A Comparison of Flare Forecasting Methods II: Data and Supporting Code

This repository contains the forecasts, event lists, and supporting code to replicate and re-evaluate the data and results from "A Comparison of Flare Forecasting Methods II" and subsequent referring papers. The forecasts are provided as both .csv and IDL ".sav" files with the naming convention of "SS_release.[csv,sav]" where SS is a code that refers to the provider, as listed in Paper II Table 1. Questions about specific methods should be directed to the author representing that method. The testing interval covered was 2016.01.01 -- 2017.12.31 inclusive (731 full-disk days). Two event definitions are used: C1.0+ (exceedance) / 0hr latency / 24hr validity and M1.0+ (exceedance) / 0hr latency / 24hr validity. Only those forecasts are available here; additional submitted forecasts (e.g. using different thresholds or validity periods) are not available except by direct request to the provider. A few forecasts were submitted for flare levels bounded on both upper and lower thresholds; these are the 'C_only', etc. They were converted to exceedance according to the appendix of Leka et al 2019. In those cases, both the original and the exceedence forecasts are included here. In the majority of the forecasts, the upper-bounded forecasts are arrays of "-1.0" which signals "no forecast". Both the forecast files (.csv or IDL .sav files) and the event-lists are suitable for the skill-score computing code "NWRA_skill_score_guts.pro". The latter spans a much larger time-span but dates are included; it thus provides the data for computing the longer-range climatology, as well.