Download mean air temperature

Predicting effects of climate
change on stream ecosystems in
the conterminous United States:
results from a pilot study in
California
Charles P. Hawkins
Jiming Jin, David Tarboton,
Ryan Hill & John Olson
1
USEPA-STAR funder project:
Predicting effects of climate change on stream
biodiversity in the USA
Maximum Annual Air
Temperature (1971-2000)
(~ 0- 30 C)
2
Annual Mean Precipitation (1971-2000)
(<10 – 400 cm / year)
3
How will climate change affect stream
invertebrate faunas at both reference
and previously stressed sites?
Will effects of climate change
confound biological assessments?
Are certain taxa and types of streams
more vulnerable to climate change
than others?
4
National Wadeable Stream Assessment
Reference Sites
5
Climate Change
Greenhouse gas forcing
GCM then RCM
6 hr 20 km resolution
USGS Streamflow
Statistical
downscaler
Physically based
hydrologic
model
Physically
based
change
predictions
DEM, land cover,
Soils, geology
Statistical
streamflow
regime model
PCA
factors
Macroinvertebrate
composition and
assemblage
structure
Streamflow
classification
WSA Probabilistic Sites
7
V
California
as a pilot
study
1999 variation in
mean annual
temperature (oC)
(PRISM data)
8
9
Probabilities of detection (PD):
A foundation for estimating
biodiversity and calculating
biological indices
PD are associated with both abundance
and richness:
– MMIs:
• Abundance-based metrics
• Richness metrics
– O/E = ∑O/∑PD
10
Hypothetical Changes
in PD & Taxa Richness
Taxon
A
B
C
D
E
etc.
NTaxa
1999
0.87
0.01
0.92
0.74
0.16
0.60
3.30
1900
0.85
0.03
0.88
0.75
0.20
0.64
3.35
2040
0.41
0.20
0.71
0.77
0.05
0.50
2.64
2090 2090-1999
0.22
-0.65
0.36
+0.35
0.58
-0.34
0.74
0.00
0.00
-0.16
0.31
-0.29
2.21
-1.09 11
NCAR
PRISM
1999-2009
CCSM
Temperature
Stream
A2
Precipitation
Benthic
(4 km)
Invertebrate (150 km)
Data
Downscaled
Climate
RIVPACS
Predictions
Model
1999-2008
Predicted
1900-1909
Taxon2040-2049
Catchment
Specific PD
2090-2999
Data
12
Estimating O/E
1999 calibration:
∑O/∑pd
1900, 2040, 2090
predictions:
∑pd(year) / ∑pd(1999)
13
14
The California
reference sites:
340 taxa.
327 sites
classified into 14
groups for
model building. 12 site
classes
Group
13
15
Group
Group
Group
2.5
14
13
1
16
14
12
9
10
11
8
7
6
5
4
3
2
13
12
11
0
9
10
10
8
20
7
Group
6
●
5
40
4
60
2
0
3
50
1
5
2
10
Basin Area (log km2)
●
Conductivity (log uS/cm)
14
13
20
1
30
14
12
9
10
11
8
7
6
5
4
3
2
1
MAAT (C)
15
13
12
11
9
10
8
7
6
5
1.0
4
●
3
2.5
2
2.0
Elev Range (SQRT m)
Group 13:
warmest
driest
highest TDS
1
14
1.5
13
12
9
10
11
8
7
6
5
4
3
2
1
Annual PPT (mm)
5 Predictors
25
4
3
●
0
Group
3.0
●
2.0
1.5
1.0
30
20
MAAT
Predicted changes
in climate-sensitive
predictors
10
0
3.0
2.5
1900 1999 2040 2090
Decade
2.0
logCond
logPPT
2.5
1.5
2.0
1.5
1.0
1900 1999 2040 2090
Decade
1.0
1900 1999 2040 2090
Decade
17
A2 Climate Change Scenario
(CCSM 250 -> 4 km empirically downscaled predictions)
Mean Annual Temperature (oC)
1900 backcast
1999 modeled
2090 forecast
A2 Climate Change Scenario
(CCSM 250 -> 4 km empirically downscaled predictions)
Mean Monthly Precipitation (mm)
1900 backcast
1999 modeled
2090 forecast
2090-1999
comparisons
20
Changes in climate-sensitive predictors
(2090-1999)
across 327 reference sites
Precipitation
Statistic
(mm)
Mean
-26
Minimum
-61
Maximum
+5
Temp
(C)
2.4
1.4
4.6
Cond
(μS/cm)
48
-30
180
21
22
Individual taxa
• Average changes in PD
– 172 decreasers (8 ≤ 0.1)
– 168 increasers (1 ≥ 0.1)
• Many predicted local
extinctions (rare taxa).
