Climate Research in Support of
Stakeholder Applications in the
Urban Northeast U.S.
Radley Horton
Columbia University / NASA GISS
April GloDecH Meeting
LDEO, April 6, 2011
Consortium for
Climate Risk in the
Urban Northeast
(CCRUN)
A NOAA Regional Integrated Sciences and Assessments
(RISA) Project
Boston New York Philadelphia
CCRUN conducts stakeholder-driven research that
reduces climate-related vulnerability and advances
2
opportunities for adaptation in the urban Northeast
CCRUN Mission
CCRUN conducts stakeholder-driven research that
reduces climate-related vulnerability and advances
opportunities for adaptation in the urban Northeast
Storm damage in Westchester County, NY, March 12-15, 2010. Striped bass fishing in Boston Harbor.
Source: James Estrin / The New York Times Source: Capt. Bill Smith / FishBoston.com
3
Overview
• Five-year project
• Geographic scope includes the
Boston – New York – Philadelphia
urban corridor
• Focus on sectors (water, coasts,
and health), vulnerable
populations, and adaptation
• Partnership between Columbia
University (Rosenzweig), City
College of New York
(Khanbilvardi), Drexel University
(Montalto), the Stevens Institute of
Technology (Blumberg), and the
University of Massachusetts
(Palmer) 4
CCRUN Research Areas and Activities
Water Coasts Health
Climate
Opportunities • Vulnerability •
Adaptation • Management
5
CCRUN Prototype Projects:
Water Resource Management
• Flooding and storm water management for the
MWRA and Connecticut River Basin
• Delaware River Basin-New York City Water
System Management
• Climate information for water harvesting and re-
use strategies in Philadelphia
• Precipitation, drought indices at a range of
timescales
CCRUN Prototype Projects: Health
• Heat/Air Quality During heat waves
(tmax, tmin, specific
• Coastal Storms and humidity), even a small
Intense Precipitation increase in
temperature can mean a
Events large increase in energy
0ºC 16ºC 32ºC
load….
…Leading to an
increased risk of
power outages and
deteriorating air
Daily Electric Energy Load quality
(gigawatt-hours) in NY State, vs.
Daily-Average Temperature. Indirect effects
Solid Points=1966; Open
Points=1997 => =More stress on systems
Peak Load Issues
CCRUN Prototype Projects:
Coastal Zone Management
• Storm Surge Forecasting and Projections in
the New York Metropolitan Region, with
Extension to the Entire Urban Northeast
Region
• Sea Level Rise Projections for the Northeast
Changes in
Nor’Easters,
tropical
cyclones?
Comparison of three sea level rise projection methods for the New York Metropolitan Region, for the
2080s relative to 2000-2004 (Horton et al. 2008, GRL; Horton et al. 2009 NPCC CRI, 2009; Horton et
al. JAMC in review).
CCRUN Prototype Projects: Climate
• Historical analysis of multi-century trends and
variability
• Statistically and dynamically downscaled
projections at seasonal to multi-decadal
timescales
• Extreme events
• Sea level rise projections
• Stakeholder-driven analysis and presentation
Downscaling Products and Techniques
• Single GCM gridbox approach
– Delta method applied to station data (NYC Panel on Climate Change)
• Bias-Correction Spatial-Disaggregation http://gdo-
dcp.ucllnl.org/downscaled_cmip3_projections/ (BCSD; Maurer, et al.
