Visualizing Intermodel Comparison For Climate Simulations Xiaoying Pu

To predict the future, we build climate models to simulate the past

The Goal is intuitive visualization Q: What’s a good tool for climate science?

How to ❏. . . compare ❏. . . visualize ❏. . . interpret

Climate model grids are like pixels PSL = 992. 1 h. Pa @(lat, long)

SSIM is based on three metrics from skimage. measure import compare_ssim as SSIM( ,

Build a Distance Matrix Symmetric! for (i, j) in SSIM( : , ) 2

Metric Multidimensional Scaling For non-Euclidean distance at d n o i s men i

Observation A model run CESM GISS-E 2 -* Euclidean distance CVDP (n = 42),

Had. GEM 2 -ES Had. GEM 2 -AO “Changes in model configuration that do

Had. GEM 2 -ES Had. GEM 2 -AO Had. GEM 2 -CC ACCESS 1

“Common atmosphere and land surface” MPI-ESM-LR Nor. ESM 1 -ME “Shared atmosphere” CMCC-CM CCSM

Small Euclidean distance sometimes Climate model lineages Interpreting the 2 D projection Sanderson, Benjamin

IPSL-CM 5 A-MR IPSL-CM 5 A-LR Questions about the 2 D projection Sanderson, Benjamin

Observation CESM What do the axes mean? Questions about the 2 D projection Sanderson,

● Intuitive and information-rich ● Cannot answer all the questions Our thoughts on SSIM

Future Work: validation User study A new collage Hierarchical clustering Scree plot Dasgupta, Aritra,

Thank you! Acknowledgement ➔ Mentors: ◆ Rick Brownrigg ◆ Bill Ladwig ➔ Scientists: ◆

Just in case Time curve Questions? Figure 1. Behind the scene: a failed attempt

CVDP members (n = 42), PSL, 1 degree rectilinear grid SSIM, Metric MDS, without

CVDP members (n = 42), PSL, 1 degree rectilinear grid SSIM, Metric MDS, with

CMIP 5 members (n = 192), PSL, 1 degree rectilinear grid SSIM, Metric MDS,

MDS that shows temporal change “Time curves” for sea ice data If we don’t

Slides: 26

Download presentation

Visualizing Intermodel Comparison For Climate Simulations Xiaoying Pu, Bucknell University

To predict the future, we build climate models to simulate the past

Hmmm… CVDP (n=42) PSL annual mean

The Goal is intuitive visualization Q: What’s a good tool for climate science?

How to ❏. . . compare ❏. . . visualize ❏. . . interpret ~ 200 model runs Plus Time-series

Climate model grids are like pixels PSL = 992. 1 h. Pa @(lat, long) CMIP 5 20+ modeling groups ~200 model runs ~100 years CVDP Aggregates over time dimension Pick a variable Get 2 D array

SSIM is based on three metrics from skimage. measure import compare_ssim as SSIM( , # python ) Luminance Contrast Structure Mathspeak: Average Variance Covariance

Build a Distance Matrix Symmetric! for (i, j) in SSIM( : , ) 2

Metric Multidimensional Scaling For non-Euclidean distance at d n o i s men i High d a Distance / similarity preserved! from sklearn import manifold import seaborn as sns

Observation A model run CESM GISS-E 2 -* Euclidean distance CVDP (n = 42), Pressure at sea level 1 degree rectilinear grid SSIM, Metric MDS, with observation Interpreting the 2 D projection Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

Had. GEM 2 -ES Had. GEM 2 -AO “Changes in model configuration that do not influence the atmospheric diagnostics” Interpreting the 2 D projection Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

Had. GEM 2 -ES Had. GEM 2 -AO Had. GEM 2 -CC ACCESS 1 -3 ACCESS 1 -0 FIO-ESM Two institutions, Share a large fraction of code base BNU-ESM Interpreting the 2 D projection Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

“Common atmosphere and land surface” MPI-ESM-LR Nor. ESM 1 -ME “Shared atmosphere” CMCC-CM CCSM 4 Nor. ESM 1 -M Interpreting the 2 D projection Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

Small Euclidean distance sometimes Climate model lineages Interpreting the 2 D projection Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

IPSL-CM 5 A-MR IPSL-CM 5 A-LR Questions about the 2 D projection Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

Observation CESM What do the axes mean? Questions about the 2 D projection Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

● Intuitive and information-rich ● Cannot answer all the questions Our thoughts on SSIM + MDS Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "Addressing interdependency in a multimodel ensemble by interpolation of model properties. " Journal of Climate 28. 13 (2015): 5150 -5170. Sanderson, Benjamin M. , Reto Knutti, and Peter Caldwell. "A representative democracy to reduce interdependency in a multimodel ensemble. " Journal of Climate 28. 13 (2015): 5171 -5194.

Future Work: validation User study A new collage Hierarchical clustering Scree plot Dasgupta, Aritra, et al. "Bridging theory with practice: An exploratory study of visualization use and design for climate model comparison. " IEEE transactions on visualization and computer graphics 21. 9 (2015): 996 -1014.

Thank you! Acknowledgement ➔ Mentors: ◆ Rick Brownrigg ◆ Bill Ladwig ➔ Scientists: ◆ Adam Phillips ◆ Dennis Shea ◆ Ben Sanderson ◆ Doug Nychka ◆ Clara Deser ➔ @NCL Workshop ◆ Mary Haley ➔ And all the admins!

Just in case Time curve Questions? Figure 1. Behind the scene: a failed attempt

CVDP members (n = 42), PSL, 1 degree rectilinear grid SSIM, Metric MDS, without observation

CVDP members (n = 42), PSL, 1 degree rectilinear grid SSIM, Metric MDS, with observation

CMIP 5 members (n = 192), PSL, 1 degree rectilinear grid SSIM, Metric MDS, with observation (20 th. C)

MDS that shows temporal change “Time curves” for sea ice data If we don’t average over time. . . Bach, B. , Shi, C. , Heulot, N. , Madhyastha, T. , Grabowski, T. , & Dragicevic, P. (2016). Time Curves: Folding Time to Visualize Patterns of Temporal Evolution in Data. IEEE Transactions on Visualization and Computer Graphics , 22(1), 559– 56

Time curve

Just in case Time curve