Improving Multi-Scale Root Zone Soil Water Process Representation

Improving Multi-Scale Root Zone Soil Water Process Representation

Partly supported by:
Improving Multi-Scale Root Zone Soil Water
Process Representation in Land Surface Models
Binayak P. Mohanty
& Vadose Zone Group
Texas A&M University
Texas Drought Forum
Oct 22, 2012
http://vadosezone.tamu.edu/
Dr. Binayak P. Mohanty
Dr. Fabio Sartori
David Hansen
Sean Tolle
Dr. Joseph Pollacco
Raghavendra Jana
Yongchul Shin
Dr. Itza Mendoza-Sanchez
Champa Joshi
Nandita Gaur
Bhavna Arora
Taehoon Kwak
Dr. Zhongyi Qu
Dipankar Dwivedi
Sandeep Patil
Soil Moisture Flow Below our Feet!
Gopher Holes
Structural Cracks
Biological Pores
Karst Geology
Soil Water Flow Processes
Evapotranspiration
Run-on
Run-off
Soil moisture
0-5 cm
observations
Heterogeneity of soil
profile
GW -100 cm
Infiltration
Upward flux
GW -150 cm
Percolation
GW -200 cm
Free drainage
Soil Moisture at Different Process Scales
SCALE
Global
Regional
Watershed
Field
Pore
Soil Moisture / Brightness Temperature
Measurement Scales
Space-borne
SMOS
SMAP
Air-borne
ESTAR
PSR
Ground-based
TOWER
TRUCK
Insitu
TDR
Gravimetric
Upcoming Soil Moisture Satellite Missions
SMOS, 2010, ESA………SMAP, 2013, NASA
scattered
Spatial Data
Temporal Data
Spectral Reflectance
Percent Reflectance
80
Vegetation
70
Soil
60
Clear River Water
Turbid River Water
50
40
30
20
10
0
0.4
0.6
0.8
1.0 1.2
1.4
1.6
1.8
Wavelength (µm)
2.0
2.2
2.4
2.6
Remote Sensing of Soil Moisture…
• TB for homogenous vegetated surface:
• TB=TS{es exp(-τc) + (1-ω)[1-exp (-τc)][1+rs exp(-τc)]}
• τc=bwc/cosθ
• soil and vegetation TS are assumed to be equal
• τc vegetation opacity
• ω vegetation single scattering albedo
• b coefficient depends of frequency and
vegetation type
• wc vegetation water content
• θ incidence angle
• es emissivity of soil
Remote Sensing of Soil Moisture…
• Reflectivity of a rough soil surface:
• rs=[(1-Q)ro + Qro)] exp(-h)
• Q,h are roughness parameter
• ro reflectivity of a smooth surface
Remote Sensing Cal/Val Studies for Soil Moisture
Synchronized Sampling
at Multiple Scales
© NASA
Soil Moisture Remote Sensing
Space-Borne
NASA: AMSR-E on AQUA
6
7!
!
!
Air-Borne
NOAA: PSR
4
!
5
AMSR-E (JAN 15 2004)
Soil Moisture cm3/cm3
0 - 0.075
0.075 - 0.105
0.105 - 0.135
0.135 - 0.165
0.165 - 0.230
0.230 - 0.500
3!
!
! 1
2
Spatio-Temporal Data in Iowa
Global Coverage every 1.5 days
U.S. Watershed Soil Moisture Validation Sites
•
•
•
All sites
include 5 cm
LW, LR, WG,
RC since
2002
WC, FC are
new
EOF Analyses Across Scales in Iowa During SMEX02
Field-scale
Watershedscale
Regionalscale
Soil Moisture Dominant Physical Controls at
Different Spatial Scales
Region
Watershed
Field
Vegetation/
Precipitation
Topography
Soil
Hypothesis
• Soil moisture variability is dominated by
Coarsening
Scale
Modeling Soil Moisture
• Importance of modeling – RS of soil moisture only
limited to near-surface soil layers
• 1D Flow equation: (Assumption: work at any scale!!)

 h  
  K(h)   1 
(h)
h
 z  

 C (h)

