Abstract. This talk is about the GEOtop and JGrass-NewAge model, their physical bases, their informatics based on older (the first) and new (the latter) programming paradigms, the lessons I learned in building them with my group of people in an academic environment, their future, and the understanding that there is no the best model, but certainly a better way to do models.
Hydrological modelling was for long time, and still is, almost a synonym of simulating rainfall-runoff. Recently, however, the scope of hydrology became wider, even among engineers. Modelling in hydrology now certainly still means modelling discharges, but also modelling snow, evapotranspiration and turbulent exchanges, and surface/subsurface interactions. With the goal of reproducing the whole picture of the terrestrial hydrological fluxes, my coworkers and I worked together in the last decade to build new models and new types of models. We started from the lesson by P. Eagleson, and we built first the process-based (grid based) GEOtop model. GEOtop is “terrain-based” (it is based on the use of digital terrain models and uses the knowledge of interaction between morphology and process) “distributed” (all the simulated variables are calculated for each pixel of the basin) model of “the water cycle” (it simulates all the components of the water cycle, accounting for both the mass budget and the energy budget, the two budget equations being coupled through the temperature of the soil, which controls evaporation, hydraulic conductivity, and accumulation of the snowpack). However, this GEOtop was intimidating many, either for the complexity of the process and its internals, and possibly not adapted to large scale modelling where faster solutions are required.
Therefore we also worked on a different, more parsimonious model, called JGrass-NewAGE. From the lesson learned by implementing and maintaining GEOtop, we also found necessary to build the new model on new informatics. This system sacrifices process details in favour of efficient calculations. It is made of components apt at returning statistical hydrological quantities, opportunely averaged in time and space. One of the goals of this implementation effort was to create the basis for a physico-statistical hydrology in which the hydrological spatially distributed dynamics are reduced into low dimensional components, when necessary surrogating the internal heterogeneities with "suitable noise" and a probabilistic description. Unlike other efforts of synthesis, JGrass-NewAge keeps the spatial description explicit, at various degrees of simplicity. This has been made possible by opportune processing of distributed information which, in this way, has become part of the model itself.
3. !3
Every Hydrologist would like to have
THE MODEL of IT
But in reality everybody wants just to investigate a limited set of
phenomena: for instance the discharge in a river. Or landsliding , or
soil moisture distribution.
Any problems requires its amount of prior information to
be solved: some problems needs more detailed information of others
R. Rigon
Introduction
4. !4
For the impatients I reveal the killer before
Up to a point*… there is no the best model
* See Klemes, Dilettantism in hydrology: Transition or destiny?, Wrr, 1986.
* See also: http://abouthydrology.blogspot.com/2012/02/which-hydrological-model-is-better-q.html
End of the story
R. Rigon
5. !5
Should we just care of process-based
models ?
The criticisms to this type of modelling have foundations.
PeakFlow
GEOtop
NewAge
Boussinesq
SHALSTAB GEOtop-FS The Horton Machine
and we have several models that we use at different scales and for
different purposes
We did not marry process based models
R. Rigon
7. !7
Every one of them:
!
!
!
Perform the mass budget (and preserves mass)
!
Make hypotheses on momentum variations
!
Simplify the energy conservation (and its dissipation)
to a certain degree
!
(Implicitly delineates a way to entropy increase)
R. Rigon
Ours have some in common
8. !8
!
(Rigon et al., Jour. Hydromet., 2006, Endrizzi et al., GMDD, 2014)
This model focuses on the water and energy budgets at few
square meters scale with the goal of describing catchment
hydrology including (a reasonable parameterization) all
known processes. (Whatever this means)
A first modelling adventure
see also: http://abouthydrology.blogspot.com/search/label/GEOtop
R. Rigon
GEOtop
9. !9
1. Radiation
4. surface energy balance
- radiation
- boundary-layer interaction
2. Water balance
- effective rainfall
- surface flow (runoff and channel
routing)
- distributed model
- sky view factor, self and cast
shadowing, slope, aspect, drainage
3. Snow-glaciers
- multilayer snow
scheme
- soil
temperature
- freezing soil
5. soil energy balance
- multi-layer vegetation
scheme
- evapotranspiration
6 . v e g e t a t i o n
interaction
R. Rigon
GEOtop
10. !10
snow, ice, permafrost
water cycle in
complex terrain
landsliding
evapo-transpiration,
energy fluxes
Bertoldi et al., 2006
Bertoldi et al 2010
DellaChiesa et al., 2014
Endrizzi 2007
Dall’Amico 2010
Endrizzi et al,
2010a,b
Endrizzi et al.,
2014
Simoni et al 2008
Lanni et al, 2010
Rigon et al., 2006
Hingerl et al., 2014
Formetta et al., 2014
Why this complexity ?
