530 likes | 729 Views
Remote Sensing, Land Surface Modelling and Data Assimilation. Christoph Rüdiger, Jeffrey Walker The University of Melbourne Jetse Kalma The University of Newcastle Garry Willgoose The university of Leeds. Overview. Remote Sensing Data Assimilation Land Surface Modelling
E N D
Remote Sensing, Land Surface Modelling and Data Assimilation Christoph Rüdiger, Jeffrey Walker The University of Melbourne Jetse Kalma The University of Newcastle Garry Willgoose The university of Leeds
Overview • Remote Sensing • Data Assimilation • Land Surface Modelling • Combining the Options
Remote Sensing • Remote Sensing defined: • Measurement of energy reflections or emissions of different spectra from a distance • Modes of Remote Sensing in Hydrology: • Ground-based • air-borne • space-borne platforms
Observed Wavelengths- Spectral Resolution - • Visual Band (~400nm – 700nm) • Infrared Band (~0.7μm – 1000μm) • Microwave Band (~1cm – 30cm) • Radio Band (>30cm) • Gravitational Measurements
What Can Be Measured(some examples) • Subsurface • Surface soil moisture, soil temperature, gravitational effects • Surface • Vegetation cover, vegetation density, evapotranspiration, temperature, sea level, elevation, fires • Atmosphere • Cloud cover, aerosols, wind, temperature
Remote Sensing- Spatial Resolution - Study Catchment
Current Missions • Visual Band (~400nm – 700nm) • Modis, Landsat … • Infrared Band (~0.7μm – 1000μm) • Landsat, GOES • Microwave Band (~1cm – 30cm) • TRMM, AMSR-E • Radio Band (>30cm) • TRMM • Gravitational Measurements • Grace
Limitations of Individual Bands • Atmospheric interference (infrared). • Radio interference (microwave). • Surface conditions, vegetation, cloud and aerosol effects (all). • Penetration depth (all). • Other effects? Rüdiger et al., in review
Summary of Remote Sensing • Advantages: • Observation of large areas • Observations of remote areas • Large quantity of environmental states can be observed • Limitations: • Either low resolution or low rate of repeat overpasses • Influence of surface and atmospheric conditions have to be filtered • Average values of observed states, need for downscaling
Data Assimilation Defined Definition 1: using data to force a model ie. precipitation and evapotranspiration to force a LSM Analogy: passenger giving instructions to a blindfolded driver on the autostrada at peak hour
Analogy 1 Initial state Update Update Update Update Update Update
Data Assimilation Defined Definition 1: using data to force a model ie. precipitation and evapotranspiration to force a LSM Analogy: passenger giving instructions to a blindfolded driver on the autostrada at peak hour Definition 2: using state observations to make a correction to the forecast model state ie. surface soil moisture obs. to correct forecasts Analogy: driver can see through his blindfold for 1/10th second every 30 seconds
Initial state Analogy 2 Avail. Info Avail. Info Forecast Forecast Avail. Info Forecast
What is the Usefulness of Data Assimilation • Organises data (model acts as interpolator) • Complements data (fills in unobserved regions) • Supplments data (provides unobserved quantities) • Quality controls data • Calibrates data
Some Methods of Data Assimilation • Direct Insertion • Statistical Correction • Optimal Interpolation (OI) • Variational over Space and Time (4DVAR) • Sequential Data Assimilation (eg. Kalman Filter)
Window 1 Window 2 State Value Time Continuous or Sequential DA? • Continuous (ie. variational) • Regression schemes • Adjoint derivation • In general: • Minimisation of objective function
State Value Time Continuous or Sequential DA? • Sequential (ie. Kalman filter) Predict: Observe: Correct:
EKF EnKF Covariance Covariance State Value State Value Time Time Extended or Ensemble KF?
DA as a Spatial Interpolator Soil Moisture Soil Moisture Houser et al., WRR 1998
Summary of Data Assimilation • Advantages • Variational: • Computationally inexpensive • Does not need prior knowledge of system states or errors • No linearisation of model needed • Can obtain model sensitivity values • Sequential: • Update of states at every observation point • Model size depends on computer not mathematics • Advantage over variational schemes for distributed models
Summary of Data Assimilation • Limitations • Variational: • Regression scheme can become unstable • Adjoint derivation can be a complex problem • Long-term forecasts become inaccurate • Sequential: • Models have to be linearised to certain extent • Can be computationally infeasible • Error estimates can cause problems
Hydrological Modelling • Different models available • Soil moisture models • Land surface models • Atmospheric models • Land surface – atmosphere models • General Circulation models • Different approaches for modelling: • Lumped • Distributed • Semi-distributed
Difference between model approaches distributed semi-distributed or lumped
Semi-distributed model Kalma et al., 1995
Two Models Koster et al., 2000 Liang et al., 1998
Importance of Land Surface States(soil moisture, soil temperature, snow) Drought monitoring Irrigation policies Flood prediction Weather forecasting
Importance of Land Surface States(soil moisture, soil temperature, snow) • Early warning systems • Flood prediction – infiltration, snow melt • Socio-economic activities • Agriculture – yield forecasting, management (pesticides etc), sediment transport • Water management – irrigation • Policy planning • Drought relief • Global change • Weather and climate • Evapotranspiration – latent and sensible heat • Albedo
Soil Moisture vs Sea Surface Temp • Knowledge of soil moisture has a greater impact on the predictability of summertime precipitation over land at mid-latitudes than Sea Surface Temperature (SST). Koster et al., JHM 2000
(JJA) Importance of Soil Moisture Koster et al., JHM 2000
R e m o t e S e n s i n g S a t e l l i t e S u r f a c e S o i l M o i s t u r e L o g g e r S o i l M o i s t u r e M o d e l f (z) [ q , D ( ) , ( ) ] s S o i l M o i s t u r e S e n s o r s The Situation
The Problem With LSMs • Same forcing and initial conditions but different predictions of soil moisture! Houser et al., GEWEX NEWS 2001
Why do we need improvement? Koster et al., JHM, 2000
Case Study – Variational DAAssimilation of Streamflow and Surface Soil Moisture Observations
Bayesian Regression Kuczera, 1982
Results “Experiment 1” Discharge Soil Moisture
Results “Experiment 2” Discharge Soil Moisture
Results Experiment 2 cont’d Root Zone Soil Moisture Surface Soil Moisture
Summary of Variational Approach • Retrieval of initial states possible to high accuracy. • Only few iterations necessary. • Limitations when additional errors are involved. • Long forecasting window will lead to less accurate results. • First estimate of initial states can be important
Case Study – Sequential DAAssimilation of Surface Soil Moisture
Single Observation Number of Observations • All observations
Summary of Sequential DA • Require a statistical assimilation scheme (ie. a scheme which can potentially alter the entire profile). • Simulation results may be degraded slightly if simulation and observation values are already close. • The updating interval is relatively unimportant when using a calibrated model with accurate forcing.