Development of a European Land Data Assimilation System (ELDAS)
COSMIC and Land Data Assimilation
description
Transcript of COSMIC and Land Data Assimilation
1
COSMIC and Land Data Assimilation
Rafael Rosolem
COSMOS 3rd WorkshopDecember 11, 2012
W. J. Shuttleworth1, M. Zreda1, A. Arellano1, X. Zeng1, T.
Hoar2, J. Anderson2, T. Franz1, S. A. K. Papuga1, Z. M. S.
Mejia1, M. Barlage2, J. S. Halasz1
1 University of Arizona2 National Center for Atmospheric Research
2
3
Why do we need a forward operator for COSMOS? “Effective” measurement depth depends on soil moisture
Can reach several individual layers of a typical land surface model
Therefore, direct assimilation of neutron intensity is more desirable!!!
4
Land Surface Model (LSM)
Modeled Soil
Moisture Profile
Requires an accurate model to interpret modeled soil moisture profiles in terms of the above-
ground fast neutron count
GOALto update LSM soil moisture profiles by assimilating the
cosmic-ray fast neutron count
Monte Carlo Neutron Particle model (MCNPx)
does that but it is too slow for use in data assimilation
Data Assimilation of Neutron Counts
5
COSMIC is a simple analytic model which: captures the essential below-ground physics that MCNPX represents can be calibrated by optimization against MCNPX so that the nuclear
collision physics is re-captured in parametric form
Exponential reduction in the number of high energy
neutrons with depth
Isotropic creation of fast neutrons from high energy
neutrons at level “z”
z
Exponential reduction in the number of the fast neutrons created at level “z” before
their surface measurement
high energy neutrons fast neutrons
Ne
z
COsmic-ray Soil Moisture Interaction Code (COSMIC)
6
The resulting analytic function that describes the total number of fast neutrons reaching measurement point is:
Two parameters measured in situ (soil bulk density and lattice water) and six to be defined:
L1 , L2 and L4 are site-independent and are easily determined from MCNPX
L1 = 162.0 g cm-2
L2 = 129.1 g cm-2
L4 = 3.16 g cm-2
N , and L3 require multi-parameter optimization against site specific-specific runs of MCNPX for a range of hypothetical soil moisture profiles
2
1 2 3 40 0
( ) ( )( ) ( )2 1exp . . exp . .
coss sw w
COSMOS s w
m z m zm z m zN N z z d dz
L L L L
A few meters will do!
Exponential reduction in the number of high energy
neutrons with depth
Isotropic creation of fast neutrons from high
energy neutrons at “z”
Exponential reduction in the number of the fast neutrons created at level
“z” before their surface measurement
COsmic-ray Soil Moisture Interaction Code (COSMIC)
7
Fort Peck
Bondville
Chestnut Ridge
Santa Rita
Coastal Sage
Calibrating COSMICHypothetical soil water profiles
8
Output DataInput Data
9
Output DataInput Data
10
Output DataInput Data
11
Output DataInput Data
12
Output DataInput Data
13
Output DataInput Data
14
Output DataInput Data
15
Output DataInput Data
16
Output DataInput Data
17
Output DataInput Data
18
Output DataInput Data
19
Output DataInput Data
20
Output DataInput Data
21
Output DataInput Data
22
Output DataInput Data
23
Output DataInput Data
24
Output DataInput Data
25
Output DataInput Data
26
Output DataInput Data
27
Output DataInput Data
28
Output DataInput Data
29
Output DataInput Data
30
COSMIC Effective Depth
31
Using COSMIC to estimate COSMOS counts from measured soil moisture profiles (TDT sensors)
COSMIC Performance at Santa Rita (AZ)
Running time for a single soil moisture profile
MCNPx ~ 30-60 minutes
COSMIC ~ 0.5 seconds
32
COSMIC
http://www.ral.ucar.edu/research/land/technology/lsm.php http://www.image.ucar.edu/DAReS/DART/
COSMOS
Data Assimilation Framework
33
34
35
Soil Moisture Dynamics 40 ensembles with perturbed forcing data (Santa Rita, AZ): 2010-07-
28_00Z through 2010-08-23_23Z (x-axis hour timesteps)
No assimilation!!! member runs are unconstrained!!! “Damped” process driven by (rainfall) pulses
36
Data Assimilation Results: Santa Rita (AZ)
R2 = 0.97, RMSE = 48 cph, BIAS = -26 cph
R2 = 0.84, RMSE = 840 cph, BIAS = -832 cph
40 ensembles: 2011-07-03_00Z through 2011-09-14_23Z With and without assimilation of observed COSMOS neutron
counts
37
NOAH Δz1
NOAH Δz2
NOAH Δz1
NOAH Δz2
Updated Soil Moisture Profiles
No Assimilation
Assimilated
TDT (independent) measurements
38
Integrated (depth-weighted) Soil Moisture
39
40
Low Spread and Negative Soil Moisture
41
Surface Energy Fluxes
42
Problems to be solved
Reduced ensemble spread at lower soil moisture test other filter types test different inflation parameters log-transform: initial tests = simulation crashes assimilation of multiple observations (e.g., SMOS) adopt a minimum soil moisture threshold in DART but add
small noise when updating variables (to ensure individual ensemble members won’t converge to minimum allowed)
Calibrate key soil moisture/surface flux parameters in Noah (currently working on that)
DART