R. Rigon
GEOtop
13. !13
First, I would say, it means that it would be better to call it, for
instance: Richards-Mualem-vanGenuchten equation, since it is:
Se = [1 + ( ⇥)m
)]
n
Se :=
w r
⇥s r
C(⇥)
⇤⇥
⇤t
= ⇥ · K( w) ⇥ (z + ⇥)
⇥
K( w) = Ks
⇧
Se
⇤
1 (1 Se)1/m
⇥m⌅2
SWRC +
Darcy-Buckingham
(1907)
Parametric
Mualem (1976)
Parametric
van Genuchten
(1981)
C(⇥) :=
⇤ w()
⇤⇥
Not only this:
What I mean with Richards++
R. Rigon
14. !14
For instance this:
Extending Richards to treat the transition from saturated to unsaturated
zone. Which means:
What I mean with Richards++
R. Rigon
15. !15
So, consider a traditional 1D infiltration problem
R. Rigon
An example
16. !16
So, consider a traditional 1D infiltration problem
usually it cannot be treated with Richards because of the saturation front
R. Rigon
An example
17. !17
But GEOtop is also 3D
After Lanni et al, 2010 , unpublished
R. Rigon
GEOtop does 3D
18. !18
Landsliding
dry case - low intensity precipitation
After Lanni et al, 2010 , unpublished
R. Rigon
GEOtop does 3D
19. !19
Landsliding
wet case - high intensity precipitation
After Lanni et al, 2010 , unpublished
R. Rigon
GEOtop does 3D
20. !20
More complex stuff
Extending Richards to treat the phase transition. Which means essentially to
extend the soil water retention curves to become dependent on temperature.
Unsaturated
unfrozen
Freezing
starts
Freezing
procedes
Unsaturated
Frozen
What I mean with Richards++
R. Rigon
21. !21
pw0 = pa wa
⇥Awa(r0)
⇥Vw
= pa pwa(r0) pi = pa ia
⇥Aia(r0)
⇥Vw
:= pa pia(r0)
pw1 = pa ia
⇥Aiar(0)
⇥Vw
iw
⇥Aiw(r1)
⇥Vw
Two interfaces (air-ice and water- ice) should be considered!!!
Curved interfaces with three phases
Four phases … well interfaces are phases too, indeed
R. Rigon
22. !22
A further assuption
To make it manageable, we do a further assumption. Mainly the freezing=drying
one.
Considering the assumption “freezing=drying” (Miller, 1963) the ice “behaves
like air” and does not add further pressure terms
Freezing=Drying
R. Rigon
23. !23
Unfrozen water content
soil water
retention curve
thermodynamic
equilibrium (Clausius Clapeyron)
+
⇥w =
pw
w g
pressure head:
w(T) = w [⇥w(T)]
How this reflects on pressure head
Freezing=Drying
R. Rigon
27. !27
T := T0 +
g T0
Lf
w0
ice content: i =
⇥w
⇥i
w
⇥
⇥w = ⇥r + (⇥s ⇥r) ·
⇤
1 + ⇤w0
Lf
g T0
(T T⇥
) · H(T T⇥
)
⇥n⌅ m
liquid water content:
Total water content:
depressed
melting point
Modified Richards equations
= ⇥r + (⇥s ⇥r) · {1 + [ · ⇤w0]
n
}
m
Water and ice mass budget
R. Rigon
29. !29
Obviously this makes it possible to simulate
a lot of new phenomenologies
Sisik, river in the artic tundra
EndrizzietAl.,JHR,2010
R. Rigon
Do you care of runoff on frozen soil ?
