Observing the Earth Sensors, Data, and Models

1
New Information Technologies for Large-Scale
Water Resource Management
Observing the Earth Sensors, Data, and Models

• New Technologies
• Distributed watershed modeling
• Advanced remote and in-situ sensing
• Intelligent sensor networks
• Improved methods for integrating models data

• Water Resource Applications
• Resource assessment
• Flood forecasting
• Better information for design planning

• Real-time operation
• Better understanding of human impacts

2
Thailand National Water Resources Summary
Land
510,000 km2 Sources FAO, USA
Arable (irrigated 50,000) Permanent
crops Forest Other
175,000 km2 40,000 132,000 163,000
3
The Role of Variability
TLDD
TMD
Source Oki et al. (2001)
4
Assessment of Current and Future Conditions
Total water available appears sufficient for
future needs, especially if industrial efficiency
increases and climate does not change dramatically
But .. Water in Thailand is distributed
very unevenly over time and space, causing
recurrent shortages and floods
Management options
Structural Expand surface/subsurface storage,
conveyance facilities, irrigation infrastructure,
etc. Policy Operational Introduce water
pricing/trading, move to real-time operations,
improve efficiency, etc.
These options can be expensive and/or difficult
to implement, with uncertain results
We need to consider small-scale variability to
achieve best management of national water
resources
Spatially distributed models data
Use technology to develop better understanding
and to support planning/operations.
5
Using Advanced Technology to Improve Management
High resolution integrated water resource
modeling Develop a detailed characterization of
the water resource system. Where does water come
from? Where does it go? What is the best way to
increase supply? Use new data sources.
Advanced models and data sources provide the
means to systematically evaluate management
options . Some examples

• Optimal management of existing facilities
• Account for inflow demand uncertainty in
  design operations. Balance different
  objectives. Apply real-time control concepts.

• Planning new conveyance irrigation projects
• Incorporate proposed designs into integrated
  modeling system. Evaluate project capabilities
  and economic implications.

• Conjunctive use of surface groundwater
  resources
• Use models to evaluate technical economic
  feasibility of temporarily storing excess water
  in natural aquifers. Operate in real-time.

6
Processes Affecting Runoff and Evapotranspiration

• Soil moisture topography control partitioning
  of precipitation into runoff and infiltration
• Infiltration, evapotranspiration, groundwater
  depth control soil moisture
• Soil moisture, soil temperature vegetation
  control evpaotranspiration sensible heat
• Solar radiation, sensible heat
  evapotranspiration control soil temperature

We seek quantitative descriptions of these
processes
7
Distributed Watershed Modeling

• Approach
• Approach Apply mass balance, momentum, energy
  conservation principles at fine scales that
  capture important natural variability
• Provide additional information (constitutive
  relationships) to account for interactions

Inputs required
Catchment-based grid (Koster et al.)
Precipitation Micro-meteorology Land
use/vegetation
Topography Soil composition Depth-to-groundwater
Natural scales for discretization
Pixel/raster-based grid
Triangulated irregular network (Vivoni et al.)
Grid that captures full range of scales
High computational and data requirements
8
How Much Resolution Do We Need?
The critical role of topography, vegetation,
soil
Depth to groundwater
Evapotranspiration
Runoff generation
Modeling results (Entekhabi, 2005)
Runoff and ET can be governed by processes acting
over 100s m. and 10s of minutes (Ivanov, 2005)
Incoming shortwave radiation
Meteorological variability
9
Averaging/Upscaling
Can we derive average runoff and
evapotranspiration from average precipitation ?
Infitration capacity decreases/increases as soil
gets wetter/drier
Runoff occurs when precipitation exceeds
infitration capacity
15 minute precip/capacity values generate runoff
but daily values do not
In order to obtain correct average runoff from
average precipitation, we need to adjust
(upscale) equations used to describe
infiltration. This process is often empirical
unreliable for evaluating new projects policies.
It is best to use resolution compatible with
scale at which runoff-generating processes occur.
10
Sources of Data for Distributed Modeling
Classical Situation No comprehensive
measurements of hydrologic states fluxes

• Precipitation - Scattered observations at gages
• Evapotranspiration No direct measurements, very
  limited indirect measurements
• Runoff - Scattered stream flow measurements at
  gages
• Recharge/GW flow - No direct measurements, very
  limited indirect measurements
• Soil canopy moisture/temperature Very limited
  point measurements
• Groundwater storage - Only indirect meas
  (depths/heads) at scattered locations

The Future Remote sensing sensor networks

• High resolution topography from airborne
  satellite measurements
• Much improved coverage of precipitation from
  remote sensing instruments (NEXRAD, TRMM, GPM)
• Much improved information on land use/vegetation,
  soil moisture (airborne and satellite microwave)
• Possible quantification of large-scale GW storage
  change (GRACE)
• Inexpensive sensor networks for micro-meterology,
  soil moisture, runoff

