Two defects affect interval-based outputs after resuming from a hotstart.
They do not alter the stored raster state, but they make reported mean rates and reporting diagnostics inconsistent with the period represented by their accumulators.
1. Unchanged Record Step
Consider a simulation with a one-hour record_step and constant rainfall of 10 mm/h:
00:00 initial report
01:00 regular report
01:30 hotstart
02:00 first report after resume
The accumulated rainfall at 02:00 correctly covers the interval from 01:00 to 02:00.
However, after resume the reporting interval can be calculated from the original simulation start at 00:00.
The 60-minute rainfall total is then divided by 120 minutes, and mean_rainfall is reported as 5 mm/h rather than 10 mm/h.
The next report timestamp remains correctly scheduled at 02:00.
2. Changed Record Step
Consider the same simulation, resumed at 01:30 with a new 10-minute
record_step:
00:00 initial report
01:00 regular report
01:30 hotstart and resume with a 10-minute record step
01:40 first report after resume
The first resumed report contains rainfall accumulated from 01:00 to 01:40,
which is 40 minutes of constant rainfall.
If the reporting interval is reset to the resume time, it is interpreted as only the 10 minutes from 01:30 to 01:40.
The reported mean_rainfall is then 40 mm/h instead of 10 mm/h.
Changing the cadence may legitimately anchor future output timestamps to the resume time,
but the interval used for accumulated outputs must match the data that has actually been accumulated.
The same inconsistency affects the reported average timestep and any other output derived from a reporting-interval total.
Two defects affect interval-based outputs after resuming from a hotstart.
They do not alter the stored raster state, but they make reported mean rates and reporting diagnostics inconsistent with the period represented by their accumulators.
1. Unchanged Record Step
Consider a simulation with a one-hour
record_stepand constant rainfall of 10 mm/h:The accumulated rainfall at 02:00 correctly covers the interval from 01:00 to 02:00.
However, after resume the reporting interval can be calculated from the original simulation start at 00:00.
The 60-minute rainfall total is then divided by 120 minutes, and
mean_rainfallis reported as 5 mm/h rather than 10 mm/h.The next report timestamp remains correctly scheduled at 02:00.
2. Changed Record Step
Consider the same simulation, resumed at 01:30 with a new 10-minute
record_step:The first resumed report contains rainfall accumulated from 01:00 to 01:40,
which is 40 minutes of constant rainfall.
If the reporting interval is reset to the resume time, it is interpreted as only the 10 minutes from 01:30 to 01:40.
The reported
mean_rainfallis then 40 mm/h instead of 10 mm/h.Changing the cadence may legitimately anchor future output timestamps to the resume time,
but the interval used for accumulated outputs must match the data that has actually been accumulated.
The same inconsistency affects the reported average timestep and any other output derived from a reporting-interval total.