30. !30
44
thaw depth: T(z,t)=0 water table depth: ψm(z,t)=0
Stefano Endrizzi, William Quinton, Philip Marsh, Matteo Dall’Amico, 2010 in preparation
R. Rigon
Do you care of runoff on frozen soil ?
31. !31
The model allows to show that the runoff
properties of a basin dramatically change when
soil freeze.
Runoff on frozen soil
R. Rigon
Do you care of runoff on frozen soil ?
35. !35
So well tested that is confidently used for real-time
forecasting (driven by ground data)
Use it !
R. Rigon
36. !36
An experimental elevation transect
Elevation as a proxy of climate change: Mazia Valley, emerging LTER
Station
B2000 m
Hs, SWC,
Biomass, GAI
Station
B1500 m
Hs, SWC,
Biomass, GAI,ET
Station
B1000 m
Hs, SWC,
Biomass, GAI
T~ 3.5K
T~ 3.5K
Courtesy of G. Bertoldi, EURAC. Complete presentation and reference at:
http://abouthydrology.blogspot.com/2014/05/process-based-hydrological-modelling-of.html
R. Rigon
Eco-hydrology of mountain prairies
37. !37
Elevation gradient: validation
Multiple variables validation: SWE, SWC, above ground biomass (Bag), ET
Two years of data: calibration in B1500, validation in B1000, B2000
B2000mB1500mB1000m
Snow Height [cm] SWC 5cm [] ET [mm]
Not Measured
Not Measured
r2=0.66
RMSE=7.1
r2=0.57
RMSE=5.9
r2=0.55
RMSE=2.9
r2=0.80
r2=0.78
r2=0.82
Bag [gDMm 2]
RMSE=0.04
RMSE=0.05
RMSE=0.04
r2=0.93
RMSE=58.39
Courtesy of G. Bertoldi, EURAC. Complete presentation and reference at:
http://abouthydrology.blogspot.com/2014/05/process-based-hydrological-modelling-of.html
R. Rigon
Eco-hydrology of mountain prairies
38. !38
The GEOtop 2.0 – DV model
Rigon et al., JHM, 2006;
Endrizzi et al. GMDD, 2014.
Processes
Dynamic vegetation
model (for grasslands)
From Montaldo et al., 2005;
Della Chiesa et al., 2014
R. Rigon
Eco-hydrology of mountain prairies
39. !39
So GEOtop is a succes story !
Is’nt it ?
R. Rigon
A synthesis
40. !40
You can find the GEOtop code at:
git clone https://code.google.com/p/geotop/
Compiling instructions:
http://abouthydrology.blogspot.com/2014/04/installing-
geotop-on-mac-and-linux.html
Manual:
http://abouthydrology.blogspot.com/2011/08/new-
version-of-geotop-with-draft-user.html
User and Developers:
geotopusers@googlegroups.com
geotopdev@googlegroups.com
If you like you can use it !
R. Rigon
41. !41
However
Developing GEOtop while learning about the
processes and the appropriate numerics required a
lot of code rewriting.
Every student working on GEOtop cancelled hours of
work of the other students.
The code was built as a “monolithic” software, and
this makes its maintenance very difficult, even
having the source code
R. Rigon
Looking behind to the whole process of building GEOtop
42. !42
While developing GEOtop, the coded evolved, and
third parties developers, doing applications, got
mad in adapting their code to the new versions.
And
As Olaf D. cites: “A fool with a tool remains a fool”.
And if someone goes crazy in developing a tool
eventually s/he fall in the above case.