It should soon be possible to estimate
unobservable hydrologic fluxes at high space/time
resolution, by combining remote sensing data with
distributed models
11
Example Estimation of Hydrologic Fluxes over
Great Plains
Estimate precipitation,soil moisture,
evapotranpiration over Great Plains, Summer 2004,
Precipitation land surface (NCAR CLM) models
Dynamic, 106 states, 2500 steps
Regional/global data sources used for estimation
Soils (STATSGO) Vegetation (USGS) Meteorological
variables (NCEP)
Precipitation (GOES, TRMM, SSMI, AMSU) L-band
microwave (HYDROS, passive active)
Objective is to determine how land surface fluxes
vary over time and space, in response to
meteorologic forcing
12
Example Satellite-based Precipitation Data
Sources
GOES Geostationary, continuous observations of
cloud top temperature 0.05 degree (4 km), 1 hr
SSMI Polar, passive microwave measurements,
related to precipitation SSMI 0.25 degree (20
km), 2/day for one location
TRMM Polar, passive and active microwave
measurements related to precipitation, in lower
half of domain 0.05 degree (5 km), 2/ day for
one location
AMSU Polar, passive microwave measurements in
several bands, related to precipitation 0.15
degree, ( 15km), 2/day for one location
NOWRAD weather radar measurements are used as
ground truth for precipitation estimation 0.05
degree (4km), 15 min. (average to get mm/hr)
13
Example Satellite-based Soil Moisture Data
Sources
HYDROS radiometer (passive) L-band, polar,
coarse resolution 0.5 deg (40km),
ºK (synthesized from model, satellite not yet
launched)
HYDROS radar (active) L-band, polar, fine
resolution 0.05 deg ( 5km), dB (synthesized from
model, satellite not yet launched)
Data gaps occur in areas with dense vegetation.
Data available daily at 1200AM UTC
14
Small Soil Moisture Temperature Sensors
These inexpensive sensors can be used in wireless
networks, to provide better information on
spatial variability and to check remote sensing
measurements. Distribute sensors in
representative watersheds.
15
Designing Wireless Sensor Networks
Hypothetical sensor layout for the Little Washita
watershed, showing sensors and base stations.
Wireless relays used on base station links.
Sensor locations selected to maximize information
about space/time variability within the watershed.
Sensor network design needs to trade off
Reliability ? Cost ? Resolution ?
Coverage ? Power/Lifetime
16
Some Important Sensor Network Isuues

• Can we design networks to automatically provide
  maxiumum information for minimum power/cost?
• Advantages of wireless vs. wired networks/data
  logging
• Network scales Coverage, spacing, tradeoffs
  between resolution, transmitter range, and
  power/size/cost
• Distributed computing as well as sensing Can
  network decide where when to take data, where
  to send data, how to deal with failures?
• Robustness, security, maintenance, lifetime?
• Integration with remote sensing?

Inexpensive wireless networks can be unreliable.
Redundancy, automatic recalibration, failure
detection, adapativity must be built in
17
Combining Data Models Data Assimilation
Models data are imperfect incomplete
Models Limited resolution, inadequate
characterization or understanding of processes,
input data inadequate
Data Limited coverage and/or resolution in space
time. Data gaps, recording calibration
errors, instrument failures, observations may
only provide indirect measurements of hydrologic
variables
Can we improve our characterization/predictions
of hydrologic conditions by combining model
predictions data?
Data Assimilation

• Key steps in ensemble data assimilation
• Relate all available measurements (different
  instruments, scales, accuracies) to a common set
  of hydrologic variables, defined on a convenient
  computational grid.

This requires probabilitic models of hydrologic
system and sensors (probabilistic to account for
errors).

• Find a collection (ensemble) of model predictions
  that is compatible with actual measurements and
  respects physical constraints. This ensemble
  defines a range of likely conditions.

18
(No Transcript)
19
Some Ensemble Examples - 2
Representative soil moisture and ET replicates
from randomly perturbed precipitation/soil/vegetat
ion data at the SGP site, Oklahoma, USA.
20
Great Plains Example Generating Realistic
Rainfall Replicates
1-8 June 2004, 106 states, EnKF, no localization
NOWRAD weather radar precipitation -- ground truth
Ensemble mean precipitation
Typical replicate from precipitation ensemble
Ensemble Kalman filter precip.,conditioned on
GOES TRMM SSMI AMSU NOWRAD data
withheld
21
Great Plains Example Estimating Soil Moisture
with a Multiscale Ensemble Kalman Filter
1-8 June 2004, 106 states , multiscale EnKF, no
localization
NOWRAD weather radar precipitation -- ground truth
Surface soil moisture from NOWRAD
Mean of conditional surface soil moisture ensemble
22
Great Plains Example Estimating
Evapotranspiration with a Multiscale Ensemble
Kalman Filter
1-8 June 2004 106 states, multiscale EnKF, no
localization
NOWRAD weather radar precipitation -- ground truth
Evapotranspiration from NOWRAD
Mean of conditional evapotranspiration ensemble
23
Suggested IT Research Topics/Challenges Thai
Water Resource Management
24
Management Implications