1. Cold Air Outbreak: Constrain Case Study
Kirsty McBeath, Paul Field, Richard Cotton, Humphrey Lean, Emilie Carter

2. Table of Contents
Introduction
Overview of radar observations
Cluster analysis of data
Comparisons of radar and Unified Model data
Variations applied to the UM
High resolution UM versions for this case

3. Constrain Overview
Constrain campaign took place in January 2010, based in Prestwick
Led by Richard Cotton and Paul Field
Aimed to Constrain constants used in model parameterisations
Variety of cloud-based sorties flown
5 x Cirrus spiral flights
4 x Cumulus cold air outbreak flights
2 x Stratocumulus/Altostratus flights

4. Introduction MODIS 31st Jan 2010
Looking at cases of cold air outbreak in the Northwest approaches
4 flights during Constrain examined these conditions during January 2010
Radar data from these cases used for comparisons with UKV model
This data is from a case on January 31st 2010 which coincides with flight b507 of the BAe-146 research aircraft
MODIS 31st Jan 2010

5. Overview of radar observations

6. Composite radar data 1.91 ±0.36km
Data available every 5 minutes for 24 hour period
Scans performed at a range of elevation angles: 0.5°, 1.0°, 1.5°, 2.5°
4 scan angles intercept cells at different distances from radar
Data from 4 scan angles combined to produce one dataset which captures all cells
0.5°: km
1.0°: 54-85km
1.5°: 32-64km
2.5°: 30-42km
150km
1.91 ±0.36km

7. Reflectivity and rain rate
Conversion from radar reflectivity (dBz) in to rain rate (mm/h), is typically approximated by:
Z = 200R1.6
This relation includes assumptions about drop size distributions and phase of precipitation
This conversion can introduce additional uncertainties into the radar data
30dBz≈66mm/day
20dBz≈16mm/day
10dBz≈4mm/day
5dBz≈1mm/day

8. Model data

9. Model data
Comparison done between radar data and UKV* (1.5km resolution, stretched to 4km at edges) model output.
Region examined is too far North for UKV (edge effects close to boundaries)
Nested version of UKV model set up to cover the region without any edge effects
Covered region from:
14.35W – 0.66E & 55.10N – 68.50N
UKV*
UKV

10. Model Reflectivity Radar Model
Reflectivity values for model data computed using model microphysics data, this reduces processing done on radar data and removes assumptions used when converting reflectivity to rain-rate

11. Cluster analysis

12. Cluster Analysis
10dBz (~4mm/day) threshold used to select regions of precipitation in both datasets

13. Cluster Analysis Identified cells tracked in time for both datasets
Whole frame advected to find overlap between cells
Fractional overlap for calculated for each overlapping pair of cells
Overlap threshold used to determine if identified cells are the same
Cells excluded from dataset if they moved outside the region of interest, or were only seen to decay
133 cells tracked in radar data
75 cells tracked in model data

14. Results of radar/model comparison using cluster analysis

15. Comparison of cell statistics for radar and model
Model has more long lived cells than seen in radar
And more cells with long decay times than radar
Mean cell lifetime
Radar: 54±3 minutes
Model: 78±6 minutes
Mean cell decay time
Radar: 28±2 minutes
Model: 44±4 minutes

16. Comparison of cell statistics for radar and model
Model shows good agreement with radar for cell diameter values.
Examining cell area shows that the model is producing more large cells than are seen in the radar data.
Ratio of cell area and enclosing circle area examined.
This Fill Fraction parameter used to examine cell shape.
This provides a measure of cell circularity
Mean fill fraction values
Radar: 0.50±0.01
Model: 0.77±0.01
Model produces more circular cells than are seen in the radar data
Cell area
Cell diameter

17. Variation in area with lifetime
UKV resolution (1.5km) may be too coarse to resolve the cells in this case.
Larger cell areas present throughout cell lifetime
More pronounced for cells with lifetimes over 1 hour
Little change in cell fill fraction seen with lifetime

18. High resolution work

19. High Resolution Model Work
The difference in cell shape between the model and radar was still seen when model physics was modified
This difference may be due to this case exhibiting characteristics at the limits of UKV resolution.
To examine the impact of model resolution, as series of nested runs has been run with resolutions of 500m, 200m and 100m.
Runs have also been done with increased vertical resolution of 140 levels (UKV has 70 levels)
500m
200m
100m

20. Initial results
500m
200m
100m
70 level model runs

21. Initial results
500m
200m
100m
140 level model runs

22. High Resolution Data Clustering
Increasing model resolution appears to produce an improvement in cell area
Cell fill fraction values reduce with increasing resolution, a bit too much at 100m
First look: cells get smaller and more circular with increasing resolution

23. Next steps Further analysis of high resolution model data
Compare this case to work done by Mesoscale Modelling Group at
Comparison to (available) aircraft data for this case
Repeat analysis for PikNMix case study (more aircraft data available)
Use high resolution model to asses cases during COPE (summer 2013)

24. Questions