R. Rigon
Looking behind to the whole process of building GEOtop
43. !43
A second model adventure
Picasso,DoraMaar
Deconstructing models
R. Rigon
Modelling a different way or perish
44. !44
Therefore we have to find a new way to build
models
That enhances
•cooperation among researchers,
•the analysis of hydrological processes,
•the comparison among different modelling solutions,
•the adoption of reproducible research strategies,
•sharing of model codes,
•reproduction of research simulations,
Modern OO tools can help
R. Rigon
Modelling a different way or perish
45. !45
Modelling by components: a solution
I am in the home of modelling by components, here, but let me repeat for those
are unaware of it. In modelling by components, every process becomes a “piece
of software” that can be programmed and inspected independently from the
other components. Components interact just at run-time, after have been
linked together, for instance with a scripting language, in an intermediate
phase.
R. Rigon
Modelling by components
46. !46
To make a long story short, we chose OMS
OMS3 can be found at: http://www.javaforge.com/project/
Resources
Knowledge
Base
Development
Tools
Products
OMS3
http://www.javaforge.com/project/oms
R. Rigon
Modelling by components
47. !47
The framework offers new exciting possibilities
So we have a foundational theoretical
declarations about JGrass-NewAGE
“…This system sacrifices process details in favour of efficient
calculations. It is made of components apt at returning statistical
hydrological quantities, opportunely averaged in time and space.
One of the goals of this implementation effort was to create the
basis for a physico-statistical hydrology in which the hydrological
spatially distributed dynamics is reduced into low dimensional
components, when necessary surrogating the internal heterogeneities
with "suitable noise" and a probabilistic description ….”
R. Rigon
Peruse and abuse of models
48. !48
In practice what we implemented
is a trade-off between the official morality and a more practical and
agnostic view, where we do not expect to derive the statistical laws first
and implement them eventually, but we adopt right away some solution
that compromise among experimental evidence, scientific knowledge,
mathematical convenience, and computational tractability ... and the
natural laziness that everybody has.
On the other hand, being easy exchanging components (and to a certain
extent to produce them) it is easy (once you have them) to compare
components with the same scope, independently from the heuristic that
generated them.
Being realistic
R. Rigon
50. !50
JGrass-NewAGE
(Formetta et al., GTD, 2011)
This model focuses on the hydrological budgets of medium
scale to large scale basins as the product of the processes
“averaged” at the hillslope scale with the interplay of the
river network.
JGrass-NewAGE a.k.a. NewAGE
51. !51
G. Formetta et al.: Snow water equivalent modeling component in NewAge-JGrass
ation component (Eberhart and
) component (Hay et al., 2006);
Adaptive Metropolis (DREAM)
, 2009).
lope-link geometrical partition
unit for the water budget eval-
illslope, rather than a cell or a
ciated link. The model requires
gical forcing data (air tempera-
midity) for each hillslope. This
deterministic inverse distance
1992; Lloyd, 2005), kriging
ed kriging as in Garen et al.
(2005).
metta et al., 2013) implements
ount shadows and complex to-
n under generic sky conditions
ng to Helbig et al. (2010) and
on choices such as Erbs et al.
nd Orgill and Hollands (1977).
et is based on Brutsaert (1982)
(including those not described
ne of the automatic calibration
particle swarm optimization al-
M. Evaluation of each model
tually carried out with the use
lidation), which provides some
oodness of fit, such as Nash–
Fig. 1. The NewAge system showing all the modeling compo-
nents, starting from the top: the uDig Geographic Information Sys-
tem (GIS), the meteorological data interpolation tools, energy bal-
ance, evapotranspiration, runoff production-routing and snow water
The structure of NewAGE
R. Rigon
53. !53
Rinaldo,GeomorphicFloodResearch,2006
For each of the variable of the hydrological cycle
a statistics is made for each hillslope and a single value is returned
so, we have 5 values of the prognostics quantities here, that are space
time-averages of what happens inside each hillslope
The structure of NewAGE
R. Rigon
54. !54
They are estimated
for each hillslope
•mean or suitable rainfall
!
•mean or suitable radiation (we exploit some old idea by Ian Moore)
!
•mean or suitable evapotranspiration
!
•mean or suitable snow cover
!
•mean or suitable runoff production
The structure of NewAGE
R. Rigon
55. !55
Subsequently, the user can choose between two different runoff
Fig. 4. Hillslope-link partition of the basin work-flow.
