freedryk / cloudtracker

Python algorithm for tracking cumulus clouds in a LES.

Cloud-tracker

Updated cloud-tracking algorithm for SAM

Example

To run the cloud-tracking algorithm, do

python -O -B -m kernprof -l -v run_cloudtracker.py > line_stats.txt

or, to run with a memory profiler instead,

python -O -B -m memory_profiler run_cloudtracker.py > memory_stats.txt

cloudtracker's Issues

create_new_clusters missing condensed only (no core) clusters

cluster_cloudlets.py
def create_new_clusters(cloudlets, clusters, max_id, MC):
line 186, missing:

    clusters[max_id] = cluster
    max_id = max_id + 1

Tested in 3 step BOMEX run. Cluster numbers up by almost 100%; cloud numbers up by 10%. Concluded we are missing mostly cloud noise.

Cloud times contain temporary variable

cloudtracker / cloudtracker / make_graph.py, lines 152-155.

IS:

        cloud_graphs.append((condensed_volume, subgraph))
        times = list(times)
        times.sort()
        cloud_times.append(tuple(times))

SHOULD BE (tested version):

    # If a graph exists less than 2 minutes, classify it as noise.     
    if len(condensed_time) < 2*60/MC['dt']:
        cloud_noise.append(subgraph)
    else:
        condensed_time = list(condensed_time)
        condensed_time.sort()
        core_time = list(core_time)
        core_time.sort()
        cloud_graphs.append((condensed_volume, subgraph, condensed_time, core_time))

# Cloud with largest condensed volume sorted largest to smallest; volume not 
# used any further
cloud_graphs.sort()
cloud_graphs.reverse()
cloud_times = [item[2:] for item in cloud_graphs]  
cloud_graphs = [item[1] for item in cloud_graphs]
if full_output: full_output(cloud_times, cloud_graphs, merges, splits, MC)

return cloud_graphs, cloud_noise