Skip to content

Commit

Permalink
Per #1809, doc updates.
Browse files Browse the repository at this point in the history
  • Loading branch information
JohnHalleyGotway committed Nov 15, 2021
1 parent 7bc3d70 commit a18b069
Showing 1 changed file with 4 additions and 2 deletions.
6 changes: 4 additions & 2 deletions met/docs/Users_Guide/tc-gen.rst
Original file line number Diff line number Diff line change
Expand Up @@ -6,14 +6,16 @@ TC-Gen Tool
Introduction
____________

The TC-Gen tool provides verification of tropical cyclone genesis forecasts in the ATCF file format as well at the probability of genesis forecasts. Producing reliable tropical cyclone genesis forecasts is an important metric for global numerical weather prediction models. This tool ingests probabilistic and deterministic model output post-processed by a genesis tracking software (e.g. GFDL vortex tracker) and ATCF format reference dataset(s) (e.g. Best Track analysis and CARQ operational tracks) and outputs categorical counts and statistics. The capability to modify the spatial and temporal tolerances that define a "hit" forecast is included to give users the ability to condition the criteria based on model performance and/or conduct sensitivity analyses. Statistical aspects are outlined in :numref:`tc-gen_stat_aspects` and practical aspects of the TC-Gen tool are described in :numref:`tc-gen_practical_info`.
The TC-Gen tool provides verification of deterministic and probabilistic tropical cyclone genesis forecasts in the ATCF file format. Producing reliable tropical cyclone genesis forecasts is an important metric for global numerical weather prediction models. This tool ingests deterministic model output post-processed by a genesis tracking software (e.g. GFDL vortex tracker), ATCF edeck files containing probability of genesis forecasts, and ATCF reference track dataset(s) (e.g. Best Track analysis and CARQ operational tracks). It writes categorical counts and statistics. The capability to modify the spatial and temporal tolerances when matching forecasts to reference genesis events, as well as scoring those matched pairs, gives users the ability to condition the criteria based on model performance and/or conduct sensitivity analyses. Statistical aspects are outlined in :numref:`tc-gen_stat_aspects` and practical aspects of the TC-Gen tool are described in :numref:`tc-gen_practical_info`.

.. _tc-gen_stat_aspects:

Statistical aspects
___________________

The TC-Gen tool processes both deterministic and probabilistic forecasts. For deterministic forecasts specified using the **-track** command line option, it identifies genesis events in both the forecasts and reference datasets, typically Best tracks. It applies user-specified configuration options to pair up the forecast and reference genesis events and categorize each pair as a hit, miss, or false alarm. As with other extreme events (where the event occurs much less frequently than the non-event), the correct negative category is not computed since the non-events would dominate the contingency table. Therefore, only statistics that do not include correct negatives should be considered for this tool. The following CTS statistics are relevant: Base rate (BASER), Mean forecast (FMEAN), Frequency Bias (FBIAS), Probability of Detection (PODY), False Alarm Ratio (FAR), Critical Success Index (CSI), Gilbert Skill Score (GSS), Extreme Dependency Score (EDS), Symmetric Extreme Dependency Score (SEDS), Bias Adjusted Gilbert Skill Score (BAGSS).
The TC-Gen tool processes both deterministic and probabilistic forecasts. For deterministic forecasts specified using the **-track** command line option, it identifies genesis events in both the forecasts and reference datasets, typically Best tracks. It applies user-specified configuration options to pair up the forecast and reference genesis events and categorize each pair as a hit, miss, or false alarm.

As with other extreme events (where the event occurs much less frequently than the non-event), the correct negative category is not computed since the non-events would dominate the contingency table. Therefore, only statistics that do not include correct negatives should be considered for this tool. The following CTS statistics are relevant: Base rate (BASER), Mean forecast (FMEAN), Frequency Bias (FBIAS), Probability of Detection (PODY), False Alarm Ratio (FAR), Critical Success Index (CSI), Gilbert Skill Score (GSS), Extreme Dependency Score (EDS), Symmetric Extreme Dependency Score (SEDS), Bias Adjusted Gilbert Skill Score (BAGSS).

For probabilistic forecasts specified using the **-edeck** command line option, it identifies genesis events in the reference dataset. It applies user-specified configuration options to pair the forecast probabilities to the reference genesis events. These pairs are added to an Nx2 probabilistic contingency table. If the reference genesis event occurs within in the predicted time window, the pair is counted in the observation-yes column. Otherwise, it is added to the observation-no column.

Expand Down

0 comments on commit a18b069

Please sign in to comment.