Transcript DAUFIN Data assimilation within a unifying modeling

Observing catchments, rivers and
wetlands from space: assimilating
hydrologic information into distributed
models
Peter A. Troch
Outline of presentation
• Data needs for (surface) water resources management
• Satellite based observations of rivers and wetlands
• Satellite based observations of soil moisture and latent
heat fluxes
• Satellite based observations of rainfall
• Satellite based observations of catchment storage
changes
• Data assimilation into distributed models
• Recommendations/conclusions
Data needs for Water Resources Management
Sound water resources management is hampered by uncertainties in quantifying
the water balance components at the catchment scale.
Water balance components at the catchment scale are traditionally estimated
by means of in-situ measurements and distributed hydrological models.
A wide variety of distributed hydrological models has been developed over the past
decade. A major problem plaguing distributed modelling is parameter identifiability,
owing to a mismatch between model complexity and the level of data which is
traditionally available to parametrize, initialize, and calibrate models, and to uncertainty and error in both models and observational data.
New data sources for observation of hydrological processes (ENVISAT, MSG, SMOS)
can alleviate some of the problems facing the validation and operational use
of hydrological models.
Data assimilation provides a means of integrating these data in a consistent manner
with model predictions.
Lack of Q?
Lack of Q and S Measurements: An example from Inundated Amazon Floodplain
Singular gauges are incapable of measuring
the flow conditions and related storage
changes in these photos whereas complete
gauge networks are cost prohibitive. The
ideal solution is a spatial measurement of
water heights from a remote platform.
How does water flow through these
environments?
100% Inundated!
(L. Mertes, L. Hess photos)
Example: Braided Rivers
It is impossible to measure discharge along these
Arctic braided rivers with a single gauging station.
Like the Amazon floodplain, a network of gauges
located throughout a braided river reach is
impractical. Instead, a spatial measurement of
flow from a remote platform is preferred.
Resulting Science Questions
– How does this lack of measurements limit our ability
to predict the land surface branch of the global
hydrologic cycle?
• Stream flow is the spatial and temporal integrator
of hydrological processes thus is used to verify
predicted surface water balances.
• Unfortunately, model runoff predictions often do
not agree with observed stream flow during
validation runs.
Solutions from Radar Altimetry
Topex/POSEIDON tracks crossing the
Amazon Basin. Circles indicate locations
of water level changes measured by T/P
radar altimetry over rivers and wetlands.
Presently, altimeters are configured for
oceanographic applications, thus lacking
the spatial resolution that may be
possible for rivers and wetlands.
Water surface heights, relative to a
common datum, derived from
Topex/POSEIDON radar altimetry.
Accuracy of each height is about the
size of the symbol.
Solutions from Interferometric SAR for Water Level Changes
0 km 20
JERS-1 Interferogram spanning February 14 – March 30, 1997. “A” marks
locations of T/P altimetry profile. Water level changes across an entire lake have
been measured (i.e., the yellow marks the lake surface, blue indicates land).
BUT, method requires inundated vegetation for “double-bounce” travel path of
radar pulse.
These water level
changes, 12 +/- 2
cm, agree with
T/P, 21 +/- 10++
cm.
River Velocity & Width & Slope Measurements
Concept by Ernesto Rodriguez of JPL
Measure -Doppler Velocity
Measure Topography
Measure +Doppler Velocity
Example of measurement of the
radial component of surface velocity
using along-track interferometry
Basic configuration of the satellite
Global Wetlands
• Wetlands are distributed
globally, ~4% of Earth’s land
surface
• Current knowledge of wetlands
extent is inadequate
Saturated extent from RADARSAT - Putuligayuk River,
Alaska
2
0
0
0
Inundated area (km 2)
= wet
= dry
400
300
200
1999
2000
a.
100
0
6/10
b.
c.
6/30
d.
7/20
e.