 S(h)
t
t
z
S(h)   w (h)
Tpot
zr
Soil Hydraulic Properties
• Soil-moisture retention property, θ(h)
• Soil hydraulic conductivity property, K(h)
θ(h)
h
K(h)
K
θ
h
Modeled Soil Moisture States under
Different Topographic Configurations
Flow Routing at Watershed Scale
Stream Flow Comparison
SG442
SG447
21
Motivation
To address the connections between the
environmental factors and “effective” soil
hydraulic parameters at various scales across
the land surface
To develop scaling (downscaling and
upscaling) algorithms, data assimilation,
inverse model, and incorporation of
stochastic evolutionary schemes
Limitation: What is Pixel-Scale Parameter!?
Vegetation
Soil
Slope
Top-down versus Bottom-up Approach
New Study
Hydraulic Property Estimation
Across the Pixel :
Top- Down Approach
Traditional Soil Physics
Hydraulic Property Measurement
Across the Pixel :
Bottom- Up Approach
Soil Hydraulic Function Scaling Hypothesis
©NASA
h
Using the information
content of the soil
moisture data
collected at that
particular SCALE, we
can estimate the scale
dependent soil
hydraulic properties
RS
footprint scale
1
λ 
K(h)=K sat Se 1- 1-Se m





θ
2
h

Scale2
Point scale
Scale1
θ
Top-down approach
θ(h)-θ res  1 
Se =
=

n
θsat -θ res 1+ αh 
Bottom-up approach
h
θ
m
m
Scale3
Remote sensing data
• Airborne: ESTAR (SGP97)
LW03 Soil moisture, cm3/cm3
DOY
LW13 Soil moisture, cm3/cm3
LW21 Soil moisture, cm3/cm3
Day of year
UNSODA
NMCGA
-1SD
Observed
100000
100000
10000
10000
CL
10000
1000
CL
1000
1000
100
100
100
h, -cm
10
1
10
10
(b)
WC12
(a)
WC11
L
L
1
Ave. CL+L
0.1
1
0
Ave. CL+L
0.1
0.1
0.2
0.3
0.4
0.5
0
0.1
0.2
0.3
0.4
0.5
100000
100000
0.1
0
10000
0.1
1000
0.2
10000
0.3
0.4
1000
SiCL
0.5
SiCL
100
100
CL
CL
10
10
h, -cm
Air-borne RS scale (Polarimetric Scanning
Radiometer: PSR)
100000
+1SD
(c)
WC13
1
(d)
WC14
1
L
Ave. CL+L+SiCL
L
Ave. CL+L+SiCL
0.1
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0
θ(h), cm3 cm-3
0.1
0.2
0.3
0.4
0.5
0.6
0
(mm)
169
174
179
184
189
Simulated vs. Observed
Soil
moisture
0
169
174
179
184
189
194
0
NMCGA
169
174
+1SD
179
-1SD
PSR
184
189
Observed
194
0
40
Rainfall (SCAN)
-40
40
Rainfall (SCAN)
Series 2
-40
80
S
80
-80
-80
0.6
0.6
0.5
0.5
(a)
WC11
0.4
θ(z,t), cm3 cm-3
0
194
0.3
0.3
0.2
0.2
0.1
0.1
0
0
169
0.6
174
Simulated
0.5
179
1+SD
184
1-SD
189
PSR
169
194
Areal_average
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
169
174
Average
179
1-SD
184
189
1+SD
0-5cm
Day
of Year
194
areal_average
174
Average
0.6
0.5
(c)
WC13
0.4
(b)
WC12
0.4
179
1-SD
184
1+SD
189
0-5cm
194
areal_average
(d)
WC14
169
174
Average
179
1-SD
184
1+SD
189
0-5cm
Day of Year
194
areal_average
Objectives
To explore the scale dependency for certain soil
hydraulic parameters used in LSMs:
saturated hydraulic conductivity, residual and
saturation soil water contents, and van Genuchten
parameters
To develop suitable mathematical approaches for
downscaling and upscaling hydarulic parameters in
complex terrains. This proposed approaches would
honor both vertical and horizontal fluxes
Research Approach
Near-surface soil moisture assimilation scheme based
on inverse model
Physically-based (Richard’s equation) hydrological models (Soil
Water Atmosphere Plant-SWAP, Community Land ModelCLM, and Noah land surface model-Noah LSM)
- SWAP: Mualem-van Genuchten (MVG) parameters
- CLM: Clapp and Hornberger empirical equation
