in Cloudy Atmospheres Alexandra L. Jones Blue Waters Graduate - - PowerPoint PPT Presentation

in cloudy atmospheres
SMART_READER_LITE
LIVE PREVIEW

in Cloudy Atmospheres Alexandra L. Jones Blue Waters Graduate - - PowerPoint PPT Presentation

High Accuracy 3D Radiative Transfer in Cloudy Atmospheres Alexandra L. Jones Blue Waters Graduate Fellow Advisor: Larry Di Girolamo Department of Atmospheric Sciences, University of Illinois, Urbana-Champaign Blue Waters Symposium May 12,


slide-1
SLIDE 1

High Accuracy 3D Radiative Transfer in Cloudy Atmospheres

Alexandra L. Jones Blue Waters Graduate Fellow

Advisor: Larry Di Girolamo

Department of Atmospheric Sciences, University of Illinois, Urbana-Champaign

Blue Waters Symposium May 12, 2015

aljones4@Illinois.edu

slide-2
SLIDE 2

Importance of Radiation

2

Why is radiation in a cloudy atmosphere important?

Adapted from graphic by NOAA

MISR, 2007

qld.gov.au

slide-3
SLIDE 3

Current Representation of Radiation

What’s the problem?

3

W×ÑI(r,W) = -se(r)I(r,W)+s a(r)B(T(r)) + s s(r) 4p p(r,W, ¢ W )I(r, ¢ W )d ¢ W

4p

ò

ICA PP 3D

slide-4
SLIDE 4

Goals

 Produce Highly Accurate Benchmark Simulations

 Quantify bias in our approximations  Improve simpler/faster parameterizations

 3D Broadband Monte Carlo Community Model

 Faster science progress  Starting point: I3RC

4

What’s the problem?

slide-5
SLIDE 5

Model Development: Spectral Integration

How do we do better? The I3RC

5

Spectral Integration Solar, thermal broadband I3RC Solar Only Source Function Domain for each λ β, ω, Ρ, x, y, z, A,T, λ

slide-6
SLIDE 6

Comparing to Analytical Solution

 TOA input flux linear

function of wavelength

 𝐺↓𝑈𝑃𝐵 λ = 𝑐λ

 Absorption coefficient

linear function of wavelength

 𝛾𝑏 λ = 𝑏λ

6

How do we do better? The I3RC

𝐺↓𝑡𝑔𝑑 = −𝑐 𝑏∆𝑨 2 𝑓−𝑏λ𝑔∆𝑨 𝑏λ𝑔∆𝑨 + 1 − 𝑓−𝑏λ𝑗∆𝑨 𝑏λ𝑗∆𝑨 + 1

slide-7
SLIDE 7

Model Development: the Real Atmosphere

How do we do better? The I3RC

7

Spectral Integration Solar, thermal broadband I3RC Solar Only Source Function HITRAN S ARTS Model CIRC T, P , z, VMRS SSP Table σ, ω, Ρ, A, λ Absorption Tables σ SSP Table Creation Tool Domain T, P , x, y, z

slide-8
SLIDE 8

Checking Transmittance

How do we do better? The I3RC

slide-9
SLIDE 9

Real Atmosphere: Clear-sky Absorption

9

How do we do better? The I3RC

LBLRTM Flux I3RC Flux

slide-10
SLIDE 10

Reducing MPI-communication Time

FORMER

 Master assigns lesser of

#perBatch or total remaining photons in bin to each worker

 Workers trace those

  • photons. Then ask for

more work to do

 Repeat until no photons

remaining

1 2 4 5 8

How does it perform?

slide-11
SLIDE 11

How does it perform?

Reducing MPI-communication Time

CURRENT

 Master assigns #perBatch

photons even if spanning multiple bins

 Workers trace those

  • photons. Then ask for more

work to do

 Repeat until no photons

remaining

50 50 50 50 50

92% improvement in total compute time

slide-12
SLIDE 12

Acknowledgements and Questions

This research is part of the Blue Waters sustained-petascale computing project, which is supported by the National Science Foundation (awards OCI- 0725070, ACI-1238993 and ACI-14- 44747) and the state of Illinois. Blue Waters is a joint effort of the University

  • f Illinois at Urbana-Champaign and its

National Center for Supercomputing Applications.

This work has also been funded by the NASA Earth and Space Science graduate fellowship program.

12

aljones4@illinois.edu

**List of references can be found on my poster**