Sensitivity of soil moisture related drought indices to...
Transcript of Sensitivity of soil moisture related drought indices to...
Sensitivity of soil moisture related drought indices to
calibration data
Zoubeida Bargaoui (Prof., ENIT, Tunisia) Ahmed Houcine (PhD Student, ENIT, Tunisia)
Problem position
• Modelling time distribution of soil moisture is a key issue for natural vegetation and crop survey and is often adopted when deriving drought awareness indices.
• In drought risk analysis, it is very important to analyze uncertainty in model outputs
Climatic context: Average annual P, R nETR estimation
based on dryness index P/Rn or P/ETP
2
1
))]sinh()cosh(1)(tnh( /[ ETRP
Rn
P
Rn
Rn
PPRn +−=
Budyko (1974) approximation of the solution of 1LWBM (P/Rn)
( )( )
K
K
K
ETPP
ETPP1
1 ETP ETR
+=
Hsuen Chun HC (1999) Approximation of Budyko model (P/ETP)
Budyko model is adopted to estimate average annual ETR using a hydrometeorological network composed by 18 stations and to derive average annual regional runoff Rs (Av P = Av ETR + Av Rs)
According to observations from 18 gauged basins estimated Rs match observed Rs for k=1.5
Models of soil moisture dynamics
• (Eagleson (1978), Milly (1993), Milly(1994), Rodriguez-Iturbe et al. (1999) and Rodriguez-Iturbe (2002)).
• Evaluation of soil moisture and evapotranspiration in terms of :– soil and vegetation parameters – climatic variables
• storm depth and duration, • dry season length, • average annual potential evapotranspiration
Water balance representation EAGLESON, P. S. (1978).Climate, Soil, and Vegetatio n. Dynamics of the Annual Water Balance. W. R. R VOL. 14, 5, 749, 764
•A one layer vertically averaged water budget over the root zone taking into account• infiltration, runoff, drainage, leakage, evapotranspiration losses and capillary rise
( ) ( ) ( )
SE- SL- tS,I dt
dSnZr = Rates of Infiltration I; Leakage L
(deep infiltration); evapotranspirationE, relative soil moisture S
n porosity; Zr soil depth,
( )( )
( ) 1
1eKSL 1
S-S
sat
FC
−−= −Β
Β
FCSe
Threshold s* depends on vegetation type s*<< sfc
Loss from leakage and capillary rise
conductivity KsatField capacity SFC :
Water balance representation Rodriguez-Iturbe et al. (1999) Probabilistic modelling of water balance at a point: the role of climate, Soil, and Vegetation. Proc. R. Soc. London A (1999) 455, 3789-3805.
Gd= exp((nS-a)/b) - c
a, b, c (Ksat, SFC, B)capillary rise component (Kobayachi et al., 2001)
Runoff dependence from soil moisture
• Whenever the soil have enough available storage to accommodate the rainfall infiltrating the soil, it results an increment n soil moisture.
• Whenever the rainfall depth exceeds the available storage, the excess water is converted to excess runoff.
For S=WBC
if P(t)>ETR(t)+Gd(t)Rs(t)=P(t)-ETR(t)-Gd(t)
else Rs(t)=0
For S <WBCif P(t)>ETR(t)+Gd(t)+( WBC-S)
Rs(t)= P(t)-ETR(t)-Gd(t)-( WBC-S)else Rs(t)=0
with WBC = η*Wmax
Runoff computation
Capillary rise component of Kobayachi modelLinkage with pedo transfer parameters and
• 0< σ <1 ; 0 < η < 1
( ) 1
1Kc
1sat −= −Β FCSe
( ) ]S) 1
1K(
1[a FC1satmax +
−Β−= −Β FCSe
LnW
1
b max Β= W
Model parameters– (1) soil water retention curve shape parameter B, slope
of log ψ versus log(θ/θs) regression.
– (2) saturated hydraulic conductivity Ks,
– (3) saturated soil water content θs (volume/volume). – (4) σ resistance of vegetation to evapotranspiration– (5) η the moisture retaining capacity– Parameters estimation: – (1) to (3) are evaluated after Cosby et al. (1984)Multiple linear regression analysis of parameters by
textural class ( %Clay, %Sand, %Silt)– (4) and (5) are data driven
Case study• To illustrate the methodology, a case
study of a basin with an area of 250 km² in arid area of South Tunisia is presented.
• Vegetation is mainly olives, with no irrigation.
• The soil type is sandy.P=0.34; %S=93.3; %C=6.7• Climate: arid
Data
• Meteorological data : (at Chaal meteo station 30 km far).
