
belongs to a class of long-term research programs that are funded by the German
54
national science foundation (Deutsche Forschungsgemeinschaft, DFG) in order to
55
concentrate and integrate research activities of several universities on an emerging
56
scientific topic of high societal relevance. With the aim to bridge the gap between
57
micro-scale soil pores and catchment-scale atmospheric variables, the TR32 unites
58
research groups from within the Geo Alliance (Geoverbund) ABC/J, namely the
59
universities RWTH Aachen, Bonn, and Cologne, as well as the environmental and
60
geoscience departments of the Forschungszentrum Jülich. Here we report about
61
recent achievements in monitoring and modeling the terrestrial system including the
62
development of new observation techniques for the subsurface, the establishment of
63
cross-scale, multi-compartment modeling platforms from the pore to the catchment
3
64
scale, and the use of these new methods and platforms to investigate the
65
propagation of subsurface patterns to the atmospheric boundary layer.
66 67
4
68
1 Introduction
69 70
State variables and parameters of the terrestrial system, which encompasses
71
groundwater, soil, vegetation, and the atmosphere, exhibit complex patterns on a
72
wide range of temporal and spatial scales that extends from seconds to years and
73
from the soil pore scale to the global scale. These patterns are also reflected in the
74
heterogeneous inter- and intra-compartmental fluxes of heat energy, water, carbon,
75
nitrogen, and momentum. Most patterns of terrestrial state variables can be traced to
76
the control of the relatively static and often extremely heterogeneous soil and sub-soil
77
parameters, which are usually generated on geological time scales. The fast-moving
78
and mixing, almost parameter-free atmosphere drives – and is driven by - these
79
fluxes and further adds considerable heterogeneity via its own internal scales of
80
motion ranging from turbulence over convection to synoptic systems. This again adds
81
considerable variability to the driving radiative fluxes due to clouds and water vapor
82
and to water fluxes via precipitation. Vegetation acts to all this as a living
83
transmission medium, which is more than any other compartment subject to human
84
interaction by agricultural and forest management and land use change but still
85
inherits geomorphological patterns from the geological past (see e.g. Figure 2 (EMI
86
results)). This prominent heterogeneous nature of terrestrial systems constitutes a
87
major challenge to monitoring and predicting their state. Improving our understanding
88
and prediction capabilities of the terrestrial system therefore requires measurement
89
techniques that allow us to characterize and monitor the spatio-temporal evolution of
90
system properties across scales, terrestrial system model platforms that include all
91
relevant processes, and state variable assimilation and parameter estimation
92
methods. The Collaborative Research Center (CRC) TR32 on “Patterns in the Soil-
93
Vegetation-Atmosphere System – Monitoring, Modeling, and Data Assimilation” 5
94
began working on these goals in 2007 including a graduate school for the PhD
95
students working within the CRC (incorporating external PhD projects on related
96
topics). A data management component (Curdt et al., 2013) stores and secures all
97
observations, analyses, and documents for at least a decade.
98 99
1.1 German Collaborative Research Centers
100 101
CRCs are established at German universities and co-funded by the German national
102
science foundation (Deutsche Forschungsgemeinschaft, DFG) for a period of up to
103
12 years subdivided into three phases. Each CRC phase is granted based on
104
applications that need to be positively reviewed by international review panels and
105
involves a highly competitive evaluation procedure that assesses and compares
106
proposed CRCs from all science disciplines including humanities. CRCs enable
107
researchers to pursue an outstanding research program with a long-term perspective
108
while crossing the boundaries of disciplines, departments, institutes, and faculties.
109
The CRC program in turn contributes towards defining and sharpening the profiles of
110
participating universities both with respect to research as well as teaching. CRCs
111
may incorporate projects at neighboring universities or non-university research
112
institutions and collaborations with industry and business. While CRCs are usually
113
applied for by one university, the TR32 is a transregional CRC within which the three
114
universities Bonn (leading), Cologne and RWTH Aachen and the Forschungszentrum
115
Jülich GmbH of the Helmholtz Association join their efforts to better understand the
116
origin and propagation of patterns in the terrestrial system and their value for model-
117
based predictions.
118 119
1.2 Patterns in System Parameters and State Variables 6
120 121
The TR32 focuses on the role of patterns in the terrestrial system state variables and
122
parameters for observations and modeling. In terrestrial system models - as
123
developed and applied in the TR32 - system parameters like the hydraulic
124
conductivity of the soil or the turbulent diffusion coefficients of the atmospheric
125
boundary layer (ABL) control the flow of energy, matter and momentum within and
126
between compartments on timescales for which predictability of the system variables
127
is aspired. On these time scales, system parameters of the soil are usually assumed
128
to be constant, or in case of the ABL, predictable from the state variables. The
129
distinction between system parameters and state variables is, however, a concept
130
born out of the model perspective: parameters refer to parameterizations of
131
processes, which evade direct numerical simulation based on the primitive equations
132
for the conservation of mass, energy and momentum due to the required but mostly
133
computationally unattainable resolution. Both in the soil and in the atmosphere,
134
homogeneity
135
parameterization concepts such as e.g. the use of hydraulic conductivity in the
136
Richards equation or the Monin-Obukhov similarity of the ABL in Reynolds-Averaged
137
Numerical Simulation (RANS) models. All parameters do actually vary - not only with
138
the grid spacing of the models due to e.g. the natural variability of soils and the non-
139
linearity of the processes, but also with time because they depend on the system
140
state variables, their subgrid variability and possibly also their history (hysteresis
141
effects). In meteorology, where such problems are more clear-cut, the endeavor of
142
parameter prediction is related to the closure problem. Thus in general, there is no
143
clear cut between system parameters and state variables, a problem increasingly
144
recognized also in data assimilation.
assumptions,
which
are
145 7
never
fulfilled,
are
used
to
justify
146
The TR32 employs the pattern paradigm as an overarching concept to address the
147
ubiquitous up- and downscaling issues in monitoring, modeling and data assimilation.