2007, based on Wood et al. 2002, 2004, and Maurer 2007)
– Direct use of monthly time series for impact models
– For daily projections at station level, we use modified delta method and random
sampling
• RCM simulations from the North American Regional Climate Change
Assessment Program archive (NARCCAP; Mearns et al. 2009, EOS)
– Changes in frequency and duration of key extremes
– Changes in (intra-annual) variability more generally
– Delta method
Climate Hazard Information
Process used to develop climate risk factors for New York City
• Projection Range, based on 16 GCMs and 3 SRES scenarios
• Time slice experiments based on single GCM gridbox, delta
method approach
• Key thresholds
– Number of days below 32 °F (transportation sector)
– Number of days above 90 °F (energy and health
sectors)
– Number of intense precipitation events (e.g., .5
in./day; water sector)
• Qualitative projections
Statistical Downscaling
Bias Correction Spatial Diaggregation (BCSD) Method--Steps
Bias correction
• Regrid 1/8º observations and GCM outputs to 2º resolution
• Generate monthly cumulative distribution functions at 2º, for
(observations, the baseline GCM, and the future simulation)
• Quantile map the baseline GCM onto the observed data (note: this
preserves higher order statistics)
Statistical Downscaling
Bias Correction Spatial Disaggregation (BCSD) Method--Steps
Spatial Downscaling
• Rank each future month based on baseline GCM CDF, then replace with observed
data as above
• Generate factor field (temperature differences, precipitation ratios)
• Interpolate these factors to 1/8º
• Apply the downscaled factors to the observed 1/8º data (for temperature this is
addition; for precipitation this is multiplication
NARCCAP Simulations
• RCMs were run at 50km resolution for three experiments:
– NCEP Reanalysis-driven, December 1979-November 2000:
How well do the RCMs simulate ‘observations’ over the NE, when driven by
‘perfect’ boundary conditions?*
– GCM hindcast-driven, December 1970-November 2000
How sensitive are the RCM results to ‘biases’* in the driving GCMs?*
– GCM future-driven, December 2041-November 2070, A2 SRES Scenario
How does the forcing associated with greenhouse gases and other
radiatively important agents manifest itself at more local scales?*
*Several caveats here…
Annual Precipitation (in./day), 1990-1999
How well do the RCMs
simulate ‘observed’ mean
precipitation over the NE,
when driven by ‘perfect’
boundary conditions?
Precipitation difference (%), Hindcast-driven minus R2-driven
How sensitive are the RCM
results to ‘biases’* in the
driving GCMs?*
BCSD
Projections--Mean Annual Changes
SRES A2 2050s minus 1980s annual temperature (°F)
Projections--Mean Annual Changes
SRES A2 2050s divided by 1980s annual precipitation (%)
Projections--Mean Changes
NYC Gridbox Mean Annual Mean Annual
(# of Temperature Precipitation
simulations) Change (°F) Change (%)
2.5 to 6.1 -9 to 10 (5)
GCM (16)
(4.1)
2.9 to 5.7 -6 to 23 (5)
BCSD (16)
(4.2)
NARCCAP 4.3 to 5.9 0 to 14 (5)
(4) (4.6)
SRES A2-driven 2050s divided by 1980s GCM hindcast-driven
Projections--% Change in # days with >.5 in prcp
SRES A2-driven 2050s divided by 1980s GCM hindcast-driven
Projections--Mean Annual Changes
SRES A2 2050s divided by 1980s annual precipitation (%)
Projections--Interannual Variability
BCSD
SRES A2 2050s divided by 1980s, standard deviation of annual temperature (°F)
GFDL/RCM3 Growing Season Changes, A2 2050s relative to 1980s
Mean T % Change in
Change (C°) SD of Daily
Temperature
Mean P Change (%)
% Change in % Change
Alpha in number
parameter of of rainy
Daily P days
Conclusions and Future Work
• Conclusions
– The range of results associated with
downscaling technique is often smaller
than other sources of uncertainty that
influence decisions, such as:
• Global climate sensitivity
• Emergent patterns not captured by
global climate models
• Interannual to decadal variability
• Climate impacts
• Socio-economic changes
• Next Steps Neiman et al. 2008, Jour. of Hydromet.
– Analyze pre-conditions for extreme
precipitation events in GCMs and RCMs
– Explore standardized statistically
downscaled daily products (CMIP3 and
CMIP5) such as BCSD, Bias Corrected
Constructed Analogues (BCCA; Maurer
and Hidalgo, 2008)
– Develop localized downscaling tailored
to sector-specific stakeholder needs
Statistical Downscaling Model
(e.g., SDSM; Wilby et al. 2002)
Snow depth at Wanakena, NY based on SDSM
- Stochastic approaches, weather downscaling. (Tryhorn and Degaetano in
generators NYState CliMAID Report, to be released in 2011)
Flexible Adaptation Pathways
Source: NPCC, 2010
Key elements to achieve Flexible Adaptation Pathways are a guiding framework,
stakeholder engagement, expert knowledge providers, recurring assessment process,
Action Plans by decision-makers, and vertically/horizontally integrated projects with
25
ongoing evaluation