Name Last Update
README.md Loading commit data...

README.md

MEOPAR Winter School in Environmental Modeling Notes

Participant intros

Jenna Joyce: UoOttawa w/ Jackie Dawson

  • Shipping in the Arctic (Corridors)
  • Corridors & Environment/Culturally significant areas coexistence

Ewelina Luczko: Baird & associates "where land meets water"

  • Coastal engineering - wave modeling
  • Wave energy generation in SWAN

Lindsay Chipman: UoColorado - Boulder

  • Cycling of O2 and C in permeable sediment
  • O2 flux from

Lei Ren: NUI Galway

  • Surface flow fields from CODAR & Model
  • Focus area: West Coast of Ireland

Russel Glazer: UoFlorida - Tallahassee

  • Saturation over ice & water in models
  • Cloud formation scheme

Blanche St-Béat: ULaval - Qc

  • Food Webs: distinguish ecosystemséestablish stability
  • Resilience vs Resistance: Wood house is better than hay or bricks?!

Feifei Sun: Andrea Scott student

  • Predict ice thickness from data
  • New assimilation methods

Dennis Monteban: Denmark university

  • Study of Fjord west of Greenland / Wave-Ice damping
  • Validated & Calibrated MIKE model

Ben Moore-Maley: UBC w/ Susan Allen

  • 3D modeling of Salish Sea @ 500m resolution
  • Wind driven circulation/interactions

Nancy Chen: DFO - St-John's

  • Satellite SSH anomaly -> water transport
  • Labrador & Scotian Shelf

Deborah Benkort: PhD Laval

  • Krill aggregation & dispersion in GSL/GSE
  • Effect of envt. on growth, distribution, reproduction

Becky Segal: MSc UVic

  • Permafrost thaw sumps work before
  • Ice prediction/obs and creating relevant products for communities

Charles Brunette: McGill with Bruno

  • Predictability using Lagrangian methods
  • Seasonal & Regional
  • Later formation -> Thinner ice -> Minimum ice extent

Onur Bora: Coastal engineer/PhD student in Istanbul

  • Hydrodynamics effects from shipping -> impact on sediment
  • Water cooling/intake system design

Olivier Asselin:

  • Figure out the atmosphere
  • Energy constant vs \lambda for different processes: 2 slopes for everything?!
  • CFD

Nathan Grivault: UoAlberta w/ Paul Myers

  • Freshwater in the Arctic: Arctic -> Lower latitudes
  • Impact on circulation: Export vs Forcing

UQAR People

Claudia Carascal:

  • Satellite validation of reflectance
  • Adjust for atmosphere/ocean to get buoy value

Jean-Luc Shaw: w/ Daniel and Dany

  • Hydrodynamics of the Bay of Sept-Iles
  • Numerical model & validation to support ecological indicators

Eliott Bismuth: w/ Dany

  • Wave-ice interactions & shore protection
  • Now erosion as an RA
  • Wave data/modeling/validation + XBeach

Sandy Gregorio: w/ Zhigang Xu (DFO)

  • Modeling storm surge & tsunami
  • Atmosphere: frequency of storms from low P events
  • ocean to get impact

Gwenaëlle Gremion: w/ Dany

  • Biological carbon pump in NOW polynia
  • Understand/represent carbon flux under water (and increased C at depth?)
  • BGCModel + PO to assess Bio + Physics effects

Jeremy Beaudry: w/ Dany

  • Wave attenuation / ice breakup
  • 1D floe size & thickness distribution for 2D WW3

Manu John:

  • Wave-Ice interactions for regional modeling
  • HF radar

Sebastien Dugas:

  • Wave-floe size/concentration relationship

Michel Tantare:

  • Drift prediction/comparison
  • Eullarian vs Lagrangian methods for drift?

Jean Chavy:

  • Energy transfer/KE cascade
  • Combine radar wind obs to get winds

Essi Aboyo:

  • Dynamics of the NOW polynia
  • Idealized model for climate change

Climate Scenarios

Climate Scenarios definition

Plausible trajectory for one or more climate variables over some region and period.

  • Statistical equivalence with a selected set of reference (ie it starts by properly reproducing ~30 years of the past)
  • Continuity from "warm up" to forecast
  • Physical credibility
  • Plausible forcing scenario

Basic variables: time series or anomalies Derived indicators of:

  • distributions (% of dry days, growing degree days, etc.)
  • sequence (# of 5 days series of XYZ)
  • correlation Spatial analogues (find somewhere that currently has the projected climate for a location)

Since this is stats based, often run ensembles

Concepts not to confuse

Emission scenario: Plausible trajectory for real human emissions (GHG, aerosols) used to force climate models Climate simulation: Output of a climate model run with an emission scenario Climate scenario: Plausible trajectory for the Earth climate based on climate simulation(s) but potentially corrected/processed.

Obs -> (interactions with climate simulations) -> Reference -> post-processing->Scenario Emission scenario -> models -> climate simulations -> (interactions with Obs) -> long-term trends -> post-processing -> Scenario

Many typical users types

Don't use multiple-model averages because averaging often kills variability and tend to get misunderstood as a robus forecast when it is not.

Many uncertainty sources:

  • Anthropogenic future forcing (impacts after ~2040 only, but then becomes the most important effect)
  • Natural future forcing (volcanoes, etc.)
  • Models (structure and parameters)
  • Chaotic nature of the system (mostly impacts short term trends, "averages" over longer scales)
  • Post-processing

=> It's important to provide people with uncertainty ranges for reducing risks/cost or under/over-adaptation. Don't focus just on the most possible warming, nor the least possible warming, nor average, give range.

Output clustering:

Take many model runs, cluster them by "similarity of climate scenario" to minimize the runs but still represent uncertainty. Otherwise, if needs prevent proper clustering have to assess what

Statistical post-processing

  • If simulations aren't close to obs for past climate, hard to use.
  • If simulations physics don't track well
  • Want good tracking of past climate and small bias

Can't only use simulations that reproduce runs that reproduce obs without bias because there will always be bias somewhere.

Quantile mapping

1 Compare quantiles for a given variable between sims and obs 2 Build a transfer function to make sims closer to reference products 3 Apply transfer functions to projected climate to get good tracking Note: Pay attention to values past the 1st and 99th percentile because future climate may have more "extreme" values

Build a transfer function for each day individually using that specific day and a range of days before/after but no range in space. Year average transfer function means seasonally changing bias aren't improved. Monthly transfer functions create discontinuities in the time series at the month change.

Taking space correlation and inter-variable correlation is possible, but a lot costlier. Lots of challenges, like how to treat winds (vector, 2x scalar components), rain (lots of 0 on sunny days?)

Statistical post-processing controversies

Physical incoherence
  • Type 1: univariate impossibility (ex.: Relative Humidity of 110% is not possible, but the transfer function can create it)
  • Type 2: multivariate impossibility (ex.: T_min > T_max after post-processing. Have to change the method: Post-process T_max with additive correction and Daily Temperature Range with multiplicative correction, then deduce T_min. ex.2: High radiative forcing & rain, meaning somehow the clouds were transparent?! No solution except awareness)
  • Type 3: multivariate correlation (Actually if multivariate correlations are significantly different from obs, it's usually from the model)
SPP vs RCM
  • Drop SPP and just look at RCM? Not really, because they are complementary next steps after global models
  • Only look at RCM and use GCM to force them only? Not really, often RCM cannot properly represent the spread of uncertainty so you use GCM to cover that. Other times (ex.: coastal and montainous areas) RCM make a lot more sense so use them.

Questions?

1 - No 2 - No 3 - Yes 4 - ???