148
More detailed information on this topic can be found in Vereecken et al. (2010) and a
149
special issue in preparation for Water Resources Research. In the geosciences,
150
patterns may be understood as repetitions of similar structures in system state
151
variables and system parameters in space and time, with structures denoting
152
identifiable objects. In the case of parameters such objects may be a single soil pore,
153
a plant root, an individual plant, a plough mark, soil heterogeneity associated with a
154
paleo-river system, a crop-cultivated field, a single hill, or an entire valley. Such
155
objects can be rather static relative to the timescales on which fluxes occur within or
156
between compartments. However, they can also be more dynamic as in the case of
157
soil temperature and moisture in response to insolation or precipitation, the exchange
158
of fluxes themselves, and any structure in the atmosphere ranging from single
159
eddies, updrafts in the ABL and cumulus clouds to thunderstorms or cyclones. Most if
160
not all of these structures and resulting patterns have characteristic scales on which
161
they can be detected, but often disappear when changing the observation scale,
162
such as e.g. cumulus clouds or cultivated fields whose characteristic patchy or
163
rectangular shape, respectively, completely disappear when spatial resolution
164
exceeds the 100 m scale. Thus the chaos approach to patterns with clear scale-
165
variance relations usually does not apply. Studying how patterns influence fluxes and
166
state variables across scales is a key goal of the TR32.
167 168
The most appropriate way for modeling the terrestrial system requires the treatment
169
of the soil-vegetation-atmosphere system as a continuum with spatial resolutions that
170
allow the simulation of all relevant flow processes by the Navier-Stokes equations
171
using Direct Numerical Simulation (DNS) models down to the sub-pore scale, and 8
172
that thus reduce remaining exchanges at compartment boundaries to diffusive
173
processes. This also takes into account small-scale processes such as the
174
movement of plants in the turbulent airflow and water motion in pores. Patterns on
175
this scale need not be considered in this approach as they turn up automatically as a
176
natural consequence of the acting processes. While the direct approach is in
177
principle possible, applications on scales required to predict terrestrial systems on
178
the catchment scale are and will be prohibitive due to restricted computational
179
capabilities for a very long time. Instead continuum scale modelling approaches are
180
used, such as the Richards equation for flow of water in soils that captures micro-
181
scale pore geometry in two macroscopic material properties, the water retention and
182
hydraulic conductivity functions. Pore-scale models are used in TR32 to understand
183
and derive these material properties in order to parameterize Richards equation. The
184
atmosphere is usually modeled using RANS (Reynolds-Averaged Navier-Stokes)
185
models, which require especially strong assumptions near boundaries. This is usually
186
tackled with the Monin-Obukhow similarity assumption. As a consequence, the
187
exchange between system compartments is parameterized by diffusion-like
188
processes with coefficients and material properties estimated from experiments
189
and/or statistics of the larger-scale flows. This is where patterns become important:
190
since exchange processes between the compartments are driven by local gradients,
191
any correlation between the patterns of the driving system state variables at the
192
boundaries of the neighboring compartments (e.g. temperature at the surface and the
193
lowest atmospheric model layer) will directly impact the fluxes and thus the system
194
state. This also means that patterns are an inevitable element of any upscaling and
195
downscaling concept applied in terrestrial system modeling, a quest taken up by the
196
TR32 and illustrated in the remainder of this paper.
197 9
198
2 Monitoring of the Rur Catchment
199 200
The development of techniques to map and understand patterns and to use this to
201
model and predict the terrestrial system requires a real counterpart for analysis and
202
testing. Due to its proximity to the cooperating institutes, TR32 identified the Rur
203
catchment at the western border of Germany to Belgium and the Netherlands (Figure
204
1 (Rur Catchment Monitoring)) as its central observation site. The Rur catchment has
205
been heavily instrumented in strong cooperation with the TERENO program of the
206
Helmholtz Association (http://teodoor.icg.kfa-juelich.de), which was developed in
207
parallel to TR32 as a network of terrestrial observatories distributed over Germany
208
(Zacharias et al., 2011). Precipitation as the main atmospheric driver for soil moisture
209
patterns is monitored in the Rur catchment in five minute intervals by the twin dual-
210
polarized X-band Doppler radars BoXPol in Bonn and JuXPol on the Sophienhöhe, a
211
hill created from open-cast mining and jutting roughly 200 m out of the surrounding
212
terrain, and at one minute resolution by a third reflectivity-only X-Band radar
213
(Rainscanner) in the southwestern corner of the catchment. The area is also covered
214
by four C-Band radars of the German national weather service (Deutscher
215
Wetterdienst, DWD), which have been recently upgraded to polarimetry. Precipitation
216
patterns derived from sets of scanning radars are usually plagued with artifacts
217
imposed by the observation method via beam-clutter, beam blocking, and
218
attenuation. Such effects have been minimized by the so-called R(A) methodology
219
developed by Ryzhkov et al. (2014). Detailed knowledge on the atmospheric state is
220
available in near-real time from the Jülich Observatory for Cloud Evolution (JOYCE,
221
http://www.geomet.uni-koeln.de/allgemein/forschung/joyce/), which is equipped with
222
a unique array of state-of-the-art active and passive remote sensing and in-situ
223
instruments. The continuous and temporally highly-resolved measurements focus on 10
224
the atmospheric boundary layer and allow the characterization of the diurnal cycle of
225
turbulence, water vapor, stability and cloudiness (Schween et al., 2011). A specific
226
feature of this atmospheric observatory is the use of scanning measurements that
227
aim to capture atmospheric patterns and their relation to the land surface (see
228
supplement). Fluxes of sensible heat, evapotranspiration, CO2, and momentum are
229
monitored by five fixed Eddy Covariance (EC) stations and by one roving EC-station.
230
Two sub-catchments (forest and grassland) are equipped with wireless extensive soil
231
moisture networks with hundreds of sensors in various depths (Bogena et al., 2010),
232
which allow for the examination of seasonal and event-scale spatial soil moisture
233
dynamics, the validation of models (Cornelisson et al., 2014, see also modeling
234
section) and remote sensing data (e.g. Hasan et al., 2014). We found clockwise
235
hysteretic soil moisture dynamics at the event-scale during intense precipitation
236
events that rapidly wetted the topsoil (Rosenbaum et al., 2012). Cosmic-ray soil
237
moisture probes (Zreda et al., 2008) supplement the installed soil moisture networks
238
(Bogena et al., 2013) and allow the characterization of temporal soil moisture
239
dynamics over the entire catchment (Baatz et al., 2014). At few specific sites, a
240
whole suite of geophysical methods like Nuclear Magnetic Resonance (NMR, e.g.