• T, V, H (1989-1990, 2003-2004, 2004-2005• Pan evaporation (3 years), Piche
evapotranspiration (17 years)• P (17 years); (1989-2005)• Profiles of soil temperature (3 years)
• Basin Hydrometeometeorological data • Ten years of P and Rs (1990-1999)
time series
• A time series of meteorological observations allow quantifying the reference crop evapotranspiration ET0.
• Model inputs are daily average basin rainfall and daily ET0 over 10 years (1989-90, 1998-1999).
•• Average long term rainfall is 228 mm/an. • According to Budyko model, average ETR = 213 mm/an.
• Average basin precipitation during the study period varies from 55 mm (1994-95) to 398 mm (1989-90).
•
Epan, Epiche et ET0 (meteo station)
y = 0,8646x
R2 = 0,7202
0
1
2
3
4
5
6
7
8
0 2 4 6 8 10
Evaporation bac (mm)
Eva
potr
ansp
iratio
n de
réf
éren
ce
(mm
)
y = 1,0022x
R2 = 0,6422
0
1
2
3
4
5
6
7
8
0 2 4 6 8Evaporation piche (mm)
Eva
potr
ansp
iratio
n de
réf
éren
ce
(mm
)
0
1
2
3
4
5
6
7
8
1 26 51 76 101 126 151 176 201 226 251 276 301 326 351
Temps (jour)
Evapotranspiration de référence
(mm)
ET0c ET0v
ET0 = 1325mm (2004-2005) ; 1368mm (2003-2004)
Daily Rainfall and runoff (1990-1999)
0
10
20
30
40
50
60
70
80
90
100
1 264 527 790 1053 1316 1579 1842 2105 2368 2631 2894 3157 3420Time (Day)
Rai
nfal
l (m
m)
0
20
40
60
80
100
120
Run
off (
m3/
s)
Rain Runoff
Calibration strategy
• the model is calibrated on the basis of observed runoff at the basin outlet.
• However, it is proposed to also constraint the model outputs on the basis of on average annual evapotranspiration(regional climatic information).
Calibration strategy
• Grid (σ,η) with δσ = ; δ η = 0.01
• Objective function: error Bias and Variance
• Equifinality so,– Consider Bias and Variance of average annual Rs
error– Limit absolute Bias of average annual Rs error to
10% (local data control based on Rs)– Limit absolute Bias of average annual ETR to 10%
(regional data control based on ETR Budyko)
Error Bias and variance
600
700
800
900
1000
1100
1200
1300
1400
1500
5 10 15 20 25 30 35
biais (Rs annuelle)
varia
nce
de l'
erre
ur d
e R
s
cal with Rs
cal with Rs and ETR
Kernel distribution of ETR
80 100 120 140 160 180 200 220 240 2600
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
ETR (mm/an)
Den
sity
est
imat
e fo
r ET
R
cal withRs
cal with Rs ans ETR
Kernel distribution of the relative soil moisture
-0.05 0 0.05 0.1 0.15 0.20
2
4
6
8
10
12
14
16Density estimate for W
W%
Den
sity
est
imat
e fo
r W
cal with Rs
cal with Rs and ETR
Kernel distribution of percolation
-40 -20 0 20 40 60 80 1000
0.02
0.04
0.06
0.08
0.1
0.12
Den
sity
est
imat
e fo
r gra
vity
dra
inag
e
Cal with Rs
cal with Rs and ETR
Kernel distribution of capillary rise
-5 0 5 10 15 20 25 30 35 400
0.01
0.02
0.03
0.04
0.05
0.06
0.07
capillary rise
Den
sity
est
imat
e fo
r cap
illar
y ris
e
Cal with Rs
cal with Rs and ETR
Kernel distribution of σ and η
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.40
2
4
6
8
10
12
14
Parameters (sigma and neta)
Den
sity
est
imat
e fo
r par
amet
ers
cal with Rs (neta)
cal with Rs and ETR (neta)cal with Rs (sigma)
cal with Rs and ETR (sigma)
0
0,2
0,4
0,6
0,8
1
1,2
0,2 0,22 0,24 0,26 0,28 0,3 0,32 0,34 0,36 0,38 0,4
neta
Sig
ma
cal with Rs
cal with Rs and ETR
σ versus η through accepted solutions
Sorted η η η η and σσσσ according to the local R s as objective function
Valeur de la fonction objectif Biais annuel en fonc tion de Neta et Sigma
