Budyko guide to exploring sustainability of
water yields from catchments under
changing environmental conditions
Irena Creed and Adam Spargo
Western University, London, ON CANADA
And a consortium of catchment
scientists, including:
USA
Mary Beth Adams (FER)
Emery Boose (HFR)
Eric Booth (NTL)
John Campbell (HBR)
Alan Covich (LUQ)
David Clow (LVW)
Clifford Dahm (SEV)
Kelly Elder (FRA)
Chelcy Ford (CWT)
Nancy Grimm (CAP)
Jeremy Jones (BNZ)
Julia Jones (AND)
Stephen Sebestyen (MAR)
Mark Williams (NWT)
Will Wolheim (PIE)
Meryl Alber (GCE)
John Blair (KNZ)
William Bowden (ARC)
Ward McCaughey (TEN)
Teodora Minkova (OLY)
Dan Reed (SBC)
Leslie Reid (CAS)
Phil Robertson (KBS)
Jonathan Walsh (BES)
CANADA
Fred Beall (TLW)
Tom Clair (KEJ)
Robin Pike (CAR)
John Pomeroy (MRM)
Patricia Ramlal (ELA)
Rita Winkler (UPC)
Huaxia Yao (DOR)
2
Budyko Curve describes the theoretical energy
and water limits on the catchment water balance
(P-ET=Q).
Budyko Curve provides a “business as usual”
reference condition for the water balance.
If we assume it depicts the expected partitioning
of P into ET and Q,
then we can begin to account for the reasons
why sites depart from the baseline.
Russian climatologist
1920 –2001
Can the Budyko Curve be used to identify
catchments undergoing shifts in water yields or
at risk of undergoing these shifts?
Budyko Curve
3
Water limit (AET=P);
a site cannot plot above
the blue line unless there
is input of water beyond
precipitation.
Energy limit (AET=PET); a
site cannot plot above the
red line unless
precipitation is being lost
from the system by means
other than discharge.
Budyko Curve
4
Energy limited (wet);
AET is limited by the amount of
thermal energy that is available.
Budyko Curve
Water limited (dry);
AET is limited by the amount
of water that is available.
5
Less runoff
VERTICAL deviations
reflect a change in partitioning
between ET and Q
Warmer, drier
HORIZONTAL deviations
reflect a change in the climatic conditions
(temperature, precipitation)
Budyko Curve
6
• Network of catchments across
North America
• Represent longest existing paired
records of meteorology and
hydrology.
• Provide opportunity to explore
effects of climate on water yields
in headwaters
• Backdrop: P-PET (30-yr climate
normals, 1971 to 2000)
North American Network
7
(1) Under stationary conditions
(naturally occurring oscillations),
catchments will fall on the Budyko
Curve.
(2) Under non-stationary conditions
(anthropogenic climate change),
catchments will deviate from the
Budyko Curve in a predictable
manner.
Hypotheses
8
Do the catchments fall on the
Budyko Curve?
9
Distribution of sites on Budyko Curve based on
“common” 10 year period of data
Budyko Curve
Jones JA et al. (2012). Ecosystem processes and human influences regulate streamflow response to
climate change at long-term ecological research sites. BioScience 62: 390–404.
10
Longterm Average AET/P
deviation from the Budyko
-0.1
HBR
AND
FER
TLW
MAR
NTL
OLY
CAS
ELA
DOR
SBC
KEJ
MRM
MAY
KBS
PIE
TEN
CWT
UPC
GCE
NWT
CAR
KNZ
BNZ
BES
SEV
LUQ
FRA
LVW
CAP
ARC
Long term average deviation in AET/P
(i.e., partitioning between ET and Q)
0.4
0.3
0.2
0.1
0.0
-0.2
-0.3
Deviations from Budyko Curve
11
Reasons for falling off the Budyko Curve
1.
Inadequate representation of P and T (Loch Vale)
2.
Inadequate representation of ET (Andrews)
3.
Inadequate representation of Q (Marcell)
4.
Forest conversion (Coweeta)
5.
Forest disturbance (Luquillo)
1. Today’s Focus: Response to changing climatic conditions
Deviations from Budyko Curve
12
We assume that the Budyko
Curve represents the
reference condition for the
time period prior to
anthropogenic climate change
being detected in water yields.
