Evolution of micrometeorological observations: Instantaneous spatial and temporal surface wind velocity from thermal image processing. GeoComputation 2019
conference contributionposted on 2019-12-01, 23:24 authored by Benjamin Schumacher, Marwan Katurji, Jaiwei Zhang, Ivana Stiperski, Christina Dunker
Pattern correlation techniques are commonly used in Particle Image Velocimetry studies to calculate tracer particle velocity in fluids or gases. The techniques can be further improved and used to track landscape scale moving thermal patterns in time sequenced infrared images to estimate a spatial advection velocity field.
This study presents a validation and parametric sensitivity test of commonly used image correlation techniques using data produced from a numerical weather model and field observations from a longwave infrared camera. The numerical simulations aimed to create three different wind speed pattern cases to analyze the performance of four different image velocimetry techniques and compare the estimated velocity field to the model's velocity field. A sensitivity test was carried out with various combinations of required user input to find the best performing image velocimetry techniques. These were then applied to a dataset of surface temperature of an artificial hockey turf field taken with an infrared (IR) camera to test the algorithms in real world conditions and to directly compare the estimated velocity to in-situ measured wind velocity.
The velocimetry algorithms are especially accurate when there is low complexity in ow structure (root mean square error: 0.4: Structural Similarity: 0.81). The comparison of the IR image velocimetry with the in-situ measurements on the hockey turf field shows accurate representation of the 10 minute mean wind direction and the mean wind speed with a maximal absolute error of 0.8 for the wind direction. If the comparison is done on a one minute average basis the accuracy of the image velocimetry decreases when the wind speed drops.