Photogrammetric Analysis of Rotor Clouds Observed during T-REXUlrike Romatschke1, Vanda Grubišić1, and Joseph A. Zehnder2

1National Center for Atmospheric Research, 2Southwest Weather Consulting, LLC

IntroductionDigital stereo photgraphs collected during the T-REX field campaign are used to:● Develop algorithms for camera calibration, automatic feature matching, and reconstruction of 3D cloud scenes.

● Track the 3D path of cloud fragments at the upstream edge of rotor clouds and monitor their growth and dynamics until their merger with the main cloud.

Data

Camera calibration, feature matching, and 3D cloud reconstruction algorithms

Results: Statistics from 37 clouds

Conclusions

The T-REX field campaign took place in Owens Valley, CA, in the lee of the Sierra Nevada. ● IOP 6: March 25, 2006, 08:40-16:50 PST ● Cameras: Cannon 20D digital 8 Mpixel● Base line between cameras: 666m ● Shutter speed: ~10s

We study the evolution of cloud fragments which form upsteam of a largely stationary rotor cloud.

1. Lens distortion correction and image enhancement.

● Grubišić, V., and Coauthors, 2008: The Terrain-Induced Rotor Experiment: A field campaign overview including observational highlights. Bull. Amer. Meteor. Soc., 89, 1513–1533.

● Hu, J., A. Razdan, and J. A. Zehnder, 2009: Geometric calibration of digital cameras for 3D cumulus cloud measurements. J. Atmos. Oceanic Technol., 26, 200–214.

Algorithms have been developed for camera calibration, automatic feature matching, and reconstruction of 3D cloud scenes from stereo photographs. The resulting point clouds have:● high spatial resolution (~m) which is only limited by the pixel size and the distance to the cameras

● high temporal resolution (~10s)High resolution 3D point clouds allow detailed analysis of cloud evolution parameters such as growth, and horizontal and vertical movement.

2. Calibration of all six camera angles relative to each other by minimizing the distances (d) of “sightlines”.

3. Transformation of camera coordinate system (x, y, z) to world coordinate system (longitude, latitude, elevation) using real world reference point visible in both images.

5. Calculation of 3D coordinates of matched pixels via triangulation, and reconstruction of 3D point cloud.

4. Automatic matching of left and right image pixels of selected cloud feature.

Cameras

Sierra Nevada

Owens Valley

Camera A Camera B

Reference point Reference point

Reference pointImage from camera A Image from camera B Lon, lat, elev,

from Google Earth

d

Camera A Camera ACamera B Camera B

019 Jan22:00

20 Jan00:00

20 Jan02:00

20 Jan04:00

Time

0s 30s 70s

Image from camera A Image from camera B

-118.162

-118.160

-118.158

36.650

36.652

36.654

3800

4000

4200

4400

Latitude [°]

Long

itude

[°]

Elevation [m]

3800

4000

4400

4200

Ele

vatio

n [m

]-118.162 -118.160 -118.158

Longitude [°]

36.6

5436

.652

36.6

50

Latitude [°]

Sierra Nevada

Owens Valley

0s 10s 20s 30s 40s 50s 60s 70s

0

0.5

1

1.5

Hor

izon

tal t

rave

l dis

tanc

e [k

m]

0

50

150

250

Ele

vatio

n [m

]

100

200

0

0.2

0.6

0.8

Clo

ud a

rea

[km

2 ]

0.4

0s 10s 20s 30s 40s 50s 60s 70s

0s 10s 20s 30s 40s 50s 60s 70s

Hor

izon

tal s

peed

[m/s

]

0-10s 10-20s 20-30s 30-40s 40-50s 50-60s

0

10

20

30

40

-10

Ver

tical

spe

ed [

m/s

]

-5

0

5

10

0

0.5

1

1.5

Tra

vel d

ista

nce

of f

ront

and

bac

k ed

ge [

km]

Front

Back

0-10s 10-20s 20-30s 30-40s 40-50s 50-60s

0s 10s 20s 30s 40s 50s 60s 70s

Contact:romatsch@ucar.edu