8/9
8/29
Variable source areas detected from ERS-1/2
Verhoest et al. (1998)
Hillslope-storage dynamics
European contribution to GPM (SRON)
•
•
1 core satellite (dual frequency 13.6 / 35 GHz
imaging pulsed radar, TMI-like radiometer)
8 constellation satellites (passive microwave
radiometers)
Potential GPM Validation Sites
GPM
Canada
Eng land
NASA Land
Spain
Ger many
Italy
South Ko rea
ARM/UOK
Japan
NASA KS C
Taiwan
Fr ance (Niger & Benin)
India
NASA Ocean
Brazil
Australia
Supersite
Eric A. Smith
Regional Raingage Site
Janu ary 28 , 2002
Supersite & Regional Raingage Site
3
River basin storage changes through gravity
• GRACE: Gravity Recovery and Climate Experiment
– Schatten van bergingsverandering in grote stroomgebieden
– Horton Research Grant (AGU) AIO onderzoek
Sensitivity of gravity changes to water storage
changes
Time (days)
1 Gal = 9,807 m/s2
Existing Instruments
• Water Surface Area:
– Low Spatial/High Temporal:
Passive Microwave (SSM/I,
SMMR), MODIS
– High Spatial/Low Temporal: JERS1, ERS 1/2 & EnviSat, RadarSat,
LandSat
• Water Surface Heights:
– Low Vertical & Spatial, High
Temporal (> 10 cm accuracy, 200+
km track spacing):
Topex/POSEIDON
– High Vertical & Spatial, Low
Temporal (180-day repeat): ICESat
• Water Volumes:
– Very Low Spatial, Low Temporal:
GRACE
– High Spatial, Low Temporal:
Interferometric SAR (JERS-1,
ALOS, SIR-C)
• Topography:
– SRTM (also provides some
information on water slopes)
Motivation for Data Assimilation
Continued progress in our scientific understanding of hydrological processes at the
regional scale relies on making the best possible use of advanced simulation models
and the large amount of environmental data that are increasingly being made available.
The objective of data assimilation is to provide physically consistent estimates of spatially distributed environmental variables.
Geophysical data assimilation is a quantitative, objective method to infer the state of
the land-atmosphere-ocean system from heterogeneous, irregularly distributed, and
temporally inconsistent observational data with differing accuracies, providing at the
same time more reliable information about prediction uncertainty in model forecasts.
Data assimilation is used operationally in oceanography and meteorology, but in
hydrology it is only recently that international research activities have been deployed.
Data assimilation of remote sensing observations
45
y = 1.0746x
R2 = 0.802
40
ETact SEBAL
35
(26)
(31)
(28)
(27)
30
25
(32)
20
(18)
15
10
(38)
(22)
(13)
(23)
5
(14)
(30)
0
0
5
10
15
20
25
30
ETact SIMGRO
35
40
45
Data assimilation of remote sensing observations
• Rivierenland-project (ICES-KIS3)
– Soil moisture measurements and scintillometer to validate RS
Open Research Issues (1)
Remote sensing technology provides many types of data that are related to land
surface variables of interest to hydrologists. However, very little of this information
is available in a form that can be used directly for hydrological purposes.
Data assimilation research in hydrology should focus on producing data products
that are directly useful for water management. Such products need to be carefully
designed to meet the needs of potential users:
• resolution and spatial configuration of data products;
• quantitative measures of data product reliability;
• quality control issues;
• sensitivity of data products to “hidden” model properties.
There is a need to bridge the gap between continental scale data sets (GLDAS)
and catchment scale applications (downscaling and parameterization issues).
Open Research Issues (2)
Classical hydrological models that have been optimized for use with sparse in situ
observations are inadequate for extension to work with remote sensing data. There
is a need for developing more appropriate distributed models at catchment scale.
More research is needed to develop data assimilation algorithms that can handle
the specific problems encountered in hydrological applications:
• subsurface processes are hard to “observe”;
• high degree of heterogeneity of physical system;
• hydrologic systems function over a wide range of temporal scales.
Geostatistical techniques for describing multi-scale spatial heterogeneity need to be
incorporated into algorithms that account for the multi-resolution nature of different
but complementary hydrologic measurements.
Case studies are needed to introduce and demonstrate the potential of data
assimilation in operational water resources management (e.g. improved flood
predictions).
What is needed?
Continued investment and coordination of data assimilation initiatives at the
European level is urged:
• wide range of research topics relevant to data assimilation;
• strong need for innovation in each of these areas;
• clear potential for water resources modelling and management;
• transboundary nature of catchments and river basins;
• need for common algorithms, models, tools, data standards, etc.
• leading role already demonstrated by European researchers.
Expertise from many disciplines will be needed to meet the challenge of data
assimilation for improved river basin water resources management:
• hydrology
• meteorology
• remote sensing
• ecology
• mathematics (systems theory, statistics)
• information technology
• water management
• etc.