stress testing: recharge scenarios to quantify streamflow … · stress testing: recharge...
TRANSCRIPT
LOGYHYDRStress testing: recharge scenarios to quantify streamflow drought sensitivityMichael Stoelzle1,2, Maria Staudinger3, Markus Weiler2, and Kerstin Stahl1 EGU2018-3391: Session “Hydrological extremes: from droughts to floods” (HS2.2.1)
Reference fromoriginal HBV simulation
Stress test assessment(Reference day: 1st June)
(1) Environmental Hydrological Systems, Faculty of Environment and Natural Resources, University of Freiburg, Germany (3) Department of Geography, University of Zurich, Switzerland (2) Hydrology, Faculty of Environment and Natural Resources, University of Freiburg, Germany Contact: [email protected] Poster-Download: researchgate.net/profile/Michael_Stoelzle
Poster highlights in 1 minute12x wettest
months
Long-term average
Sorting the climate forcing according to decreasing recharge from wet to dry conditions (Staudinger et al., 2015).
Prec
epita
tion
Stre
amflo
w
JFMAMJJASOND
JFMAMJJASOND
JFMAMJJASOND
JFMAMJJASOND
Mon
thly
rech
arge
JFMAMJJASOND
12x 2nd wettest months
12x averagemonths
12x 2nd driest months
12x driest months
Temperature series is permuted according to precipitation series
DRY-UPFrom wet to totally dry
NAT-VARNatural variability
01-Jun 2003
20012000
2004
…
1999
2005…
↺2011
20092008
2012
…
2007
2013…
↺20132012
2016
…
2011
2017…
200519851976
31-May
1976
1977
…2001
20032004…
20142015
Further drought years…
Change initial conditions for historical droughtyears
2015↺
Streamflow years
50:50 drier/wetter years
What would have happened in a drought year if the preceding recharge year was changed?
More years drier
More yearswetter
ReferencesReading list
▪ Vormoor et al., 2017, Climatic Change▪ Staudinger et al., 2015, Hydrology and
Earth System Sciences▪ Stoelzle et al., 2014, Geophysical
Research Letters▪ github.com/tidyverse▪ github.com/clauswilke/ggridges/
STRE
SS T
EST
SCEN
ARIO▪ Stress testing scenarios are a
valuable alternative to conventional climate scenarios.
▪ “Stress” means recharge reduction before drought, “Test” means a quantification of ”Stress”-effect on the drought year (e.g. deficit, recovery, minimum flow).
▪ Scenarios are embedded in a HBV model framework to compare streamflow series from scenario runs with streamflow from reference runs.
▪ NAT-VAR stress tests elucidate that drought is catchment- and event-specific, thus improved stress tests should consider more event-specific characteristics (< monthly scale).
▪ Drought propagation is controlled by short-term and long-term recharge reduction and by recharge timing.
▪ DRY-UP stress test filters which catchments are more sensitive to progressive drying (dry weather).
▪ Successful identification of catchments with higher drought sensitivity.
HBV
refe
renc
eH
BV s
cena
rio
Runoff ratio (Q/P) during DRY-UP:Test catchments’ sensitivity to progressive drying
snow-dominatedrain-dominated hybrid
Z-score (Q/P)
Drie
r las
t dec
ade
Intensity
Deficit
RecoveryReference
D Terms & Framework
▪ “Stress” is defined as systematic decrease of recharge, “Test” is defined as a framework to quantify the streamflow response to this changed main control.
▪ Stress test scenarios are not about hydrological prediction, but identify catchments’ sensitivity to drought.
▪ All scenarios are compared against the HBV reference run (never against Qobs)! HBV recharge is percolation water from soil box into GW box (incoporates also snow melt)
▪ All scenarios focus on recharge reduction or decreased recharge rates, but for scenario simulation P and T is changed (HBV input)
▪ 40 meso-scale (1-1000km2) catchments grouped into rainfall-, hybrid- and snowmelt-dominated regimes (<800m, 800-1600m, >1600m).
▪ Daily data 1971-2015: Precipitation (gridded, Rhires), Temperature (gridded, MeteoSwiss) and observed streamflow (FOEN)
Why stress testing??Climate change studies often fail to distinguish between inherent climate variability and projected climate change signal. For example, different temporal structures of future climate input can affect low flows, but the sequencing of simulated wet and dry spells is not altered (Vormoor et al., 2017). Climate change scenarios bring huge uncertainties to hydrological assessment of future streamflow droughts and low flow events.
Instead of future climate, we use historical data ▪ to alter antecedent water deficit
conditions while preserving catchment-specific climate variability
▪ to quantify how sensitive catchments are to such alterations (“Stress”), i.e. under decreased recharge before drought.
Through alteration of antecedent water defict a new sequencing in drought propagation can be tested.
Data & Catchments1101001011
HBV model▪ Scenario development is based on
GAP-calibrated HBV reference run▪ Calibration 1980-2000 with 70%
adapted KGE and 30% MARE.▪ Model spin-up with 5 average years▪ P and T gradients based on input data
0 time 1
~0
0.5~1
0 R
echa
rge
amou
nt 1
0 time 1
Recharge timinglike Gini-coefficient and
Lorenz curve
1975
1983
1986
1988
1989
1991
2003
2005
2006
2008
2010
2011
2013
1976 1985 2003 2005 2011 Drought year
Catc
hmen
t-spe
cific
min
imum
rech
arge
sce
nario
yea
r
25
50
75
100
125
Recharge in scenario year (%)
drierwetter
Scenario year(1975-2015)
Drought yearrepeated
6
13
2
<1
7
1
7
8
16
2
22
<1
15
Ratio (%) of recharge year as “minimum scenario year” for all drought years.