G. Formetta et al. / Environmental Modelling & Software 55 (2014) 190e200 195
So components for watershed partition
The treatment of the topographic data first
R. Rigon
Formettaetal.,Hydrologicalmodellingwithcomponents:AGIS-basedopen-source
framework,2014
57. !57Fig. 6. The workflow for the Fort Cobb river basin application.
G. Formetta et al. / Environmental Modelling & Software 55 (2014) 190e200
Rainfall-Runoff*
R. Rigon
Formettaetal.,Hydrologicalmodellingwithcomponents:AGIS-basedopen-source
framework,2014
58. !58
When runoff is collected
then is routed, for small basins, with a modification of the Muskingum-Cunge
algorithm, or directly with a semi-implict solver of the de Saint-Venant 1D
R. Rigon
Watershed model of NewAGE
59. !59
Thus we have discharges
Here, Here ... and here again
R. Rigon
Watershed model of NewAGE
60. !60
Input Data treatment
Goodness of fit
Next time step
JGrass-NewAge
Calibration tools
R. Rigon
The structure of NewAGE
Formetta et al., GTD, 2011,
Formetta et al, EM&S, 2014
61. !61Fig. 6. The workflow for the Fort Cobb river basin application.
G. Formetta et al. / Environmental Modelling & Software 55 (2014) 190e200
Forcings and Calibration
R. Rigon
Formettaetal.,Hydrologicalmodellingwithcomponents:AGIS-basedopen-source
framework,2014
63. !63
G. Formetta et al.: Modeling shortwave solar radiation using the JGrass-NewAge system 919
Fig. 1. OMS3 SWRB components of JGrass-NewAge and flowchart
to model shortwave radiation at the terrain surface with generic sky
conditions. Where not specified, quantity in input or output must be
intended as a spatial field for any instant of simulation time. ”Mea-
Fig. 1. OMS3 SWRB components of JGrass-NewAge and flowchart to model shortwave radiation at the terrain surface with generic sky
conditions. Where not specified, quantity in input or output must be intended as a spatial field for any instant of simulation time. “Measured”
efers to a quantity that is measured at a meteorological station. The components, besides the specified files received in input, include an
appropriate set of parameter values.
Radiation
R. Rigon
Formettaetal.,ModelingshortwavesolarradiationusingtheJGrass-NewAge
system,2013
64. !64
G. Formetta et al.: Modeling shortwave solar radiation using the JGrass-NewAge system 919
Fig. 1. OMS3 SWRB components of JGrass-NewAge and flowchart
to model shortwave radiation at the terrain surface with generic sky
conditions. Where not specified, quantity in input or output must be
intended as a spatial field for any instant of simulation time. ”Mea-
Fig. 1. OMS3 SWRB components of JGrass-NewAge and flowchart to model shortwave radiation at the terrain surface with generic sky
conditions. Where not specified, quantity in input or output must be intended as a spatial field for any instant of simulation time. “Measured”
efers to a quantity that is measured at a meteorological station. The components, besides the specified files received in input, include an
appropriate set of parameter values.
Radiation
R. Rigon
Formettaetal.,ModelingshortwavesolarradiationusingtheJGrass-NewAge
system,2013
65. !65
G. Formetta et al.: Modeling shortwave solar radiation using the JGrass-NewAge system
Fig. 2. OMS3 SWRB components of JGrass-NewAge and flowchart
for automatic Jack-Knife procedure. The Jack-knife component
(which is not used in the present paper) simply needs to be added to
the basic model solution, and actually just substitutes the the Verifi-
OMS3 SWRB components of JGrass-NewAge and flowchart for automatic jackknife procedure. The jackknife component (which
sed in the present paper) simply needs to be added to the basic model solution, and actually just substitutes the the verification
ent in Fig. 1
R. Rigon
Radiation - II
Formettaetal.,ModelingshortwavesolarradiationusingtheJGrass-NewAge
system,2013
66. !66
G. Formetta et al.: Modeling shortwave solar radiation using the JGrass-NewAge system 923
Fig. 5. The Fort Cobb river basin, Oklahoma (USA). riangles repre-
sent the verification set (V-set) and circles represent the calibration
set (C-set). The comparison between measured and modeled incom-
ing solar radiation is represented in term of scatter plots.
