1. Prologue Study GCMs (runs) Colorado R Runoff by mid 21st century
Christensen et al. 2004
1 (3)
-18%
Milly 2005
12 (24)
-10 to -20%
Hoerling and Eischeid 2006
18 (42)
-45%
Christensen and Lettenmaier 2007
11 (22)
-6%
(-40 to 18%)
Seagar et al. 2007
19 (49)
-16%
(-8 to -25%)
Information from Table 5-1 in Western Water Assessment (WWA) report for Colorado Water Conservation Board “Colorado Climate Change: A Synthesis to Support Water Resource Management and Adaptation.” Oct 2008 (available online at:

2. + *RISA: Regional Integrated Science and Assessments Brad Udall
Dennis Lettenmaier
Julie Vano
Brad Udall
Jonathon Overpeck
Holly Hartmann
Kiyomi Morino
Dan Cayan
Hugo Hidalgo
Tapash Das
Greg McCabe (USGS)
Robin Webb (NOAA)
Marty Hoerling (NOAA)
Levi Brekke (Reclamation)
Kevin Werner (NWS RFC) +
From abstract: In response, a group of scientists from academic and federal agencies, brought together through
a NOAA cross-RISA project, set forth to identify the major sources of disparities and provide actionable
science and guidance for water managers and decision makers.
This project was developed by the four western RISA centers to address divergent results in Colorado River streamflow, trying to understand and interpret the differences between studies
RISA goal: Research that addresses complex climate sensitive issues of concern to decision-makers and policy planners at a regional level.
*RISA: Regional Integrated Science and Assessments

3. Understanding Uncertainties in Future Colorado River Streamflow
JAN 2014
Understanding Uncertainties in Future Colorado River Streamflow
JA Vano and Collaborators
Dennis Lettenmeier, (UW, CIG, CIRC)
Brad Udall (WWA)
Dan Cayan, Tapash Das,
Hugo Hidalgo (CAP)
Jonathon Overpeck, Holly Hartmann,
Kiyomi Morino (CLIMAS)
Robin Webb, Marty Hoerling (NOAA)
Greg McCabe (USGS)
Levi Brekke (Reclamation)
Kevin Werner (NWS RFC)
understand how modeling can impact the spread, or uncertainties
14 authors (Regional Integrated Science Assessment)

4. CLIMATE SURFACE LAND MGMT IMPACT GCM RCM Hydrology model Stream-flow
Next set of slides
Look more closely at the process
Stream-flow
MGMT
IMPACT
Adapted from Vano et al 2014: Figure 1

5. CLIMATE SURFACE LAND MGMT IMPACT Seager et al. 2007 Seager et al 2013
GCM
Seager et al. 2007
Seager et al 2013
Milly et al. 2005
CLIMATE
boundary
conditions
RCM
Stat. down
P-E, R
SURFACE
LAND
Hydrology model
P-E, R
flow
routing
Stream-flow
MGMT
IMPACT
Adapted from Vano et al 2014: Figure 1

6. CLIMATE SURFACE LAND MGMT IMPACT Seager et al. 2007 Seager et al 2013
GCM
Seager et al. 2007
Seager et al 2013
Milly et al. 2005
Christensen et al. 2004
Christensen and Lettenmaier 2007
Cayan et al 2010
USBR 2011
CLIMATE
boundary
conditions
RCM
Stat. down
P-E, R
SURFACE
LAND
Hydrology model
P-E, R
flow
routing
Stream-flow
MGMT
IMPACT
Adapted from Vano et al 2014: Figure 1

7. CLIMATE SURFACE LAND MGMT IMPACT Seager et al. 2007 Seager et al 2013
GCM
Seager et al. 2007
Seager et al 2013
Milly et al. 2005
Christensen et al. 2004
Christensen and Lettenmaier 2007
Cayan et al 2010
USBR 2011
Gao et al 2011
Rasmussen et al 2011
Gao et al 2012
CLIMATE
boundary
conditions
RCM
Stat. down
P-E, R
SURFACE
LAND
Hydrology model
P-E, R
flow
routing
Stream-flow
MGMT
IMPACT
Adapted from Vano et al 2014: Figure 1

8. Statistical Downscaling
GCM, Emission Scenario
& Period of Analysis
Spatial scale
Land Surface models
Statistical Downscaling

9. GCM, Emission Scenario & Period of Analysis
Uncertainty #1:
GCM, Emission Scenario & Period of Analysis
Hydrology model
RCM
GCM
Stream-flow
CLIMATE
SURFACE
LAND
Research Objectives: We compare the hydrologic sensitivities of five commonly used land-surface models to temperature and precipitation changes to better understand:
To what extent does the land-surface hydrology modulate or exacerbate regional scale sensitivities to global climate change?
How do these sensitivities vary spatially across the Colorado River basin?
How much of the range of results of these hydrologic sensitivities is attributable to model bias?
MGMT
IMPACT