241
Perlo et al., 2013), Spectral Induced Polarization (e.g. Kemna et al., 2013),
242
Electromagnetic Induction (EMI, e.g. Mester et al., 2011), and Ground-Penetrating
243
Radar (e.g. Busch et al., 2013) are employed to develop and test methods for
244
probing the structure and composition of the subsurface. Recent results from the
245
inversion of Electro-Magnetic Induction observations (Figure 2 (EMI results), see also
246
von Hebel et al., 2014 and Rudolph et al., in preparation) clearly show the linkage
247
between sub-soil patterns in terms of sand, silt and gravel originating from paleo-
248
rivers with both the EMI information and the visual impact on vegetation and Leaf
249
Area Index after an extended drought period. Methods known from medicine such as 11
250
Magnetic Resonance Imaging have first been developed in the laboratory and are
251
now being applied in the field to measure soil water content with high vertical
252
resolution near the soil surface. Observations also address ecosystem exchange
253
processes on various scales for agricultural surfaces (e.g. Langensiepen et al.,
254
2012). The role of soil patterns for soil carbon pools (Bornemann et al., 2010, 2011)
255
and soil heterotrophic respiration is analyzed using Mid-Infrared Spectroscopy
256
(MIRS) and geostatistical modeling (Herbst et al., 2012), respectively. In several
257
intensive measurement campaigns, aircraft-borne instruments are used to measure
258
the spatio-temporal structure of the ABL including the CO2 concentration in order to
259
analyze the relation between surface flux patterns and corresponding patterns in the
260
atmosphere. Vegetation, soil moisture and land use over the whole catchment is
261
monitored via aircraft and satellites (Hoffmeister et al., 2012; Koyama et al., 2010). In
262
addition, vegetation assimilation activity states are observed using ground-based
263
fluorescent-related techniques and airborne instrumentation (Figure 3 (Sun-Induced
264
Fluourescence)) that is a prototype for future space-based mission (Damm et al.,
265
2011). The observations document the large between and within field variability of
266
plant photosynthetic activity, and may in future be used for directly quantifying plant
267
transpiration. Patterns have also been investigated within the soil moisture
268
monitoring activities with a particular focus on the scaling properties for agricultural
269
land (Korres et al., 2013). The dependence of evapotranspiration on soil states, soil
270
and root properties, and meteorological conditions is analyzed using sap-flow
271
devices installed in trees and crops.
272 273
3 Modeling the Terrestrial System on Different Scales
274 275
The TR32 employs a cross-scale, multi-compartment modeling approach to upscale 12
276
the water, energy, and CO2 fluxes in the terrestrial system from the local to the
277
catchment scale based on a numerical approach combined with stochastic
278
techniques. The analysis of the simulations with grids that honor the respective
279
scales reveals the role of patterns on the fluxes in the system and helps to design a
280
general upscaling framework that quantifies information transfer between scales due
281
to non-linear interactions in the system. Accordingly, the TR32 uses and develops
282
physics-based process models and model platforms for all relevant scales. For
283
example, pore-scale models using Lattice Boltzmann methods have been used for
284
jointly simulating multi-phase flow and NMR relaxation behavior (Figure 4 (Lattice
285
Boltzmann Simulations)) in order to improve the interpretation of NMR observations
286
in terms of water retention and hydraulic conductivity properties that are essential for
287
describing water flow in continuum models (Mohnke and Klitzsch, 2010). On a
288
somewhat larger scale, a soil-root model is developed in order to derive improved
289
parameterizations for root density and distribution in the soil, as well as their effect on
290
the soil moisture profiles and respiration as a function of larger-scale soil state
291
variables and parameters (see Box Soil-Root Model). Large-Eddy-Simulation (LES)
292
models (see below) with highly resolved land surface and subsurface models for
293
energy fluxes, water flow, and carbon dynamics are used for resolving patterns and
294
heterogeneities on numerical grids of the order of centimeters (vertical) to decimeters
295
(horizontal) for land surface and subsurface, while the atmosphere from the LES is
296
assumed homogeneous on this scale. Patterns on scales of tens to several hundred
297
meters, which are mainly characterized by land use patterns and subsurface
298
structures generated e.g. from geomorphologic processes like erosion or river
299
evolution are addressed by simulations with LES-ALM (Shao et al., 2012) derived
300
from the WRF-NOAH modeling platform and the Terrestrial System Modeling
301
Platform (TerrSysMP, see Box 2 TerrSysMP) designed for the regional climate scale 13
302
with resolutions from about 100 meters to kilometers, but also by 3D hydrological
303
models like HydroGeoSphere (HGS, see e.g. Cornelissen et al., 2014).
304 305
Compared to Parflow in TerrSysMP, which requires a rectangular grid definition, the
306
finite-element model HGS is particularly useful for small-scale soil, land use, and
307
topography applications. The comparison of the mean simulated and observed soil
308
moisture dynamic of a 27 hectares headwater sub-catchment of the Rur for the
309
period 2010/2011 (Figure 5 (HGS Dynamics)) shows the ability of HGS to capture
310
long-term dynamics in a reasonable manner, although this model failed to reproduce
311
short-term dynamics probably due to a missing preferential flow component. The
312
general spatial patterns between simulations and observations are similar (Figure 6
313
(HGS Soil Patterns)) and are determined by the complex interactions between soil,
314
topography, and vegetation.
315 316
With LES-ALM we investigate the propagation of land surface heterogeneity in the
317
atmospheric boundary layer. Here we present example results for the natural land
318
surface at the Selhausen-Merken field site (for model settings see Shao et al.
319
(2013)). The evolutions in space and time of the patterns of e.g. sensible and latent
320
heat fluxes, temperature, humidity and turbulent kinetic energy are analyzed using
321
wavelet decomposition and averaging over multiple time scales. Figure 7 (LES
322
Pattern Propagation) shows the variation of sensible heat flux (H) patterns with
323
height in a convective boundary layer with an inversion at about 1.6 km above
324
ground. H patterns and wavelet energy spectra substantially differ on different height
325
levels and for different averaging times. Close to the surface the H patterns (and
326
those of other quantities not shown) bear great similarity to the land use pattern even
327
without time averaging, but time averaging enhances the similarity. For large heights 14
328
the resemblance decreases for small time intervals as turbulent patterns emerge, as
329
the hexagonal cells at 32 m. With increasing averaging times, however, the land
330
surface flux pattern clearly re-emerge with large-scale features of land surface
331
pattern persisting over a considerable depth of the ABL with enhanced energy
332
according to the wavelets spectra for the 30 minutes averaged H patterns. The
333
persistency of a land surface pattern in the atmospheric boundary layer depends on
334
its scale and strength, and on the capacity of turbulence to diffuse the land surface
335
signals. Close to the surface with low eddy diffusivity the land surface dominates the
336
pattern of atmospheric quantities. At higher levels turbulence rapidly diffuses the land
337
surface footprints by its own patterns governed by domain averaged buoyancy flux
338
and inversion height. Convective eddies have typical time scales τt of 103s and
339
spatial scales lt of 103m. A land surface pattern has in general multiple scales but has
340
often a dominant scale ls. Given an averaging time of atmospheric quantities of τa for
341
a given height, land surface heterogeneity persists if ls >> lt, and the land surface
342
pattern is most visible if τa >> τt.