- Noah LSM: Clapp and Hornberger empirical equation
Comparison of Layer-Specific and
Near-Surface Assimilation approaches
Layer-specific
Near-surface
, n, res, sat, Ksat)
Soil moisture
Observations
(obsj=1) 0-5 cm
Land surface
1, n1, res1, sat1, Ksat1)
1st: 0-10 cm
(obsj=2) 10-15 cm
2, n2, res2, sat2, Ksat2)
(obsj=3) 60-70 cm
3, n3, res3, sat3, Ksat3)
2nd: 10-60 cm
Soil profile
depth
(Z: 0-200 cm)
3rd: 60-200 cm
Results
Case 1: Comparison of layer-specific/near-surface approach
1 : Clay
loam
This
result
showed
that
the
model
performance
is
1st layer:
0-10 cm
: Sandy loam
affected
by the variability of soil textures and 2layers
in the
soil column.
0.6
nd
0.4
3rd: Silt loam
0.2
0.0
50
0.6
100
150
200
DOY
250
2nd layer: 10-60 cm
300
Obs.
Average (layer-specific)
+PCI (layer-specific)
-PCI (layer-specific)
Average (near-surface)
+PCI (near-surface)
-PCI (near-surface)
350
Near-surface
Layer-specific
0.4
0.2
0.0
1st layer
0
R2: 0.997, MAE: 0.014
R2: 0.981, MAE: 0.013
2nd layer
R2: 0.995, MAE: 0.004
R2: 0.832, MAE: 0.114
3rd layer
R2: 0.998, MAE: 0.003
R2: 0.786, MAE: 0.072
50
100
Soil moisture (cm3 cm-3)
0
Soil moisture (cm3 cm-3)
Soil moisture (cm3 cm-3)
st
150
200
DOY
250
300
3rd layer: 60-200 cm
0.6
350
Near-surface
Layer-specific
0.4
0.2
0.0
0
50
100
150
200
DOY
250
300
350
Spatial scaling algorithms
Evapotranspiration
Run-on
Run-off
Soil moisture
0-5 cm
observations
Heterogeneity of soil
profile
GW -100 cm
Infiltration
Upward flux
GW -150 cm
Percolation
GW -200 cm
Free drainage
Deterministic Disaggregation
algorithm
800 m
Disaggregation process
Remote sensing soil moisture
(RS)
(1,1),ET(1,1)
(2,1),ET(2,1)
(1,2),ET(1,2) (2,2),ET(2,2)
Soil moisture and
ET of individual sub-pixel
Vegetation cover
+
Loamy
sand
Loam
j
Silt
loam
i
Sandy
loam
Soil texture
Results - Soil Moisture
Obs. (in-situ)
Average
+95PCI
-95PCI
49 in-situ SM
sampling points
Soil map
Silt loam (dominant)
Vegetation
Sub 1 Sub 2 Sub 3
Grass Wheat
Sub 4 Sub 5 Sub 6
Sub 7 Sub 8 Sub 9
Soil moisture (cm3 cm-3)
LW 21 site (ESTAR, 800 m  800 m)
Sub-pixel 01
0.5 R2: 0.865, MBE: -0.157
Sub-pixel 02
0.5 R2 : 0.845, MBE: -0.130
Sub-pixel 03
0.5 R2 : 0.433, MBE: -0.136
0.3
0.3
0.3
0.1
0
60
100
140
0.1
0
60
200
Sub-pixel 04
0.5 R2 : 0.555, MBE: -0.165
0.5
0.3
0.3
0.1
0
60
100
140
R2 :
0.1
0
60
200
100
140
0.1
0
60
200
Sub-pixel 05
0.550, MBE: -0.134
0.5
R2 :
100
140
200
Sub-pixel 06
0.802, MBE: -0.122
0.3
100
140
0.1
0
60
200
100
140
200
Sub-pixel 07
0.5 R2 : 0.647, MBE: -0.167
Sub-pixel 08
0.5 R2 : 0.461, MBE: -0.131
Sub-pixel 09
0.5 R2 : 0.343, MBE: -0.131
0.3
0.3
0.3
0.1
0
60
100
140
0.1
0
60
200
100
140
DOY
0.1
0
60
200
100
140
200
Results
Comparison of in-situ and average (for the upscaled sub-pixels)
soil moisture dynamics for SWAP, CLM, and Noah LSM.
Noah LSM-R2: 0.877, RMSE: 0.058
CLM-R2: 0.927, RMSE: 0.051
0
0.7
20
0.6
40
0.5
60
0.4
80
0.3
100
0.2
120
0.1
140
0.0
160
152
162
Precipitation
172
Obs. (in-situ)
182
192
SWAP
202
Noah LSM
212
CLM
In order to support the robustness of our approach, we tested various
36
hydrological models in the upscaling process only.
Precipitation, mm
Soil moisture, cm3 cm-3
SWAP-R2: 0.913, RMSE: 0.029
0.8
Better understanding of
the water cycle
Water resources management
Vertical
heterogeneity
Natural disasters
Evapotranspiration
component
Spatial
variability
Agriculture
Non-parameteric
evolutionary
scheme
Increasing Drought Severity