Climate related deviations:
the “d” statistic
13
For naturally occurring climate
oscillations, the partitioning
between ET and Q should
move up and down with the
Budyko Curve.
Climate related deviations:
the “d” statistic
14
If the partitioning between ET
and Q moves away from the
Budyko Curve,
then this can be attributed to
anthropogenic climate
change.
Climate related deviations:
the “d” statistic
15
The “d” statistic, the deviation
in AET/P due to climate
change, is calculated.
Climate related deviations:
the “d” statistic
16
We know that not all
catchments fall on the Budyko
Curve for reasons unrelated to
climate (i.e., o and ô on plot).
We assume these offsets are
constant before and after
climate change, and so this
term becomes zero.
Climate related deviations:
the “d” statistic
17
Negative d represents a
downward shift and an
increase in Q (more water
yield).
Positive d represents an
upward shift an a decrease in
Q (less water yield).
Climate related deviations:
the “d” statistic
18
Can we identify catchment properties
that influence water yields?
19
Spider plots showing year-to-year deviations
from long term average
Inter-annual variation along Budyko Curve
Jones JA et al. (2012). Ecosystem processes and human influences regulate streamflow response to
climate change at long-term ecological research sites. BioScience 62: 390–404.
20
Spider plots showing year-to-year deviations
from long term average
Inter-annual variation along Budyko Curve
Jones JA et al. (2012). Ecosystem processes and human influences regulate streamflow response to
climate change at long-term ecological research sites. BioScience 62: 390–404.
21
Responsivity is
measured as the
maximum range in AET/P
after accounting for
natural deviation in the
Budyko Curve.
Pre climate change responsivity
22
LOW RESPONSIVITY
Water yields are
not synchronized to P
High vs. low responsivity
HIGH RESPONSIVITY
Water yields are
synchronized to P
23
PREDICTION #1:
Larger deviations in
catchments with
higher responsivity
(catchments cannot buffer
against climate change and
water yields strongly linked
to the atmosphere).
Pre climate change responsivity vs. “d”
24
Elasticity is measured
as the ratio of range of
PET/P to AETP/P.
Pre climate change elasticity
25
LOW ELASTICITY (<1)
small PET/P range
relative to AET/P range
High vs. low elasticity
HIGH ELASTICITY (>1)
large PET/P range
relative to AET/P range
26
PREDICTION #2:
Larger deviations in
catchments with
lower elasticity
(catchments cannot
acclimate/adapt to
changing climatic
conditions)
Elasticity
Pre climate change elasticity vs. “d”
27
2.5
Y = 4.90x – 2.86
R² = 0.039
P = 0.07
Elasticity
2.0
1.5
1.0
0.5
0.0
0.5
0.6
0.7 0.8 0.9
Responsivity
1.0
Responsivity does not imply elasticity
28
Applying these metrics to test
hypotheses
29
Defining the onset of
anthropogenic climate change to
identify pre vs. post behavior
Wang and Henjazi adopted a constant breakpoint (1970)
to detect the effects of global environmental change on
water yields across USA
30
Wang D and M Hejazi. (2011). Quantifying the relative contribution of the climate and direct human
impacts on mean annual streamflow in the contiguous United States. Water Resour. Res. 47: W00J12,
doi:10.1029/2010WR010283.
Record length of P, ET (T)
and Q data highly
variable among the
catchments
Constant breakpoint
31
Record length of P, ET (T)
and Q data highly
variable among the
catchments
A 1970 breakpoint
captures only 7
headwater sites!
Constant breakpoint
32
Some sites observing net positive changes while
others observing net negative changes.
