Only Detected few TCs from GCM results

[linva7]

Hi, everone:

Recently, i'm trying to use tempestextrems to capture tropical cyclones from cesm1.2 results. I restricted the WNP region and first removed the warm-core condition, only using the closed PSL low-pressure center for selection. The results obtained are already very few, and the latitudes and longitudes of the tracks are close to the boundary. I have provided my SH script, log file, and result files. My CESM spatial resolution is 1.9×2.5, and the time resolution is 6 hours. All variables are merged into one NC file for reading. I can see some signs of TC from Panoply, such as the results of a specific time for VBOT in the PNG image, but it seems I cannot capture them. I hope someone can help me identify the issue. Thank you!

cesm_fw5_2018_JJASO_tracks.txt
cesm_test.log
cesm_test.sh

[linva7]

@paullric

[zarzycki]

~2deg resolution is extremely questionable for TC tracking (both from a grid spacing standpoint and the model's capacity to simulate processes that approach TC dynamics). My strong suggestion would be to use finer resolution data if possible.

bin/DetectNodes \
    --in_data_list ${TEMPEST_DIR}/cesm_file.txt \
    --timestride 1 \
    --out ${OUTPUT_PREFIX}_candidates.txt \
    --searchbymin PSL \
    --mergedist 8.0 \
    --closedcontourcmd "PSL,200.0,6,0" \
    --outputcmd "PSL,min,0;_VECMAG(UBOT,VBOT),max,2;PHIS,min,0" \
    --minlat 0 \
    --maxlat 70 \
    --minlon 100 \
    --maxlon 190 \
    --regional \
    --verbosity 3

In your script you are looking for local minima in SLP of 2mb (hPa) over 6 degrees, which for this model is essentially a delP of 2hPa between grid cells.

You may have success tracking weak minima by relaxing this criteria

    --closedcontourcmd "PSL,50.0,8,0" \

You could then investigate the candidates file to see if TE is detecting your requested storms at those timesteps.

[paullric]

To emphasize @zarzycki -- your data resolution is too coarse to effectively capture tropical cyclones. You should be using data that is at most 0.5 degree, but 0.25 degree is recommended. Additionally, this higher resolution data needs to come from the model directly -- interpolating the data to higher resolution won't provide any benefit.