1 of 16

Paul Ullrich

Climate Resilience Lead�Lawrence Livermore National Laboratory

Professor of Regional and Global Climate Modeling�University of California, Davis

Requirements for

Storm Tracking

2 of 16

A highly extensible and multifaceted framework for rapid feature detection, tracking, and scientific analysis of regional or global Earth system datasets supporting native grid systems.�

conda create -n "tempest"�conda activate tempest�conda install -c conda-forge tempest-extremes

Method 1: From conda

Method 2: Build from source

https://github.com/ClimateGlobalChange/tempestextremes

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

3 of 16

Colin Zarzycki, Penn State	Co-developer; Tropical cyclones and extratropical cyclones (and metrics)
Yushan Han, UC Davis	Low-pressure systems
Elizabeth McClenny, UC Davis	Atmospheric rivers
Marielle Pinheiro, UC Davis	Atmospheric blocking (Z500 exceedance, geopotential gradient, PV anomaly)
Alan Rhoades, LBNL	Atmospheric rivers
Kevin Reed, Stony Brook	Tropical cyclones and precipitation
Alyssa Stansfield, Stony Brook	Tropical cyclones and precipitation
Karthik Balaguru, PNNL	Tropical cyclones (in E3SM)
Malcolm Roberts, UKMO	Tropical cyclones (in HighResMIP)
Daehyun Kim and Yumin Moon, UW	Tropical cyclones
William Boos and Vishnu Nair, UCB	Monsoonal depressions
…others?

Contributors and Collaborators

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

4 of 16

TempestExtremes Features

Unstructured grid support: Support for feature identification and tracking on arbitrary unstructured / native grids.

Parallelization Over Files: Automatically distribute feature detection scripts over multiple files using MPI.

Command line configurability: No hard-coded thresholds or constants. Algorithms designed to maximize robustness for multiple features.

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

5 of 16

Tropical + Extratropical Cyclones

TempestExtremes v2.1 criteria:

Only deepest local minima within 6 degrees GCD retained
psl must drop by 200Pa within 5.5 degrees GCD
z300-z500 geopotential layer thickness must drop by 6m within 6.5 degrees GCD of layer thickness maximum (within 1 degree of PSL minimum)
10m wind speeds must be >= 10m/s for at least 60 hours
Storm must be equatorward of 50 degrees latitude for at least 60 hours
Storm must have elevation (zs) below 150m for at least 60 hours

Variables Needed:

At minimum data available at 6 hourly frequency
psl, z300, Z500, 10m wind speed, zs
Instead of z300-z500, t400 (or an approximation) can be substituted

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

6 of 16

Nodal Features

Figure: The track of the strongest tropical cyclone (TC) in the high-resolution E3SM, produced using TempestExtremes. Background colors represent cooling of sea-surface temperatures induced by the storm. Figure courtesy of Karthik Balaguru, PNNL.

Figure: Observed overland precipitation extracted using TempestExtremes from tracked storms using (left) radius of 8m/s wind and (right) a fixed 500km radius. Figure courtesy of Alyssa Stansfield, Stony Brook.

Needs precipitation (pr)

Needs sea surface temperatures

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

7 of 16

Feature-centered snapshots or composites can be easily generated within TempestExtremes once features are tracked

Figure: Severe northern hemisphere extratropical cyclone composites.

Nodal Features: Composites

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

8 of 16

System for Classification of Low-Pressure Systems (SyCLoPS)

Han and Ullrich (2024) Manuscript:

https://doi.org/10.22541/essoar.171319442.20959523/v1

ERA5 Catalogue of classified LPSs:

https://zenodo.org/records/10906285

TC

STLC(SS)

PL(PTLC)

MS

TLO/TD

TLO/TD*

SC

EX

DS

THL

DOTHL

HAL

Required variables:

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

9 of 16

DetectBlobs�--timefilter "6hr”�--thresholdcmd "_LAPLACIAN{8,10}(_VECMAG(VIWVE,VIWVN)),<=,-20000,0”�--minabslat 15

--geofiltercmd "area,>=,4e5km2"

Data operators that process data on the fly (here, computing a Laplacian of a vector magnitude).

Time subsetting and filtering

Geometric filters (e.g., minimum blob area)

Areal Features: Atmospheric Rivers

TempestExtremes v2.1 criteria:

8-point Laplacian of IVT within 10 degrees GCD must be less than -20000 kg/m/s/deg²

Variables Needed:

Integrated vapor transport (IVT)

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

10 of 16

Areal Features: Atmospheric River Tracking

Figure: Temporal tracking of an atmospheric river object through its lifetime using an overlap criteria and TempestExtremes’ StitchBlobs functionality. Figure courtesy of Alan Rhoades.

Nodal locations represent the mass center of the binary mask.

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

11 of 16

Areal Features: Atmospheric Blocks

Below: Climatology of atmospheric blocking events in MERRA2 data, defined as regions of persistent positive geopotential anomalies at the 500hPa level.

TempestExtremes v2.1 criteria:

z500 exceeds mean plus 1.5 times standard deviation of climatological z500

Variables Needed:

z500 at daily frequency

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

12 of 16

Mesoscale Convective Systems in PyFlexTrkr

PyFLEXTRKR is an open-source Python framework to track 2D atmospheric objects
Specialized capability to track convective systems in high-resolution datasets

Works with multiple variables
Multi-object identification algorithms
Treats merging/splitting
Optimized for large datasets with scalable parallelization
Visualization, statistical analysis notebooks

Variables Needed

Preferred at hourly frequency
precipitation (pr) and outgoing longwave radiation (rlut)

Observed Tracks

Simulated Tracks

Available on GitHub

Feng et al. (2023), GMD

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

13 of 16

Other Unstructured Grid Tools: ncvis

A new visualization package for “quick look” examination of structured and unstructured climate and weather data.

Available on conda:

conda install conda-forge::ncvis-climate

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

14 of 16

Several parallelized operators for processing large climate datasets:

Long-term daily, monthly or seasonal mean climatology
Fourier filtering in space and time
Quantile / percentile calculator
High-throughput data processing

Figure: E3SM High-Res 95^th percentile summertime daily maximum temperature

Figure: 500hPa Geopotential Height blocking threshold produced using long-term daily mean, data processing, Fourier filtering operations.

Data Analysis Operators

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

15 of 16

Summary

Cyclonic Storms (TEv2.1)

@ 6-hourly frequency

psl

z300, z500

10m wind speed

zs

Substititons

t400 for z300/z500

Atmospheric Blocks

@ daily frequency

z500

Atmospheric Rivers

@ 6-hourly frequency

integrated vapor transport

Mesoscale Convective Systems

@ hourly frequency*

pr

rlut

* 3-hourly or 6-hourly frequency could be used, but tracking becomes difficult

SyCLoPS Framework

@ 6-hourly frequency

u925, u850, u200

v925, v850, v200

rv500

t850

hur850, hur100*

z925, z700, z500, z300

* daily or monthly frequency�could be used

Paul Ullrich

October 22^nd, 2024

Storm Tracking for GCRMs

16 of 16

Paul Ullrich

Climate Resilience Lead�Lawrence Livermore National Laboratory

Professor of Regional and Global Climate Modeling�University of California, Davis

Thanks!