10. -19% -13% Different GCMs, A1B scenario !st source of uncertainty…
Adapted from Vano et al 2014: Figure 3

11. Different GCMs, A1B scenario
-19%
-13%
!st source of uncertainty…
-24%

12. Same GCMs, Different Emissions Scenarios
-8%
-13%
-15%

13. -8% -17% -15% -10% Same GCMs, Different Emissions Scenarios
& Different Periods of Analysis
-8%
-17%
-15%
-10%

14. Lesson #1:
1. Model subset size and composition will impact projections of future streamflow.
2. Emission Scenario and Period of Analysis may matter for some model subsets.

15. Uncertainty #2: Spatial scale.
Hydrology model
RCM
GCM
Stream- flow
CLIMATE
SURFACE
LAND
Research Objectives: We compare the hydrologic sensitivities of five commonly used land-surface models to temperature and precipitation changes to better understand:
To what extent does the land-surface hydrology modulate or exacerbate regional scale sensitivities to global climate change?
How do these sensitivities vary spatially across the Colorado River basin?
How much of the range of results of these hydrologic sensitivities is attributable to model bias?
MGMT
IMPACT

16. Basin-wide, PRCP amounts are roughly equivalent in summer and winter
BUT
winter PRCP is much greater in headwaters and more efficiently produces runoff.
Images from: (L) ®

17. 15 85

18.
Runoff (mm/year)
another key source of uncertainty… Coarse spatial resolution does not resolve high elevation hydrologic processes that dominate basin runoff production
Figure from Vano et al., BAMS, January 2014

19.
Runoff (mm/year)
1/8˚
1/2˚ 1˚ 2˚
another key source of uncertainty… Coarse spatial resolution does not resolve high elevation hydrologic processes that dominate basin runoff production

20. Annual Average Runoff above
1/8˚
1/2˚ 1˚ 2˚
another key source of uncertainty… Coarse spatial resolution does not resolve high elevation hydrologic processes that dominate basin runoff production
Annual Average Runoff above
Lees Ferry (mm/yr)
Grid spacing (degrees)
Adapted from Vano et al 2014: Figure 4

21. Annual Average Runoff above
Sensitivity:
The % change in runoff for an imposed increase in T.
another key source of uncertainty… Coarse spatial resolution does not resolve high elevation hydrologic processes that dominate basin runoff production
Annual Average Runoff above
Lees Ferry (mm/yr)
Grid spacing (degrees)
Adapted from Vano et al 2014: Figure 4

22. Lesson #2:
Coarser spatial resolutions tend to be more sensitive to change from both warming and precipitation reduction.

23. Land surface representation.
Uncertainty #3:
Land surface representation.
Hydrology model
RCM
GCM
Stream-flow
CLIMATE
SURFACE
LAND
Research Objectives: We compare the hydrologic sensitivities of five commonly used land-surface models to temperature and precipitation changes to better understand:
To what extent does the land-surface hydrology modulate or exacerbate regional scale sensitivities to global climate change?
How do these sensitivities vary spatially across the Colorado River basin?
How much of the range of results of these hydrologic sensitivities is attributable to model bias?
MGMT
IMPACT

24. Daily timesteps with some sub-daily processes
Image from:
Research Objectives: We compare the hydrologic sensitivities of five commonly used land-surface models to temperature and precipitation changes to better understand:
To what extent does the land-surface hydrology modulate or exacerbate regional scale sensitivities to global climate change?
How do these sensitivities vary spatially across the Colorado River basin?
How much of the range of results of these hydrologic sensitivities is attributable to model bias?
Grid-based simulations of land-surface processes using principles of energy and water balance
Daily timesteps with some sub-daily processes
Forcing data: PRCP, T, specific humidity, wind speed, air pressure, and surface incident shortwave and longwave radiation

25. elasticity* (runoff:prcp)
Hydrologic sensitivities
For example, catchment elasticity of 3, 5% decrease in precipitation, 15% decline in streamflow, whereas with Noah 2.8 a 5% decline translates to a 11%
* elasticity is a ratio of:
the percent change in annual model runoff to
the percent change in annual precipitation.
Adapted from Vano et al 2014: Figure 5

26. sensitivity* (%∆ per deg C)
Hydrologic sensitivities
For example, catchment elasticity of 3, 5% decrease in precipitation, 15% decline in streamflow, whereas with Noah 2.8 a 5% decline translates to a 11%
* sensitivity is the percent change in runoff for an imposed increase in T.
Adapted from Vano et al 2014: Figure 5

27. Lesson #3:
Hydrology models show:
1. substantial differences in sensitivities to T increases;
2. similar responses to
precipitation change;
3. differences in P elasticity and T sensitivity are generally smaller in headwater regions.