▪ 66% of all mimimum recharge years in 2013, 2010, 2006 and 1983
▪ Highest average recharge reduction with 2005 (-53%) compared to 2003 (-19%)
▪ Repeated drought years are less or not effecive
▪ Wet decade 1992-2002▪ More driest recharge years after
2002 (~70%) than before 1992 (~30%)
Germany
Italy
Austria
France
▪ Rsce / Rref in the 2 months before drought▪ Rsce / Rref in the year before drought▪ Recharge timing (0=early, 1=late) in the year before
drought (relative to Rref)
▪ ∆Qabs Qsce — Qref▪ ∆Qrel Qsce / Qref▪ Trec Recovery time (until ∆ < ± 2%)▪ ∆NQ NQsce / NQref
Cause (before 1th June)
Effe
ct (a
fter 1
st J
une)
Only sc
enari
os with
drier
reco
very
(compa
red to
refer
ence
)
Data: 2
003,
2005
, 201
5
Correlat
ion analy
sis
JFMAMJJASOND
Recha
rge un
iform
ly distr
ibuted
●
●
●
●●●
●●
●●●
●
●●●●
●
●●
●
● ●
●
●● ●
●
●
●●●●
●
●
●
●
●●
●
●
●
●●
●
●●
●●
●
●●
●
●● ●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●●●
●
●
●●
●●
●
●●●
●
●
●●
●
●●
●
●
●●●●
●
●
●
●
●●●
●●
●
●
●●
●
●●●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●●●●
●
●●
●
●
●●
●
●● ● ●
●●
●
●
●
●●●
●
●
●
●
2005 2011
1976 2003
0.0 0.5 1.0 0.0 0.5 1.0
0
1
2
0
1
2
ratiorec_sumr
●
●
●
hybridrain−dominatedsnow−dominated
Rsce / Rref (during Scenario)Q
sce /
Qre
f (dur
ing
Reco
very
)
4461
22 29 2437
0
25
50
75
100
1976 1985 2003 2005 2011 2015
%
94% 72% 53% 70% 66% 48%Streamflow related to long-term average (Jun-Nov)
Drought year gets due to NATVAR scenariodrierwetter
40 catchments
Drought intensification / attenuation due to permuted antecedent recharge years
Effect of minimum “worst-case” recharge year before a typical drought year
RESU
LTS
STRE
SS T
EST
SCEN
ARIO
RESU
LTS
StationID
40 years of progressive dry-up
Recharge sum (—1 year)
Recharge timing (—1 year)
Less annual Q/P variation More annual Q/P variation
2468
_SIT
2609
_ALP
2635
_GRO
2305
_GLA
2374
_NEC
2487
_KLE
2283
_RAP
2300
_MIN
2251
_RO
T26
04_B
IB21
59_G
UE23
08_G
OL
2282
_SPE
2414
_RIE
2603
_ILF
2343
_LAN
2608
_SEL
2412
_SIO
2386
_MUR
2179
_SEN
2034
_BRO
2471
_MUR
2070
_EM
M23
12_A
AC24
50_W
IG26
29_V
ED24
37_P
AR23
27_D
IS23
69_M
EN21
55_E
MM
2126
_MUR
2232
_ALL
2356
_RIA
2497
_LUT
2461
_MAG
2321
_CAS
2366
_PO
S22
44_K
RU
∆NQTrec∆Qrel∆Qabs
∆NQTrec∆Qrel∆Qabs
StationID
Recharge sum (—2 months)
∆NQTrec∆Qrel∆Qabs
Pearson’s rank correlation coef.-0.7 +0.7
p < 0.2p < 0.01
▪ Correlation between short-term recharge and drought characteristics suggest smaller catchment-storage and shorter storage memory.
▪ Correlation between long-term recharge and drought characteristics suggest larger catchment-storages (e.g. 2343_LAN, 2308_GOL)
▪ For a faster recovery the recharge amount is more important than the recharge timing.
Snow-dominated catchments in bold font type, here snowpack- and snowmelt-effects (before June) must be considered.
21 Which recharge years should be used to test the “Stress”-effect on drought years?
-2 0 +2 2 0 2 2 0 22437_PAR
2369_MEN2155_EMM2070_EMM2450_WIG
2461_MAG2282_SPE
2603_ILF2497_LUT2179_SEN2386_MUR2312_AAC2321_CAS2471_MUR2034_BRO
2414_RIE2126_MUR2409_EMM2608_SEL2487_KLE2412_SIO2604_BIB
2629_VED2343_LAN2283_RAP2374_NEC2159_GUE2305_GLA2468_SIT
2609_ALP2635_GRO
2327_DIS2232_ALL
2605_VER2366_POS2300_MIN2244_KRU2356_RIA
▪ Monthly Q/P from Jun-Nov▪ Catchments sorted along 90th
percentile of Z-score of the driest decade (in yellow)
▪ Stronger shifts to negative Z-scores suggest higher sensitivity of catchments to drying
▪ Snowmelt influence in higher-elevation catchments
▪ Earlier / later tipping points
Wettest Decade
Driest decade
Relationship between antecedent recharge- and following drought-characteristics
Effect of driest recharge years (x-axis) on streamflow (y-axis) for 4 drought years:
▪ Antecedent recharge leads to wettest reference in 2003, driest reference in 2011
▪ Streamflow reduction is smaller for snow-dominated catchments
▪ Hybrid catchments often more sensitive to reduced recharge (e.g. 1976, 2011)
▪ Analysis supports the “drought year”- and “catchment”-specific approch
snow-dominated (> 1600 m asl)hybridrain-dominated (< 800 m asl)
▪ Severity of summer streamflow droughts is different
▪ Changing antecedent conditions leads to wetter and drier drought periods
▪ Thus, a catchment- and event-specific approach is needed
Data