Fig. 5. The Fort Cobb Reservoir basin, Oklahoma (USA). Triangles represent the V-set and circles represent the C-set. The comparison
between measured and modeled incoming solar radiation is represented with scatter plots.
in which R represents the linear correlation coefficient
between the S and O values, A and B are, respectively
expressed in Eqs. (33) and (34):
A =
o
s
, (33)
point of the Piave River basin. In order to perform this ap-
plication it was necessary to interpolate the air temperature
and relative humidity measurement data for each pixel of the
basins by using a detrended kriging component. The simula-
tion time step was hourly and the simulation period was one
day: from 1 January 2010 to 1 February 2010.R. Rigon
Radiation: test at stations
after Formetta et al., GMD, 2013
67. !67
G. Formetta et al.: Modeling shortwave solar radiation using the JGrass-NewAge system
Fig. 7. The Figure represents the global shortwave radiation on the
Piave area the first october 2010, at four different hours of the day.
During the day differently oriented hillslope received the maximum
amount of radiation and, at 4 p.m. most of the area is covered by
shadows.
Fig. 7. The figure represents the global shortwave radiation on the
Piave area on 1 October 2010 at four different hours of the day.
During the day, differently oriented hillslope received the maximum
amount of radiation and, at 16:00 LT most of the area is covered by
shadows.
also due to the lower measure
elevation zones.
Because of this topographic
surement data uncertainty of t
influenced the atmospheric tran
is confirmed also by the data
basin measurements show lowe
example, the correlation betwe
Washita River basin, where the
play a crucial role.
Regardless, the model was
shortwave solar radiation also
pography. The PBIAS index w
case. According the hydrologic
on PBIAS index, presented in
Stehr et al. (2008), the results a
sified as “good” and therefore
suitable to be used for the estim
solar radiation.
Finally, Fig. 7 presents the r
model. Maps of incoming sola
four hours during the daytime.
pographic feature of the Piave
radiation maps. Their patterns c
cording to the solar position, th
shadow.
5.2 About the possibilities
based JGrass-NewAGE
Since the goal of the paper wasR. Rigon
Radiation: daily total radiation over an area
Formettaetal.,ModelingshortwavesolarradiationusingtheJGrass-NewAge
system,2013
69. !69Fig. 6. The workflow for the Fort Cobb river basin application.
G. Formetta et al. / Environmental Modelling & Software 55 (2014) 190e200
Forcings and Calibration
R. Rigon
Formettaetal.,Hydrologicalmodellingwithcomponents:AGIS-basedopen-source
framework,2014
70. !70
Hillslope Storage
Dynamics
Surface flows
Aggregation
Channel flow
Next time step
JGrass-NewAge
(Formetta et al., GTD, 2011
Evapotranspiration
Radiation
Including snow
Including snow (with various models)
R. Rigon
Formetta et al., Snow water equivalent modeling components in NewAge-JGrass, 2014
71. !71
G. Formetta et al.: Snow water equivalent modeling component in NewAge-JGrass
Fig. 2. The SWE-C integration into the NewAge system, showing connections with the shortwave radiation component and kriging inte
ation algorithm. The connection with the particle swarm optimization algorithm is presented as a red dashed line.
Table 1. Meteorological stations used in test simulations for the
Formetta et al.: Snow water equivalent modeling component in NewAge-JGrass
R. Rigon
Including snow (with various models)
Formetta et al., Snow water equivalent modeling components in NewAge-JGrass, 2014
72. !72
G. Formetta et al.: Snow water equivalent modeling component in NewAge-JGrass
Formetta et al.: Snow water equivalent modeling component in NewAge-JGrass 15
Figure 8. SWE-C application in distributed mode: snow water
equivalent maps from 1 November to 1 June for the Upper Cache la
Poudre basin.
www.jn.net J. Name
Fig. 8. SWE-C application in distributed mode: snow water equiva-
lent maps from 1 November to 1 June for the upper Cache la Poudre
Basin.
and temperature maps was relatively small, and further work
temperature only or bo
The model is integrated
ical model as an OMS3
can make use of all the
GIS-based visualization,
evaluation packages. All
verified at three SNOTE
Poudre River basin (Co
forms well for both dai
model performance degr
uation periods. This is
step compared to an hou
that both the degree-day
els are very sensitive to
they have to be evaluated
dividual sites but also fo
time and space.