343 344
Results on pattern interactions emerging from this effort eventually feed into
345
TerrSysMP (see Box 1), which includes a downscaling scheme for predicting near-
346
surface atmospheric variables at the scale of its higher-resolved landsurface and
347
subsurface scheme (Schomburg et al., 2010, 2012). This scheme is currently
348
advanced to achieve a better reconstruction of the spatiotemporal patterns at the fine
349
scale. Figure 8 (Atmospheric Downscaling)) illustrates the performance of a new
350
algorithm based on multi-objective Genetic Programming (GP). This machine
351
learning method allows for physical consistency checks (in contrast to alternative
352
methods that rely on the output of artificial neural networks) and offers the possibility
353
to quantify the quality of a downscaling rule based on several aspects, such as 15
354
spatial structure, spatially distributed variance and spatio-temporal correlation of the
355
fields. The regional climate length scale corresponds to the smallest grid size that
356
can be computed using TerrSysMP, with the Monin-Obukhov similarity theory for the
357
ABL currently being the limiting approximation. Patterns resolved include differences
358
in land use (bare soil/cropped soil and differences in crops), soil type, groundwater
359
table depth, and topography.
360 361
TerrSysMP simulations with increasing spatial resolution and thus grid sizes over the
362
whole Rur area include calculations of both water and carbon dynamics from the
363
subsurface into the atmosphere. The goal is to identify the major spatial and temporal
364
scales at which two-way feedbacks from the subsurface and the atmosphere impact
365
water and energy states and fluxes at the land surface and vice versa. Using, for
366
example, time localized wavelet spectra (Figure 9 (WTD-LE Cross-Wavelets)),
367
Rahman et al. (2014) were able to show that net radiation (Rnet) induces physically
368
intuitive diurnal variability in evapotranspiration (ET), which results in a cyclic
369
pumping effect depleting groundwater storage under dry conditions. In turn,
370
groundwater depletion and associated lowering of the shallow groundwater table
371
leads to a negative feedback on ET at the monthly time scale, which increases to
372
multi-months in case of extended dry periods. Applying similar concepts in the space
373
domain, the simulation results suggest that in summer, the structure in ET is mainly
374
determined by the spatial water table configuration, while ET can be predicted from
375
the spatial structure of Rnet during cooler months. This hierarchy and interactions of
376
space and time scales is used to derive a theoretical framework to upscale fluxes
377
and to account for the role of patterns, which may include e.g. a coarse graining
378
approach based on wavelets and information theory.
379 16
380
4 Data Assimilation
381 382
Aside from appropriate boundary conditions, the prediction of the state of the
383
terrestrial system requires knowledge concerning the initial state. Data assimilation,
384
i.e. the convolution of observations with a given model state, is the method of choice
385
for this endeavor. In contrast to atmospheric models, terrestrial system models are
386
faced with the additional problem of a priori unknown parameters of the surface, soil
387
and subsoil, which vary strongly in space and even time due to violations of
388
homogeneity assumptions in the models or even physical hysteresis effects. Thus,
389
parameters need to be included into the data assimilation process. Most
390
observations of the terrestrial system are rather indirect (any remote sensing
391
technique) and depend on additional characteristics of the system, which are often
392
ignored by the terrestrial models because of their minor role for the evolution of the
393
system state variables (e.g. the dependence of the measured neutron intensity by the
394
cosmic ray probe on litter layers). Accordingly, as many observations as possible
395
should be utilized in data assimilation in order to constrain the very large degree of
396
freedom.
397 398
Currently, the TR32 uses the Local Ensemble Transform Kalman Filter (LETKF)
399
(Hunt et al., 2007) for updating both model state variables and parameters of the land
400
surface model in a column-based approach. The implementation allows assimilating
401
brightness temperature measured by satellites with the CMEM-operator (de Rosnay
402
et al., 2011), land surface temperature with a dual source operator (Kustas and
403
Anderson, 2009) and neutron counts with the COSMIC-operator (Shuttleworth et al.,
404
2013), besides direct soil moisture and soil temperature measurements. Further
405
details are given in the online supplement. The catchment tomography approach, in 17
406
which localized precipitation events are considered as transmitters and runoff gauges
407
act as integrating receivers, is explored as an alternative approach to determine
408
catchment properties and their impact on water fluxes.
409 410
5 Outlook
411 412
With our observation and monitoring capabilities set up to the optimum and the fully
413
developed integrated terrestrial system models for the meter and kilometer scales in
414
place, the ensuing last DFG-funded phase of the TR32 will focus on model-data
415
fusion via combined parameter estimation and state variable assimilation employing
416
the Data Assimilation Research Testbed (DART, Anderson et al., 2009) in order to
417
address predictability of the terrestrial system on the catchment scale. We envision a
418
complete high-resolution reanalysis data set of a real mesoscale terrestrial system on
419
the sub-kilometer scale, which will be of interest for the wider scientific community.
420
The TR32 will employ this data set for the development of an extended view on
421
patterns in terrestrial systems. To this goal we will extend the object-based view
422
currently being developed for the atmospheric sciences for the analysis of e.g. the
423
structure and life cycles of convective systems or atmospheric rivers (e.g. Sellars et
424
al., 2013) to include pattern linkages between the surface and subsurface and
425
structures of the atmospheric boundary layer. This will require the consideration and
426
inclusion of the different time scales on which processes act in the contributing
427
compartments. Following our initial LES-based results on the propagation of patterns
428
from the land surface to the ABL, this requires the consideration of objects living on
429
timescales related to sub-objects of the slowest component like groundwater or soil
430
moisture.