Drought is one of the most severe environmental outcomes
of the climate change
Texas suffered under an intense drought driven by La Niña
with a total damage of $ 7.6 billion
Drought severity assessment using precipitation data
Agricultural drought influenced by not only weather
conditions, but also by root zone soil moisture
No study of finer-scale drought severity evaluations using
remotely sensed soil moisture
To develop a drought severity assessment
framework using remotely sensed soil moisture
products
- Global circulation model (GCM) based climate forecasts
To evaluate the finer-scale drought severity under
various environmental factors
- Various soil textures, vegetations, soil depths, shallow ground
water tables, etc.
Agricultural Drought Severity Platform
Selection of climatic model
(GCMs, RCMs, etc.)
Remotely sensed soil moisture
(e.g., satellite, air-borne, etc.)
Weather forecast scenarios
Grid-based disaggregation
algorithm
Bias correction of climate
model outputs using the GCM
scenario (1998-2009) and
the historical climatic data
Output variables
(Regional soil characteristics
and sub-grid fractions)
Multiplicative method
x obs
x  xi
x GCM
'
i
Where xi and x’i are raw and corrected
GCM scenarios on day i, and XGCM and
Xobs mean monthly mean weather data
from the GCM and observations
Potential risk assessment
(e.g., Agricultural drought)
Prediction of field-scale root
zone soil moisture
(2010-2020)
Simulation-optimization
scheme (the linkage of SWAP
and EMOGA)
Soil moisture deficit index (SMDI)
SMDI j  0.5  SMDI j1 
Drought severity assessment
(e.g., PDSI, SPI, SMDI, etc.)
Evaluation of various water
management & cropping practices
(e.g., croppig rotations)
SDi, j 
SDi, j 
SWi, j  MSWj
MSWj  min SWj
SD j
50
100,if
SWi, j  MSWj
max SWj  min SWj
100,if
SWi, j  MSWj
SWi, j  MSWj
Providing agricultural decision
support system
Results
Average
SMDI
Wet
4
(e) s1v1 (Sub-pixel 1)
4
pos95PCI
(f) s2v1 (Sub-pixel 2)
4
(g) s3v1 (Sub-pixel 3)
4
2
2
2
0
0
0
0
M
A
M
Dry -4
J
J
-2
-4
M
A
M
J
J
-2
-4
07/01
neg95PCI
2
-2
06/01
03/01
07/01
06/01
05/01
04/01
0.5
0.4
0.3
0.2
0.1
0.0
03/01
0.5
0.4
0.3
0.2
0.1
0.0
07/01
03/01
07/01
06/01
04/01
05/01
0.5
0.4
0.3
0.2
0.1
0.0
03/01
0.5
0.4
0.3
0.2
0.1
0.0
05/01
Ksat: 43.6~53.1
mm/day
(d) Mixed
(a+b+c: 22.2+19.8+58.0)
(a) s3v1 (Sub-pixel 3)
Ksat: 0.4 mm/day
04/01
Ksat: 106.1 mm/day
06/01
(a) s2v1 (Sub-pixel 2)
05/01
(a) s1v1 (Sub-pixel 1)
04/01
Soil moisture, cm3 cm-3
Predicted root zone soil moisture at the WC1 site during
2010-2020
M
A
M
J
J
-2
(h) Mixed
(a+b+c: 22.2+19.8+58.0)
M
A
M
J
-4
J
Although the SMDI values Average
have some uncertainties,
results showed that the sub+95PCI this
-95PCI
pixels within the RS product can be partially affected by the drought severity.
10 August 2012
41
Soil moisture estimates in Northern Texas
Estimating soil moisture dynamics using land surface
model (Community Land Model, CLM)
Upper Brazos river
Brazos River Basin
Elevation
Upper Colorado river
Colorado River Basin
Elevation
Ongoing work…
Improve subsurface process including soil moisture
interaction with groundwater in the land surface models
(CLM)
Develop scaling approach for hydraulic parameters of
hydrologic
components
(surface,
subsurface,
and
recharge/discharge)
Forecast drought severity using the Bayesian multi-scale
averaging of land surface models at regional scale in Texas