LUSciD-LLNL UCSD/SIO Scientific Data Project: SIO LLNL SDSC Tim BarnettDoug RotmanReagan Moore David PierceDave BaderLeesa Brieger Dan CayanBen SanterAmit Chourasia Hugo HidalgoPeter Gleckler Mary TyreeKrishna AcutaRao Climate Studies

Objective Can we detect a global warming signal in main hydrological features of the western United States?

Program Elements Control run: Natural variability CCSM3 from NCAR on Thunder. (approx 4.5 TB) Downscaling: 12 km grid over west for spatial resolution (control+anthro; another 5 TB) Hydrological modeling: The downscaled data on rainfall, temperature, terrain, etc. force a hydrological model for time histories of steam flows and snow pack evolution in the western US (control+anthro: another 5 TB). Detection and attribution (D&A) analysis.

Deliverables Year 1 Complete a long GCM control run and begin statistical downscaling for selected geographic regions.…..DONE Begin VIC simulations with both downscaled data and PCM forced realizations.……………………….....…..DONE Implement a data grid linking resources between LLNL and SDSC. The data grid will be used to manage the simulation output that is generated.………………….…..DONE (1st order)

Deliverables (con’t) Year 2 Complete downscaling of Control. Complete VIC run on downscaled Control run. Prepare paper on downscaling intercomparisons Begin preliminary D&A analysis. Develop a digital library for publishing results, and integrating with PCMDI Year 3 Complete D&A analysis. Write a paper describing the results.

Change in California Snowfall

Change in Snow Water Equivalent Observed, Courtesy P. Mote

River flow earlier in the year

Runoff already coming earlier

Sierra snow pack Now and ………………….………….future?

Projected change by 2050

Global model (orange dots) vs. Regional model grid (green dots) Downscaling: the motivation

Key Question Do the signals we see happen naturally or are they human-induced? To answer, we need to know the levels and scales of natural variability in the western hydrological cycle.

Long GCM control run CCSM3 with finite volume dynamical core (“-FV”) Atmospheric resolution is 1.25 o x1 o with 26 vertical levels Ocean resolution is 320x384 stretched grid with 40 levels (so- called “gx1v3” grid; averages 1 1/8 o x0.5 o ) 760 years of a long pre-industrial control run transferred to SDSC

CCSM3-FV: Temperature, DJF

CCSM3-FV: Temperature, MAM

CCSM3-FV: Temperature, JJA

CCSM3-FV: Temperature, SON

CCSM3-FV: Precipitation, DJF

CCSM3-FV: Precipitation, MAM

CCSM3-FV: Precipitation, JJA

CCSM3-FV: Precipitation, SON

CCSM3-FV: Precip Variablity, DJF

CCSM3-FV: Precip Variablity, MAM

CCSM3-FV: Precip Variablity, JJA

CCSM3-FV: Precip Variablity, SON

CCSM3-FV and El Nino

CCSM3-FV and La Nina

Next steps with CCSM3-FV …(Tim: are we done with CCSM3-FV? Will it be dynamically downscaled? Will there be historical runs with anthro forcing?)

Statistical downscaling Uses “analogue” technique: Start with daily CCSM3-FV data on coarse grid, and daily obs. data on fine grid (Mauer et al. 2002; PRISM data disaggregated to daily level using daily obs) Coarsen obs to model grid Compare model field to coarsened obs 30 closest matches (least RMSE) and optimal weights found Weights applied to obs on original fine grid Hidalgo et. al 2006, J. Climate, submitted

Example: Jan 1 st, M.Y. 240 PREVISIONARY MAY HAVE ERROR

CCSM3-FV downscaling: Precipitation monthly EOFs

CCSM3-FV downscaling: T-max monthly EOFs

CCSM3-FV downscaling: T-min monthly EOFs

CCSM3-FV downscaled: Precipitation EOFs vs. Obs

CCSM3-FV downscaled: T-max EOFs vs. Obs

CCSM3-FV downscaled: T-min EOFs vs. Obs

Spectrum of Precipitation PC#1 Observed M.Y M.Y (x axis is cycles per month)

Spectrum of T-max PC#1 Observed M.Y M.Y (x axis is cycles per month)