431 18
432
Box 1 (Soil Root Model)
433 434
Plant roots play an important role in the terrestrial water cycle as they take up water
435
from the soil that is pumped up back into the atmosphere by vegetation. Since plant
436
roots increase the depth of the soil layer from which water can be taken up and
437
transpired back into the atmosphere, the depth of the root zone is an important
438
structural feature of terrestrial systems. Root zone depths depend on the interaction
439
between vegetation, climate, and soil (Schenk and Jackson, 2005) and show a
440
seasonal dynamics. The plasticity of root systems to changing environmental
441
conditions is therefore an additional important feature of terrestrial systems. Besides
442
root zone depths, the root density distribution is also important since water is
443
expected to be taken up more easily from soil layers with a high root density. The
444
structure of the plant root system therefore has an imprint on soil water distributions,
445
with soil layers with a high root density drying out more rapidly. However, how root
446
water uptake changes when part of the root zone dries out during dry spells is a
447
source of uncertainty in models. Neglecting compensatory uptake from deeper soil
448
layers in simulation models is considered to be the reason for the underestimation of
449
transpiration during dry spells (Wang and Dickinson, 2012). In order to improve the
450
prediction of root water uptake dynamics, a biophysical model that couples flow and
451
transport processes in the soil and plant root system while spatially resolving water
452
fluxes to single roots has been developed (Javaux et al., 2008). Simulations by this
453
model (Figure Box 1 (Soil-Root Model)) were used to infer upscaling rules that can
454
be used to parameterize larger-scale simulation models that do not resolve single
455
plant roots (Javaux et al., 2013; Couvreur et al., 2012). Processes like root growth
456
and regulation of transpiration by hormonal signals that are produced in the root zone
457
as a function of soil environmental conditions have been implemented in the model to 19
458
investigate whether these processes lead to a fundamentally different behavior of
459
root water uptake. In order to verify the behavior of the model, a field research facility
460
has been constructed in which root growth, soil water content and soil water
461
potential, and plant transpiration are non-invasively measured in different soil types
462
with different water holding capacity and for different water application regimes.
463
20
464
BOX 2 (TerrSysMP)
465 466
The TR32 Terrestrial System Modeling Platform (TerrSysMP, Figure Box 1
467
(TerrSysMP), Shrestha et al., 2014) couples the hydrological model ParFlow
468
(Lawrence Livermore Laboratory, e.g. Kollet and Maxwell, 2008; Kollet et al., 2010),
469
the land surface scheme Community Land Model (CLM, Version 3.5 (Oleson et al.,
470
2008) of NCAR, and the weather forecasting and regional climate model COSMO of
471
DWD and the COSMO-community (e.g. Baldauf et al., 2011) using the OASIS3
472
coupler (Ocean Atmosphere Sea Ice Soil, Version 3, e.g. Valke et al., 2012)
473
developed under the coordination of CERFACS (Centre Européen de Recherche et
474
de Formation Avancée en Calcul Scientifique). OASIS uses a dynamical two-way
475
approach including down- and upscaling algorithms for fluxes and state variables
476
between computational grids of different resolution (Shrestha et al., 2014). The
477
upscaling algorithm uses the mosaic or explicit sub-grid approach (Avissar and
478
Pielke, 1989) in which high-resolution land surface fluxes are averaged to the coarser
479
resolution of the atmosphere before they are passed over to the atmospheric model.
480
In addition, a downscaling scheme following Schomburg et al. (2010, 2012) is
481
implemented which downscales atmospheric variables of the lowest layer to the
482
higher-resolved land surface model. The scheme involves three steps: (1) spline
483
interpolation while conserving mean and lateral gradients of the coarse field, (2)
484
deterministic downscaling rules developed via statistics and genetic programming to
485
exploit empirical relationships between atmospheric variables and surface variables;
486
(3) addition of high-resolution variability (i.e. noise) in order to restore spatial
487
variability. All component models can be run in stand-alone or arbitrary coupled
488
mode. When CLM is coupled with ParFlow, both models share the same upper soil
489
layers. ParFlow takes over the complete hydrological calculations while sources and 21
490
sinks by rainfall and evapotranspiration are provided by CLM. The simulation (Figure
491
Box 2 (TerrSysMP)) shows the 3D distribution of the relative soil saturation along
492
with the turbulent eddy heat fluxes. A clear spatial structure can be observed at the
493
land surface, with topographic convergent zones (i.e., river corridors) experiencing
494
higher saturation values. In the atmosphere, patterns seem mainly controlled by the
495
vegetation distribution. For instance, higher values of turbulent eddy fluxes can be
496
identified over forested areas characterized by higher sensible heat fluxes. This can
497
be explained by the increased available energy (via lower albedo) over such areas
498
compared e.g. to crops. Note also that this effect tends to be amplified over steeper
499
terrain with its lower level of soil saturation due to a more efficient lateral drainage.
500 501
22
502 503
References (main text and boxes):
504 505
Anderson, J., T. Hoar, K. Raeder, H. Liu, N. Collins, R. Torn, and A. Arellano, 2009:
506
The Data Assimilation Research Testbed: A Community Facility. Bulletin of the
507
American Meteorological Society, 90, 1283-1296. doi:10.1175/2009BAMS2618.1.
508 509
Avissar, R., and R. A. Pielke, 1989: A parameterization of heterogeneous land
510
surfaces for atmospheric numerical models and its impact on regional meteorology,
511
Monthly Weather Review 117: 2113-2136.
512 513
Baatz, R., H. Bogena, H.-J. Hendricks Franssen, J.A. Huisman, Q. Wei, C. Montzka
514
and H. Vereecken, 2014: Calibration of a catchment scale cosmic-ray soil moisture
515
network: A comparison of three different methods. J. Hydrol.,
516
http://dx.doi.org/10.1016/j.jhydrol.2014.02.026.
517 518
Baldauf, M., A. Seifert, J. Förstner, D. Majewski, M. Raschendorfer, and T.
519
Reinhardt, 2011: Operational Convective-Scale Numerical Weather Prediction with
520
the COSMO Model: Description and Sensitivities. Mon. Wea. Rev., 139, 3887–
521
3905.doi: http://dx.doi.org/10.1175/MWR-D-10-05013.
522 523
Bogena, H.R., Herbst, M., Huisman, J.A., Rosenbaum, U., Weuthen, A., &
524
Vereecken, H. 2010: Potential of Wireless Sensor Networks for Measuring Soil Water
525
Content Variability. Vadose Zone Journal, 9, 1002-1013.
526
23
527
Bornemann, L., M. Herbst, G. Welp, H. Vereecken, W. Amelung, 2011: Rock
528
fragments control organic carbon pool sizes in agricultural topsoil. Soil Sci. Soc. Am.
529
J., 75, 1898-1907.