Using an hourly time
degradation when movin
tion period. Therefore, a
could be to adopt a time
bin et al. (2013).
Finally, the model is
ulate spatial patterns of
snow water equivalent pR. Rigon
Including snow (with various models)
Formetta et al., Snow water equivalent modeling components in NewAge-JGrass, 2014
73. !73
1995 1996
1997 1998
1999 2000
0
500
1000
1500
0
500
1000
1500
2000
2500
0
500
1000
1500
0
500
1000
1500
2000
0
500
1000
1500
2000
0
1000
2000
3000
Jan 1995 Apr 1995 Jul 1995 Oct 1995 Jan 1996 Jan 1996 Apr 1996 Jul 1996 Oct 1996 Jan 1997
Jan 1997 Apr 1997 Jul 1997 Oct 1997 Jan 1998 Jan 1998 Apr 1998 Jul 1998 Oct 1998 Jan 1999
Jan 1999 Apr 1999 Jul 1999 Oct 1999 Jan 2000 Jan 2000 Apr 2000 Jul 2000 Oct 2000 Jan 2001
Time
BasinmeanwaterFluxes(mm)
Figure 15: The water budget closure based on the use of all the rainfall, discharge and evapotranspiration estimation at annual time scale. The
plots reports for the first sex years of the analysis. The solid line with red band is the annual cumulative TAW (total available water) which
includes the rainfall input plus the residual storage(P-Q-ET) remains from the last hydrological year, the dash line with blue band is the annual
cumulative ETa, the dot line with black band is annual cumulative discharge at the outlet (Q), and the longdash line with gray band is annual
cumulative OUTFLOW (ETa + Q). The lines are the maximum and the minimum estimation value, while the bands are the ranges of error in the
estimation.
14
The whole budget
R. Rigon
Aberaetal.,inpreparation,2014
74. !74
Observe,
that I did not mention the complexity implied by
the Richards equation.
!
WHERE IS IT NOW ?
Concluding
Toward some conclusion
R. Rigon
75. !75
A rigorous statistical theory would be needed that
allows for
!
•doing rigorously such simplifications*, not just on the basis of the personal Art
of modelling^;
!
•quantify the uncertainty remained after the simplifications**
*for a derivation of part of it see Cordano and Rigon, 2008 and BTW compare it with the abstract view
Reggiani et al., 1999
^Art will remain, anyway ...
** The distribution around the mean quantities could not be sharp. Variances can be important ...
A need of a “statistical theory”
R. Rigon
76. !76
The more “reductionist” GEOtop
!
could be used to test the solutions implemented in the simplified NewAGE and
evaluate the non-acceptable behaviors.
Obviously, this is not as simple as
it can be, because GEOtop itself
comes with its simplifications and
errors
A need of a “statistical theory”
R. Rigon
78. !78
In general, when building our models, we should
have a clear and disenchanted vision of their
limits, a theory for their errors, and the idea of
the measures (if we do not have controlled
experiments) to falsified them. The best would
be to have a theory* correlating the information
(of the signal) we need to reproduce with the
complexity of the model needed to get it, so we
do not exaggerate with detailed descriptions of
the (micro-)physics, at finer scales, which are not
required at the larger ones.
Conclusion for physics
R. Rigon
For tentative studies about the relation of modelling and information theory see also: http://
abouthydrology.blogspot.com/2014/07/uncertainty-and-information-theory.html
80. !80
To my former Ph.D. students
R. Rigon
A thank to my Ph.D students: they made it possible
81. !81
Find this presentation at
http://abouthydrology.blogspot.com
Ulrici,2000?
Other material at
Thank you audience !
R. Rigon
http://www.slideshare.net/GEOFRAMEcafe/which-is-the-best-model