• No predicted regional
extinctions.
23
Most Sensitive Taxa
(mean ∆PD)
-0.16
+0.10
-0.14
+0.08
-0.13
+0.07
24
What is the
consequence for
reference site
taxa richness and O/E?
25
~10% loss in mean site
richness by 2090
No loss in regional
richness
26
Mean ∆O/E = -0.12
40%
-1.0
-0.5
12%
0.0
0.5
1.0
OE5_2090_1999
Change
in O/E (2090-1999)
27
28
Predictors of
Site Vulnerability (Δ O/E)
(PRE = 0.24)
-0.12
(n=327)
MAAT > 9.8
MAP < 58 mm
-0.16
-0.04
(260)
(67)
-0.23
Coldest
Driest
-0.06
MAAT < 16.5 (150)
(110)
-0.26
0.18
(139)
(11)
Warmest
Artifact associated with
end-member predictions?
29
How realistic are these
predictions?
• Climate predictions?
– Accuracy of CCSM?
– Accuracy of downscaling?
• Biota predictions?
– General accuracy of RIVPACS
model?
– End-member problem?
30
Related Issues and Work
• Downscaling – some caveats
– Empirical versus dynamic
– Yearly/seasonally versus regime
• Invertebrate-environment
relationships
– Stream temperature models
– Flow models
31
A2 2090 Temperature
CCSM Original (150 km)
o
( C)
CCSM Downscaled (4km)
Stream
Temperature
Modeling
Pilot
Study
N = 455
50 validation
Modeling (MLR) XTEMP
Model
Geography
Climate-WS
Predictors
elevation
latitude
WS area
longitude
elevation
mean air temperature
latitude
WS area
# of frost-free days
others
R2
0.75
RMSE
1.3o
0.86
1.0o
Geography
model
SD = 1.5
Climate-WS
model
SD = 1.0
XTEMP Residuals
Variable
XTEMP
SUMMER
WINTER
Predictors
R2
RMSE
elevation
mean air temperature
latitude
# of frost-free days
WS area
0.86
1.0o
0.73
2.2o
0.75
1.7o
maximum air temperature
# of frost-free days
latitude
soil bulk density
WS area
mean air temperature
minimum air temperature
soil permeability
% granitic geology
depth to water table
Do modeled stream
temperatures improve
predictions of biota?
7 classes of CO reference sites
n = 132
How well did 2 RIVPACS-type models predict
invertebrate composition in Colorado reference
streams?
Model
Predictors
Both
Day of year
Long-term precipitation
Previous year precipitation
Local topographic relief
Latitude
Elevation
WS area
SUMMER
WINTER
Geography
Predicted
Temperature
% correct by model
Groups Surrogates 3 Temps
1
47
47
2
73
82
3
73
73
4
64
80
5
8
33
6
33
60
7
53
53
Total
54
63
Change
0
+9
0
+16
+25
+27
0
9
Expanded stream
temperature modeling
EPA-STAR grant
&
USGS/NAWQA collaboration
41
Stream temperature candidate reference sites
Stream flow modeling
by
David Tarboton and students
43
1
2
3
4
5
6
7
1
2
3
4
5
6
7
Magnitude
Predictability
Flood duration
Seasonality
Flashiness
Baseflow
Zerodays
Magnitude
Predictability
Flood duration
Seasonality
Flashiness
Baseflow
6
Zerodays
6
1, 160
6
4
4
2
2
0
0
-2
-2
-4
-4
3, 113
6
4
4
2
2
0
0
-2
-2
-4
-4
2, 140
4, 130
1
0.1
0.01
10
1
0.1
0
1
0.1
Months
0.01
8, Small flashy streams
3
Monthly mean flow m s
7, Small unpredictable streams
100
0.01
Centroid sites
50th quantile
5th - 95th range
6, Big seasonal streams
0.01 0.1
Dec
Feb
Apr
Jun
AugOct
Dec
Feb
Apr
Jun
1
10
Months
0.01 0.1
0
1
10
Months
100
5, Baseflow dominated streams
3
Monthly mean flow m s
Months
10
Months
100
10
3
Monthly mean flow m s
100
10
1
0.1
0.01
100
3
Monthly mean flow m s
4, Big streams with low
predictability
Oct
100
2, Smaller predictable
3, Mid-size perennial
intermittent streams with low baseflow streams with low seasonality
1, Seasonal streams
Aug Oct
Dec
Feb
Apr
Jun
Aug
Modeled stream flows and
temperatures improve predictions of
stream invertebrates
Model
Null
Flow (7 factors)
Temperature (3 variables)
Flow + Temperature
O/E 10th %
0.58
0.73
0.67
0.80
47
Flow and temperature modeling sites