Troubleshooting
Limits on the flux estimation
As noted in Section Estimating solar X-ray flux, the flux may be estimated within a factor of two. However, this depends on the availability of past data for different class of flares. If only M and X flares are used to build the probability density functions, then C-flares will be overestimated. If there is no data for strong X flares, then the flux for those flares may be underestimated. A way around this would be to generate synthetic VLF data, by pairing a D-region chesmitry code (WACCM, GPI) to a VLF propagation model (LMP, LWPC).
High breakpoint slopes, but DP is zero
It can happen that several breakpoints are detected in a quick succession for a single station, all with very high slopes but the DP values stay very low and the flux estimation as well.
The most likely reason for this is that the value of df for the station may not be correct anymore. This may be checked by plotting the unwrapped and detrended phase for the last day. In this case, the phase would be correctly unwrapped, but the detrend would induce a spurious slope. This would explain why the breakpoints’ slopes are very high, as the phase data would present this incorrect slope. Moreover, it explains why DP is close to zeros: though the values of the slopes are high, the data almost exactly follows this slope, so there is no phase increase.
This case would cause one alert to be sent for the first high slope detected, but no spurious flux estimation or incorrect electron density computation.
There were no breakpoint detected for more than two hours
This may be due to a very quiet day, for may be caused by a wrong df, inducing high slopes (and thus it is harder to detec a breakpoint). In this case, this may cause very high values of DP to be computed, which impacts the flux estimation.
The solution found was to automatically declare the station to be in quiet time if no breakpoints were detected in the last two hours. A new slope if computed based on the last 9 minutes of data (if it is usable). It is unlikely that this will cause a flare detection, as no new breakpoint (thus no sudden phase changes) occured for two hours before.
A flare occured just before midnight
(This only concerns receivers for which midnight occurs in daytime). One of the cases not taken into account is the following:
A flare happens before midnight
It decays, just enough that \(\Delta \Phi_{now} > \Delta \Phi/2\) before midnight
After midnight, it is still flare time, but the new
DPis much lower (as it is the maximum of phase in the data)The phase never decays enough to go back to flare time
This would cause the data for the station to never go to quiet time. BUT
This is a quite unlikely situation
This would only matter for stations for which midnight happens in daytime
If no new breakpoints is detected at any point during the day, the station is forced back into quiet times
For this situation to happens, DP would have to be low. Thus, there wouldn’t be a lot of impact on the flux estimation if there is no new flare during the day
If there is a flare during the day, DP would get higher. Thus, the criterion for reaching quiet time (\(\Delta \Phi_{now} < \Delta \Phi/2\)) would be easier to cross, and any new flare’s decay would be correctly treated
Correcting this would require reading the data for the previous day, which is longer, has more potential for failure and requires more informations from the user (because the data changes path from one day to the next). Thus, though it is possible to correct this, it is not a priority (and would only be tested in real-time next solar cycle anyways, because we don’t have an antenna for which midnight is in daytime anyways at the moment). If it is ever necessary though, it should be noted that it is doable, by reading the previous day’s data and finding DP for the flare, or keeping track of DP across a flare (not necessary otherwise)