530 531
Bornemann, L., G. Welp, W. Amelung, 2010: Particulate Organic Matter at the Field
532
Scale: Rapid Acquisition Using Mid-infrared Spectroscopy. Soil Sci. Soc. Am. J., 74,
533
1147-1156.
534 535
Busch, S., J. van der Kruk, and H. Vereecken, 2013: Improved characterization of
536
fine texture soils using on-ground GPR full-waveform inversion. IEEE Transaction on
537
Geoscience and Remote Sensing. PP, 1-12. DOI: 10.1109/TGRS.2013.2278297.
538 539
Cornelissen, T., B. Diekkrüger, and H. R. Bogena, 2014: Importance of a bedrock in
540
the 3D simulation of discharge and soil moisture patterns on different spatial and
541
temporal
542
10.1016/j.jhydrol.2014.01.060
scales
–
The
Wüstebach
case
study.
J.
Hydrology.
doi
543 544
Couvreur, V., J. Vanderborght, and M. Javaux, 2012: A simple three-dimensional
545
macroscopic root water uptake model based on the hydraulic architecture approach.
546
Hydrology and Earth System Sciences, 16, 8, 2957-2971.
547 548
Curdt, C., D. Hoffmeister, C. Jekel, G. Waldhoff and G. Bareth, 2012: Scientific
549
Research Data Management for Soil-Vegetation-Atmosphere Data – The TR32DB.
550
The International Journal of Digital Curation, 7 (2), 68-80, doi:10.2218/ijdc.v7i2.208.
551
24
552
Damm A., Erler A., Hillen W., Meroni M., Schaepman M.E., Verhoef W. & Rascher
553
U., 2011: Modeling the impact of spectral sensor configurations on the FLD retrieval
554
accuracy of sun-induced chlorophyll fluorescence. Remote Sensing of Environment,
555
115, 1882-1892.
556 557
De Rosnay, P., M. Drusch, A. Boone, G. Balsamo, B. Decharme, P. Harris, Y. Kerr,
558
T. Pellarin, J. Polcher, and J. P. Wigneron, 2009: AMMA Land Surface Model
559
Intercomparison Experiment coupled to the Community Microwave Emission Model:
560
ALMIP-MEM, J. Geophys Res.-Atmos., 114, D05108, DOI: 10.1029/2008JD010724.
561 562
Langensiepen, M., M. Kupisch, M.T. van Wijk, F. Ewert, 2012: Analyzing transient
563
closed chamber effects on canopy gas exchange for optimizing flux calculation
564
timing. Agricultural and Forest Meteorology, 164, 61-70.
565 566
Hasan S., C. Montzka, C. Rüdiger, M. Ali, H. Bogena, and H. Vereecken 2014: Soil
567
moisture retrieval from airborne L-band passive microwave using high resolution
568
multispectral data. J. Photogr Remote Sens, 91: 59-71, DOI:
569
10.1016/j.isprsjprs.2014.02.005.
570 571
Hebel, C. van, S. Rudolph, A. Mester, J. A. Huisman, P. Kumbhar, H. Vereecken,
572
and J. van der Kruk, 2014: Three-dimensional imaging of subsurface structural
573
patterns using quantitative large-scale multi-configuration electromagnetic induction
574
data. Water Resources Research, 50, 27322748, doi:10.1002/2013WR014864.
575
25
576
Herbst, M., L. Bornemann, A. Graf, G. Welp, H. Vereecken, W. Amelung, 2012: A
577
geostatistical approach to the field-scale pattern of heterotrophic soil CO2 emission
578
using covariates. Biogeochemistry, 111, 377-392.
579 580
Hoffmeister, D., G. Waldhoff, C. Curdt, N. Tilly, J. Bendig and G. Bareth, 2013:
581
Spatial variability detection of crop height in a single field by terrestrial laser
582
scanning. In: Stafford, J.V. (ed.): Precision agriculture’13, Proc. of the 9th European
583
Conference on Precision Agriculture, 7-11 July 2013, Lleida, Spain, 267-274.
584 585
Hunt, B. R., E. J. Kostelich, and I. Szunyogh, 2007: Efficient data assimilation for
586
spatiotemporal chaos: A local ensemble transform Kalman filter, Physica D:
587
Nonlinear Phenomena, 230, 1-2, 112-126.
588 589
Javaux, M., T. Schröder, J. Vanderborght, and H. Vereecken, 2008: Use of a three-
590
dimensional detailed modeling approach for predicting root water uptake. Vadose
591
Zone Journal 7, 3, 1079-1088.
592 593
Javaux, M., V. Couvreur, J. Vanderborght, and H. Vereecken, 2013: Root Water
594
Uptake: From Three-Dimensional Biophysical Processes to Macroscopic Modeling
595
Approaches. Vadose Zone Journal 12(4), doi:10.2136/vzj2013.02.0042.
596 597
Kemna, A., Binley, A., Cassiani, G., Niederleithinger, E., Revil, A., Slater, L.,
598
Williams, K., Flores-Orozco, A., Haegel, F., Hoerdt, A., Kruschwitz, S., Leroux, V.,
599
Titov, K., and E. Zimmermann, 2012: An overview of the spectral induced
600
polarization method for near-surface applications. Near Surface Geophysics. 11 (7),
601
453-468. DOI: 10.3997/1873-0604.2012027. 26
602 603
Kollet S., R. M. Maxwell, 2008: Capturing the influence of groundwater dynamics on
604
land surface processes using an integrated, distributed watershed model. Water
605
Resour Res 44:W02,402.doi:10.1029/ 2007WR006004
606 607
Kollet, S. J., R. M. Maxwell, C. S. Woodward, S. Smith, J. Vanderborght, H.
608
Vereecken and C. Simmer, 2010: Proof of concept of regional scale hydrologic
609
simulations at hydrologic resolution utilizing massively parallel computer resources,
610
Water Resour. Res. 46, W04201, doi:10.1029/2009WR008730.
611 612
Korres, W., T.G. Reichenau, and K. Schneider, 2013: Patterns and scaling properties
613
of surface soil moisture in an agricultural landscape: An ecohydrological modeling
614
study. Journal of Hydrology, 498, 89, doi:10.1016/j.jhydrol.2013.05.050.
615 616
Koyama, C.N., W. Korres, P. Fiener, K. Schneider, 2010: Variability of Surface Soil
617
Moisture Observed from Multitemporal C-Band Synthetic Aperture Radar and Field
618
Data. Vadose Zone Journal, 9, 1014.