Constant breakpoint: pre vs. post changes
33
0.08
0.06
0.06
0.04
0.04
0.02
0.02
0.00
0.00
-0.08
AND WS08
AND WS02
MRM
HBR WS6
HBR WS3
CWT WS18
-0.06
FER WS4
-0.04
MAR S2
-0.02
d
0.08
MAR S5
d
POSITIVE deviations (lower water yields) in coniferous forests and
NEGATIVE deviations (higher water yields) in deciduous forests
-0.02
Deciduous
Coniferous
-0.04
-0.06
-0.08
Constant breakpoint: deviation (“d”)
34
With Hubbard WS06
Without Hubbard WS06
PREDICTION #1
Absolute value of d
0.07
0.06
y = 0.12x - 0.06
R² = 0.39
p = 0.07
0.05
0.04
0.03
y = 0.12x - 0.05
R² = 0.85
0.00
0.02
0.01
0.00
0.00
0.50
Responsivity
Constant breakpoint: responsivity vs. “d”
1.00
35
With Hubbard WS06
Without Hubbard WS06
Elasticity
PREDICTION #2
Absolute value of d
0.07
0.06
0.05
y = 0.02x + 0.02
R² = 0.48
p = 0.04
y = 0.01x + 0.03
R² = 0.57
p = 0.03
0.04
0.03
0.02
0.01
0.00
0.00
1.00
2.00
Elasticity
Constant breakpoint: elasticity vs. “d”
3.00
36
Use AutoRegressive Integrated Moving Average technique to check
for breakpoints at each year from 1960 to 2000 (Ford et al. 2006).
1981
No breakpoint
Breakpoint at 1981
Variable breakpoint
Ford CR, SH Laseter, WT Swank, JM Vose. (2011). Can forest management be used to sustain waterbased ecosystem services in the face of climate change? Ecological Applications 21: 2049–2067.
37
Variable breakpoints
allowed inclusion of
additional sites (CAR,
NTL), but also exclusion of
sites with no detectable
breakpoint (AND).
Variable breakpoint
38
Similar separation of
deciduous (positive water yields) vs.
coniferous (negative water yields)
in the variable breakpoint analysis, except Carnation Creek.
0.10
Marmot
Coweeta WS 17
Coweeta WS18
Hubbard WS6
Fernow
Marcel S5
Hubbard WS3
-0.10
Marcel WS2
-0.05
Carnation
d
0.00
North temperate…
0.05
-0.15
Variable breakpoint: deviation (“d”)
39
Absolute value of d
0.14
0.12
With Carnation
0.10
Without Carnation
Linear (With Carnation)
0.08
Linear (Without Carnation)
y = 0.07x - 0.01
R² = 0.05
y = 0.14x - 0.07
R² = 0.48
0.06
0.04
0.02
PREDICTION #1
0.00
0.00
0.20
0.40
0.60
0.80
Responsivity
Variable breakpoint: responsivity vs. “d”
1.00
40
1.20
0.14
Without Carnation
With Carnation
Linear (Without Carnation)
Linear (With Carnation)
Absolute value of d
0.12
Elasticity
PREDICTION #2
0.10
0.08
y = 0.01x + 0.03
R² = 0.06
y = 0.01x + 0.01
R² = 0.69
0.06
0.04
0.02
0.00
0.00
1.00
2.00
3.00
Elasticity
Variable breakpoint: elasticity vs. “d”
4.00
5.00
41
6.00
Constant vs. variable breakpoints
Similar relationships observed between responsivity and elasticity
vs. absolute value of “d”.
Variable breakpoint relationships reveal a “significant” outlier
(Carnation Creek, BC).
Need to delve further into the data to identify causes for this
outlier – examine magnitude of temperature increase after the
breakpoint.
Is the outlier key to our understanding?
42
As temperature increases above 0, d increases
Average increase
in temperature
estimated by
slope of line after
breakpoint
Rate of temperature increase vs. “d”
43
As temperature increases above 0, d increases
Average increase
in temperature
estimated by
slope of line after
breakpoint
Rate of temperature increase vs. “d”
44
As temperature increases above 0, d increases
Average increase
in temperature
estimated by
slope of line after
breakpoint
Rate of temperature increase vs. “d”
45
Outlier: Carnation Creek, BC
• Western hemlock
• Largest “d” and highest rate of
anthropogenic climate change
(2°C/decade)
Is water yield of the outlier
catchment at greater risk
because its tree species is “at the
edge” of its climate tolerance?
Climatic Parameter
Parameter Mean
Parameter Range
(min and max)
Annual mean
temperature
5.13 °C
-4.66 to 12.93 °C
Annual total
precipitation
1600 mm/year
237 to 4196 mm/year
46
Prior to anthropogenic climate
change, Carnation Creek fell at
the edge of the range of
Western Hemlock.
Maximum Range
5th – 95th Percentile
Baseline 1971 - 2000
McKenney DW, et al. (2007). Potential impacts of climate change on the distribution of North
American trees. Bioscience 57: 939-948.
47
Based on CGCM model
simulations, the range of
Western Hemlock will recede
on Vancouver Island.