0,1
0,15
0,2
0,25
0,3
0,35
0,4
0 0,2 0,4 0,6 0,8 1 1,2
Sigma
Net
a
classe biais 10-15%
classe biais 15-20%
classe biais 20-25%
classe biais 25-30%
classe biais 30-35%
Sorted η η η η and σσσσ according to the local R s and regionalETR as objective function
Valeur de la fonction objectif Biais annuel en fonc tion de Neta et Sigma
0,1
0,15
0,2
0,25
0,3
0,35
0 0,1 0,2 0,3 0,4
Sigma
Net
a
classe biais 10-15%
classe biais 15-20%
classe biais 20-25%
classe biais 25-30%
classe biais 30-35%
Selection of one optimal solution (median criteria value )
600
700
800
900
1000
1100
1200
1300
1400
1500
5 10 15 20 25 30 35biais (Rs annuelle)
varia
nce
de l'
erre
ur d
e R
s
cal with Rs
cal with Rs and ETR
Optimal solution: cal with Rs
Optimal solution: cal with Rs and ETR
0,00
0,05
0,10
0,15
0,20
0,25
0,30
0,35
1 138 275 412 549 686 823 960 1097 1234 1371 1508 1645 1782 1919 2056 2193 2330 2467 2604 2741 2878 3015 3152 3289 3426 3563
W/Wmax Sfc P/Wmax
gravity drainage periods
capillary rise periods
cal with Rs
0,00
0,05
0,10
0,15
0,20
0,25
0,30
1 138 275 412 549 686 823 960 1097 1234 1371 1508 1645 1782 1919 2056 2193 2330 2467 2604 2741 2878 3015 3152 3289 3426 3563
W/Wmax Sfc P/Wmax
gravity drainage periods
capillary rise periods
cal with Rs and ETR
Effect of regional information on estimated water conten t
-0,20
0,00
0,20
0,40
0,60
0,80
1,00
1 139 277 415 553 691 829 967 1105 1243 1381 1519 1657 1795 1933 2071 2209 2347 2485 2623 2761 2899 3037 3175 3313 3451 3589
W/Wmax Sfc Gd
gravity drainage periods
capillary rise periods
cal with Rs
-0,20
0,00
0,20
0,40
0,60
0,80
1,00
1 139 277 415 553 691 829 967 1105 1243 1381 1519 1657 1795 1933 2071 2209 2347 2485 2623 2761 2899 3037 3175 3313 3451 3589
W/Wmax Sfc Gd
gravity drainage periods
capillary rise periods
cal with Rs and ETR
Effect on Water content, percolation capillary rise
0
0,1
0,2
0,3
0,4
0,5
0,6
1 253 505 757 1009 126115131765 2017226925212773 302532773529
Time (day)
Val
ue
0
20
40
60
80
100
120
cal with Rs and ETR(W/Wmax)cal with Rs (W/Wmax)
Sfc
P
Effect on relative soil moisture
Effect on ETR
0
1
2
3
4
5
6
7
8
9
10
1 267 533 799 106513311597 18632129 239526612927 31933459
Time (day)
ET
R
0
20
40
60
80
100
120
140
P (m
m)
cal with Rs (ETR)
cal with Rs and ETR(ETR)P
What is the performance on thedecadal scale?
Budyko, Milly, Eagleson
Annual scaleDecadal scaleDaily scale
R.I, K.,
Representation of Bias and variance on thedecadal scale
0
5
10
15
20
25
30
-0,06 -0,04 -0,02 0 0,02 0,04 0,06 0,08Biais (Rs décadaire)
Var
ianc
e de
l'er
reur
de
Rs
Cal with RsCal with Rs and ETR
Water budget for the median solution
Cal with RS Cal with Rs and ETRETR (mm/year) 206,1 215,3Wmoy (mm/year) 0,052 0,030Rs (mm/year) 16,72 16,65Rc (mm/year) 22,44 28,00Per (mm/year) 7,0621 3,6301Mmoy 0,245 0,266W/Wbc 0,181 0,117biais decadaire 0,014 0,014variance decadaire 19,88 19,88variance annuelle 1096,90 1096,90Biais annuelle 23,60 23,60
Conclusions
• Due to uncertainty about pedo transfer parameters as well as uncertainty about vegetation response to waterstress, a lot of uncertainty is attached to soil moisturecontent prediction
• Kernel function is a way to quantify this uncertaintyand to transfer it to other soil moisture relatedvariables (median solution)
• Collection of data on vegetation response (LAI,NDVI) (ground and remote sensing) may contribute to decrease this uncertainty