619 620
Kustas, W. and M. Anderson: 2009: Advances in thermal infrared remote sensing for
621
land surface modeling. Agricultural and Forest Meteorology, 149, 2071-2081.
622 623
Mester, A., van der Kruk, J., Zimmermann, E., Vereecken, H., 2011: Quantitative
624
Two-Layer Conductivity Inversion of Multi-Configuration Electromagnetic Induction
625
Measurements. Vadose Zone J., 10, 4, 1319-1330. DOI: 10.2136/vzj2011.0035.
626
27
627
Mohnke, O., and N. Klitzsch, 2010: Microscale simulations of NMR relaxation in
628
porous media considering internal field gradients. Vadose Zone Journal, 9, 846–857.
629 630
Oleson, K. W., G. Y. Niu, Z. L. Yang, D. M. Lawrence, P. E. Thornton, P. J.
631
Lawrence, R. Stöckli, R. E. Dickinson, G. B. Bonan, S. Levis, A. Dai, and T. Qian,
632
2008: Improvements to the Community Land Model and their impact on the
633
hydrological cycle. J. Geophys. Res., 113, doi: 10.1029/2007JG000563
634 635
Perlo, J., E. Danieli, B. Blümich, and F. Casanova, 2013: Optimized slim-line logging
636
NMR tool to measure soil moisture in situ. Journal of Magnetic Resonance. 233, 74-
637
79. DOI: 10.1016/j.jmr.2013.05.004.
638 639
Rahman M., S. Kollet, and M. Sulis, 2014: The concept of dual-boundary forcing in
640
land surface-subsurface interactions of the terrestrial hydrologic and energy cycles,
641
Water Resour. Res., (in review).
642 643
Rosenbaum, U., H. R. Bogena, M. Herbst, J. A. Huisman, T. J. Peterson, A.
644
Weuthen, A. W. Western, and H. Vereecken, 2012: Seasonal and event dynamics of
645
spatial soil moisture patterns at the small catchment scale. Water Resour. Res., 48,
646
10, W10544, doi:10.1029/2011WR011518.
647 648
Ryzhkov, A., M. Diederich, P. Zhang, and C. Simmer, 2014: Potential utilization of
649
specific attenuation for rainfall estimation, mitigation of partial beam blockage, and
650
radar networking. Journal of Atmospheric and Oceanic Technology, 31, 599-619.
651
28
652
Schenk, H.J. and Jackson, R.B., 2005: Mapping the global distribution of deep roots
653
in relation to climate and soil characteristics. Geoderma 126(1-2), 129-140.
654 655
Schomburg, A., V. Venema, R. Lindau, F. Ament, and C. Simmer, 2010: A
656
downscaling scheme for atmospheric variables to drive soil-vegetation-atmosphere
657
transfer models. Tellus, 62, 4, p.242-258, doi 10.1111/j.1600-0889.2010.00466.
658 659
Schomburg, A., V. Venema, R. Lindau, F. Ament, and C. Simmer, 2012:
660
Disaggregation of screen-level variables in a numerical weather prediction model
661
with an explicit simulation of subgrid-scale land-surface heterogeneity. Meteorology
662
and Atmospheric Physics, doi: 10.1007/s00703-012-0183-y, 116 (3-4), 81-94.
663 664
Schween, J.H., S. Crewell, and U. Löhnert, 2011: Horizontal-humidity gradient from
665
one single-scanning microwave Radiometer, IEEE Geosci. Remote Sens. Lett. 8(2),
666
336-340. doi:10.1109/LGRS.2010.2072981
667 668
Sellars, S., P. Nguyen, W. Chu, X. Gao, K. Hsu, and S. Sorooshian, 2013:
669
Computational Earth Science: Big data transformed into insight. EOS, 94, 32, 277-
670
279.
671 672
Shao, Y. S. Liu, S. Crewell, and J.H. Schween, 2013: Large-Eddy Atmosphere - Land
673
Surface Modeling over Heterogeneous Surfaces: Model Development and
674
Comparison with Measurements. Boundary-Layer Meteorology, 148 (2), 333-356,
675
DOI:10.1007/s10546-013-9823-0.
676 29
677
Shuttleworth, J., Rosolem, R., Zreda, M., and T. E. Franz, 2013: The COsmic-ray
678
Soil Moisture Interaction Code (COSMIC) for use in data assimilation. Hydrology and
679
Earth System Sciences, 17, 3205-3217.
680 681
Shrestha, P., M. Sulis, M. Masbou, S. Kollet, and C. Simmer, 2014: A scale-
682
consistent Terrestrial System Modeling Platform based on COSMO, CLM and
683
ParFlow. Accepted for publication in Mon. Wea. Rev..
684 685
Valcke, S., V. Balaji, A. Craig, C. DeLuca, R. Dunlap, R. W. Ford, R. Jacob, J.
686
Larson, R. O’Kuinghttons, G. D. Riley, and M. Vertenstein, 2012: Coupling
687
technologies for Earth System Modeling. Geosci. Model Dev., 5, 1589–1596.
688 689
Vereecken H., S. Kollet, and C. Simmer, 2010: Patterns in Soil–Vegetation–
690
Atmosphere Systems: Monitoring, Modeling, and Data Assimilation. Vadose Zone J.,
691
9, 821–827, doi:10.2136/vzj2010.0122
692 693
Wang, K.C. and Dickinson, R.E. (2012) A Review of Global Terrestrial Evaporation:
694
Observation, Modelling, Climatology, And Climatic Variability. Reviews of Geophysics
695
50.
696 697
Zacharias, S., H. Bogena, L. Samaniego, M. Maude, R. Fuss, T. Putz ,M. Frenzel, M.
698
Schwank, C. Baessler, K. Butterbach-Bahl, O. Bens, E. Borg, A. Brauer, P. Dietrich,
699
I. Hajnsek, G. Helle, R. Kiese, H. Kunstmann, S. Klotz, J. C. Munch, H. Papen, E.
700
Priesack, H.P. Schmid, R. Steinbrecher, U. Rosenbaum, G. Teutsch, H. Vereecken,
30
701
2011: A Network of Terrestrial Environmental Observatories in Germany. Vadose
702
Zone J., 10, 3, 955-973, DOI: 10.2136/vzj2010.0139
703 704
Zreda, M., D. Desilets, T. P. A. Ferre, and R. L. Scott, 2008: Measuring soil moisture
705
content non-invasively at intermediate spatial scale using cosmic-ray neutrons.