Maximum Range
5th – 95th Percentile
First 30 years, 2011 - 2040
McKenney DW, et al. (2007). Potential impacts of climate change on the distribution of North
American trees. Bioscience 57: 939-948.
48
Based on CGCM model
simulations, the range of
Western Hemlock will recede
on Vancouver Island.
By mid 21st century, Carnation
Creek will lie at the edge of the
maximum range of Western
Hemlock.
Maximum Range
5th – 95th Percentile
Second 30 years, 2041 - 2070
McKenney DW, et al. (2007). Potential impacts of climate change on the distribution of North
American trees. Bioscience 57: 939-948.
49
Based on CGCM model
simulations, the range of
Western Hemlock will recede
on Vancouver Island.
By mid 21st century, Carnation
Creek will lie at the edge of the
maximum range of Western
Hemlock.
By end 21st century, Carnation
Creek will lie outside the
maximum range for Western
Hemlock.
Maximum Range
5th – 95th Percentile
Third 30 years, 2071 - 2100
McKenney DW, et al. (2007). Potential impacts of climate change on the distribution of North
American trees. Bioscience 57: 939-948.
50
Absolute value of d
LARGE DEVIATIONS
IN WATER YIELDS
Unresponsive
inelastic
Shifting to alternative
stable states?
+ 2.0 deg C/yr
+ 1.5 deg C/yr
+ 1.0 deg C/yr
+ 0.5 deg C/yr
Responsivity and/or elasticity
SMALL DEVIATIONS
IN WATER YEILDS
Responsive
Elastic
A better conceptual model of
climate change effects on water yields
51
• NSERC Discovery Grant and Canada Research Chair Program
• “LTER Synthesis Workshops” funded by the LTER Network Office
• NSF LTER grants to participating USA sites
• USFS and USGS for initial establishment and continued support of
watershed studies at many of the study sites
• NCE-SFM funded project on HydroEcological Landscapes and
Processes (HELP) and the participating Canadian sites
Acknowledgements
52
53
Loch Vale (LVW):
Failure to apply adiabatic lapse rate to meteorological data to
account for orographic effects results in shift away from the
curve.
Evaporative Index
(AET/P)
1.0
Annual Raw Data
Adiabatic Lapse Rate
0.8
0.6
0.4
0.2
0.0
0.0
0.2 0.4 0.6 0.8
Dryness Index (PET/P)
1.0
!
(1) Inadequate representation of P, T
54
HJ Andrews (HJA):
Failure to consider net radiation, relative humidity and/or wind
speed results in shift away from the curve.
Evaporative Index
(AET/P)
Formula
Variables
1.0
Thornthwaite
T
Hamon
T
0.8
Priestley-Taylor
T, NR
0.6
PenmanMonteith
T, NR, RH, WS
0.4
PET Penman-Monteith
PET Hamon
0.2
0.0
0.0
0.2 0.4 0.6 0.8
Dryness Index (PET/P)
1.0
(2) Inadequate representation of ET
55
Marcel (MAR):
Failure to consider surface vs. groundwater losses of
precipitaton.
Evaporative Index
(AET/P)
1.0
0.8
0.6
0.4
Discharge
Discharge + Groundwater
0.2
0.0
0.0
0.2 0.4 0.6 0.8
Dryness Index (PET/P)
1.0
(3) Inadequate representation of Q
56
Coweeta (CWT):
Conversion of forest from deciduous to coniferous forest results
in shift away from the curve.
Evaporative Index
(AET/P)
1.0
0.8
0.6
0.4
Deciduous
Coniferous
0.2
0.0
0.0
0.2 0.4 0.6 0.8
Dryness Index (PET/P)
1.0
(4) Forest management effects
57
Luqillo (LUQ):
Disturbance Effects
Evaporative Index
(AET/P)
1.0
Before hurricane
1-5yrs after hurricane
>5yrs after hurricane
0.8
0.6
0.4
0.2
0.0
0.0
0.2 0.4 0.6 0.8
Dryness Index (PET/P)
1.0
(5) Natural disturbance effects
58
• Responsivity and elasticity were directly correlated to
climate related deviations from the Budyko Curve
(among the catchments studied)
• Catchments where P and Q are synchronised catchments
are more sensitive to climate change, that is they are
tightly coupled with the atmosphere.
Constant breakpoint findings
59