Spectrum of T-min PC#1 Observed M.Y M.Y (x axis is cycles per month)

Runoff applied to river flow model

Sacramento River at Sacramento Columbia River at the Dalles Colorado River at Lees Ferry

Next steps with statistical downscaling Have processed ~100 yrs of the 760 yrs available Process rest of CCSM3-FV control run Evaluate observed changes in hydrology against this estimate of unforced variability

PCM/VIC runs (Andy Wood, UW) Historical simulations with estimated GHG and sulfate aerosols 3 ensemble members covering

PCM/VIC: Trend in Snow Water Equivalent

PCM/VIC: Trend in T-air

PCM/VIC: River flow

Amplitude shows strong decadal variability Phase shows flow earlier in the year for some, but not all, rivers

Next steps for PCM/VIC Process other ensemble members to reduce natural internal decadal variability Is the forced change statistically significant? How does it compare to observations?

Cooperative projects #1: SIO LLNL Tim BarnettKrishna AchutaRao David PiercePeter Gleckler Ben Santer Karl Taylor Ocean Heat Content

Motivation Can GHG and sulfate aerosol forcing explain the warming signal? YES! (surprisingly well)

PCM Signal Strength & Noise

HadCM3 Signal Strength & Noise

Inter-Model Comparison: N. Hem PCM signal + HadCM3 noise

What about other models? 38 realizations of 20th century climate from 15 coupled models in the IPCC AR4 archive are being analyzed. Work in progress Krishna AchutaRoa; David Pierce; Peter Gleckler; Tim Barnett

NCAR CCSM 3.0 MRI CGCM 2.3a

Preliminary findings Most models show a detectable warming signal in all the ocean basins with some exceptions NCAR CCSM 3.0 shows large natural variability in the North Atlantic Details of signal penetration in some ocean basins vary More complex picture than the previous study (Barnett et al. 2005) that considered two models Does the fidelity of model heat uptake relate to climate sensitivity?

Note: Plot shows only a subset of the 15 models analyzed. Heat uptake vs. climate sensitivity

Volcanic Eruptions and Heat Content P. Gleckler 1, K. AchutaRao 1, T. Barnett 2, D. Pierce 2, B.D. Santer 1, K. Taylor 1, J. Gregory 3, and T. Wigley 4 ( 1 PCMDI 2 UCSD/SIO, 3 U.Reading, 4 NCAR) How do volcanic eruptions affect ocean heat content? Can this give insight into how ocean heat content anomalies are formed and propagate?

Background Heat Content (10 22 J) Volcanic eruptions substantially reduced 20th Century ocean warming and thermal expansion. Recovery from Krakatoa (1883) takes decades. Effect of Pinatubo is much weaker than Krakatoa because it occurs against backdrop of substantial ocean warming. Models including V forcing agree more closely with late 20th Century observations than those without V Gleckler et al., Nature, 2006 KrakatoaPinatubo Depth (m) Heat Content Temperature

Work in progress: Evolution of the Krakatoa anomaly GISS-HYCOMNCAR CCSM3 6 realizations of CCSM3, GISS-HYCOM (and GFDL2.1) Large inter-model differences Uncertainties associated with external forcings, model physics and initial conditions S/N analysis (in-progress) should help isolate model responses to eruptions - necessary to evaluate realism Decadal average ocean heat content anomalies: zonally integrated cross- sections Heat content (10 16 Joules/m)

Cooperative projects #2: SIO LLNL JPL Tim BarnettPeter GlecklerEric Fetzer David PierceBen Santer Atmospheric water vapor

Water vapor a key greenhouse gas How well do models simulate it? New 3-D satellite data set available Compare to AR-4 model fields in PCMDI database

Annual mean: models too moist

Fractional errors greater with height

Seasonal cycle in North Pacific Blue = 10 yrs of model Red = 3 yrs of AIRS

Summary Goal: can we detect a global warming signal in main hydrological features of the western United States? Long CCSM3-FV control run for estimate of natural internal variability is done CCSM3-FV simulation comparable to other NCAR-series models Statistical downscaling to give river flow is progressing PCM runs give another estimate of natural varaibility, and also in this case of the forced signal Other cooperative projects (ocean heat content, atmospheric water vapor) progressing well