706
Geophysical Research Letters, 35, L21402
707
31
708
Figure captions (main text and boxes)
709 710
Figure 1 (Rur Catchment Monitoring): Rur catchment including instrumentation and
711
precipitation radar coverage from TR32 and TERENO.
712 713
Figure 2 (EMI results): a) satellite-derived Leaf Area Index (LAI) distribution in Mai
714
2011 estimated after a two-month drought period over the lower right field in e). b)
715
and c) EMI ECa measurements with depth sensitivity up to 0.5 and 1.8 m,
716
respectively, measured in June 2012. d) Photo taken in August 2013 (camera
717
position and view delineated in a) – c)) that indicates stressed (A and C) and
718
unstressed (B) regions in a sugar beet field. e) ECa measurements for a larger area
719
with green colors indicating high and red colors low values. Red dashed lines
720
indicate paleo river channels. f) is a blowup of the Selhausen testsite where the EMI
721
measurement lines are clearly visible. Open areas indicate man-made structures. g)
722
quasi 3-D EC distribution of the subsurface obtained by a 3-layer inversion of multi-
723
configuration ECa measurements using a Maxwell forward model at every grid point
724
(modified from von Hebel et al., 2014 and Rudolph et al., in preparation).
725 726
Figure 3 (Sun-Induced Fluorescence): Air-borne observation of sun-induced
727
fluorescence from Sept 2012 600 meters above ground using the high-performance
728
imaging spectrometer HyPlant. HyPlant allows the quantification of the emitted red
729
fluorescence of active chlorophyll in the oxygen absorption line at 760nm, which is
730
directly related to the efficiency of photosynthesis. Thus the map illustrates the
731
variable photosynthetic rates of the different vegetation types. Highest fluorescence
732
signals come sugar beet while other vegetation types were already approaching
733
autumn senescence. 32
734 735
Figure 4 (Lattice Boltzmann simulations): Lattice Boltzmann simulations at the pore
736
scale: a) Fluid distribution in the pore space of a sand sample (air in blue and water
737
in red) at a water saturation of 0.6 and b) simulated (lines) and measured (circles)
738
NMR saturation recovery data of the same sample for water saturations between 0.3
739
and 1 (color coded).
740 741
Figure 5 (HGS Dynamics): Soil moisture simulation results for the 27 ha headwater
742
sub-catchment of the Rur catchments at 25 m spatial resolution for the period 2010
743
and 2011.
744 745
Figure 6 (HGS Soil Moisture Patterns): Spatial distribution of absolute soil moisture
746
(vol. %) on 13.01.2011 (upper figure) and 30.5.2011 (lower figure) for measured (top)
747
and simulated data for 25 m. The values in brackets refer to the mean standard
748
deviation of the kriging algorithm.
749 750
Figure 7 (LES Pattern Propagation): Shown in column (a) are patterns of
751
instantaneous sensible heat flux (deviation from domain average) in Wm-2 for 1300Z
752
at levels of 2, 8, 32 and 512m. Column (b) is as (a) but for patterns of sensible heat
753
flux averaged over 30 minutes. Column (c) shows the Haar wavelet energy spectra.
754
The sensible heat fields are Haar-decomposed with window sizes of 2, 4, 8, 16 and
755
32Δx (with Δx = 60 m).
756 757
Figure 8 (Atmospheric Downscaling): Disaggregation of a temperature field on an
758
almost cloud-free night. Under such conditions temperature inversions cause cold air
759
to drain into the valleys, which leads to pronounced channel structures in the 33
760
temperature field with substantial variability contained in the fine scales: (a) shows
761
the temperature field simulated by the COSMO (atmospheric component of
762
TerrSysMP) at coarse resolution (2.8 km); (b) shows the downscaled field at 400 m
763
resolution resulting from the GP based disaggregation algorithm; (c) shows the
764
reference field from a high-resolution (400 m) COSMO run.
765 766
Figure 9 (WTD-LE Cross-Wavelets): Cross-wavelet spectrum of evapotranspiration,
767
LE, and water table depth, WTD, simulated from TerrSysMP (CLM coupled to
768
ParFlow and driven with COSMO output) for the year 2009.
769 770
Figure Box 1 (Soil-Root Model): a) Installation of 54 8-m long transparent rhizotubes
771
(9 at 6 depths), b) recording images of roots along rhizotubes using a BTC2 video
772
microscope (Bartz Technology Corporation, Carpinteria, CA, USA), c) examples of
773
recorded images in rhizotubes at different depths from which the evolution of the
774
relative root density distribution during the growing season is derived, d) three-
775
dimensional simulation of root water uptake by a root architecture using the coupled
776
soil-root model R-SWMS (Javaux et al., 2008): color scale represents the soil water
777
pressure head (PH) and transparent cyan colors represent water uptake by root
778
segments.
779 780
Figure Box 2 (TerrSysMP Structure)): a) Schematic of TerrSysMP (modified after
781
Shrestha et al., 2014) showing the fluxes and state variables exchanged between the
782
three model components COSMO (atmosphere), CLM (landsurface and subsurface),
783
and ParFlow (subsurface hydrology) via the OASIS coupler. SW, LWdn, Rain, T, P,
784
QV and U (downward solar and terrestrial radiation, precipitation, screen
785
temperature, air pressure, specific humidity, and wind, respectively) are passed from 34
786
COSMO to CLM while CLM passes back SH, LH, TAU, LWup, and Albedo (sensible
787
and latent heat fluxes, friction velocity, upward terrestrial radiation, and albedo,
788
respectively). CLM share the upper soil layers with ParFlow within which qrain and qe
789
(infiltration and plant transpiration) are passé from CLM to ParFlow, which transmits
790
back Sw and Ψ (soil moisture and pressure head, respectively). b) Turbulent eddy
791
heat flux (km s-1) and relative soil saturation (-) along the Rur catchment. c) Main land
792
use classes in the catchment. d) Meridional cross-section of turbulent eddy heat flux
793
and relative soil saturation.
794
35
795
Supplement for Online Display
796 797
Supplement 1: The Jülich ObservatorY for Cloud Evolution (JOYCE)
798
A suite of ground-based remote sensing instruments is operated from the roof of
799
Forschungszentrum Jülich (FZJ) GmbH (50.908547°N, 6.413536°E, 105m MSL) next
800
to a 120 meter meteorological mast. Continuous long-term measurements
801
(http://gop.meteo.uni-koeln.de/~hatpro/dataBrowser) with high-temporal resolution (