28. Statistical downscaling.
Uncertainty #4:
Statistical downscaling.
Hydrology model
RCM
GCM
Stream-flow
CLIMATE
SURFACE
LAND
Research Objectives: We compare the hydrologic sensitivities of five commonly used land-surface models to temperature and precipitation changes to better understand:
To what extent does the land-surface hydrology modulate or exacerbate regional scale sensitivities to global climate change?
How do these sensitivities vary spatially across the Colorado River basin?
How much of the range of results of these hydrologic sensitivities is attributable to model bias?
MGMT
IMPACT

29. 200 km
1 month
10-20 km
sub-daily

30. +6% (A2 emissions scenario) Global Climate Models
Two examples, although there are many new downscaling approaches being developed… see result in differences, but these differences are not as large as differences between GCMs.
Comparison of BCSD downscaling from Christensen and Lettenmaier (2007) with a delta method downscaling approach for Lees Ferry in the future period for the A2 where, on average, the BCSD approach has a decline of 7% whereas with the delta method, declines are 13%.

31. Lesson #4:
The choice of downscaling method can affect the magnitude of the climate signal leading to differences in long-term projected runoff.

32. Re-cap
We identified four major reasons for discrepancies in Colorado River projections:
GCMs and emissions scenarios;
Spatial scale or the ability of models to simulate the disproportionate contributions to Colorado River discharge of the relatively small, high elevation runoff source areas;
Sensitivities of land surface hydrology models to P and T changes;
Methods used to statistically downscale GCM scenarios.

33. Implications for Decision makers
From past AR4 studies, we can say with high likelihood that in the Colorado:
Temperatures (T) will rise in the Colorado over the coming decades
Precipitation (P) less certain, but will likely decline on annual basis
From our analysis
Warmer T (ignoring P) will reduce runoff production (our estimates -6.5 3.5% per C at Lees Ferry)
Change in P results in streamflow response of 2 to 3 times (5% decline in P results in 10-15% decline in streamflow)
Coarse spatial resolution of models does not resolve high elevation hydrologic processes that dominate Colorado River basin runoff production

34. Parting words…
The diversity of approaches to projecting future streamflow in the scientific literature requires further analysis to make apples-to-apples comparisons in order to better understand sources of uncertainty.

35. The Uncertainty Prayer*
Grant us… The ability to reduce the uncertainties we can; The willingness to work with the uncertainties we cannot; And the scientific knowledge to know the difference. -
Thank you!
Research funded by NOAA through its RISA (Regional Integrated Sciences & Assessments) Project and its National Integrated Drought Information System *

36. Summary
14 co-authors, from academia and federal agencies, many authors of divergent papers
Reached agreement on four key sources of future uncertainties AND important certainties for decision makers in the Colorado River basin
Documented for the larger science and management communities how to approach seemingly disparate results, particularly timely with new results being released with the 5th Assessment Report (AR5)
In summary,
The paper, with it’s 14 co-authors from academia and federal agencies, reached agreement on key sources of uncertainty and important certainties
This provides information for the larger science and management communities on how to approach seemingly disparate results. If you remember back to the table of the range of results, with this study we can evaluate the studies, for instance the lowest change value 6% (c&l) is affected by the selection of gcm and downscaling approach, while  the largest value (hoerling) 45% does not capturing the headwaters with there larger spatial resolution . 
This will be particularly important with all the new studies that are likely to come out – the results we evaluated here where primarily from AR4.
Also, the process helped foster novel research that can help us to better identify how we think about future changes

37. Future research directions
Where research evolving (not ranking of research priorities):
New climate change projections
Increased spatial resolution of climate models
Improved land surface simulations
New paleoclimate reconstruction and model evaluation
Improved observational records
Strengthening connection with management community
Lesson 7: As climate science evolves, our understanding of future uncertainties will continue to improve. However, the evidence indicates there is no single magic bullet that will “reduce uncertainties”, nor will uncertainty ever be reduced to zero. Therefore, it is critical that both researchers and water managers redouble efforts and research to incorporate uncertainty and reconcile differences in future projections when possible. This will require continued communication and collaboration between the management and science communities, and will require scientists to more clearly articulate how their studies fit into existing knowledge, and explain how and why their studies do or do not agree with past work.
This is a overview of ongoing/future work, not a ranking of research priorities. 
As climate science evolves, understandings will improve. However, there will be no single magic bullet. Therefore, a continued effort to communicate and incorporate uncertainty is needed.