Processing bug
There is a data processing bug which is a bit difficult to describe and was difficult to reproduce, but I managed to provide an example.
Situation: see attached example, modified from the Stillberg example. In the example, 1 station (*DAV) provides PSUM for the full period, 1 station (*WFJ) only for the first 6 hours.
Using run.sh, running from 2014-10-01 to 2014-10-10, all the time steps show: [i] Interpolating PSUM (PSUM_SNOW, 2 stations, AVG_LAPSE, 2 stations, (null average input for frac lapse rate))
Even though only 1 station is available after 2014-10-01T06:00:00, 2 stations are taken into account. Closer inspection revealed that the station with no data is considered to have PSUM = 0, such that the AVG_LAPSE only has halve the precipitation it should have. However, there is no data generator specified for PSUM.
In run2.sh, we start the simulation later in time (2014-10-05T01:00:00), and then all the time steps show: [i] Interpolating PSUM (PSUM_SNOW, 1 station, AVG_LAPSE, 1 station, (null average input for frac lapse rate))
This means that depending on starting time, different data is returned by meteoio for future dates (i.e., dates after 2014-10-05).
Also, even with one stations, occasionally it shows: [i] Interpolating PSUM (PSUM_SNOW, 1 station, AVG_LAPSE, 1 station, Computed regression with NoisyLinear model (fixed lapse rate=3.35e-05) - r=1) indicating that the lapse rate provided in the ini file is ignored.
In run3.sh, WFJ data has been modified (WFJmod) to contain nodata (-999), instead of just no lines after 2014-10-01T06:00:00. Then, behavior is also correct: [i] Interpolating PSUM (PSUM_SNOW, 2 stations, AVG_LAPSE, 2 stations, Computed regression with NoisyLinear model (fixed lapse rate=0.00010825) - r=0.99) up to 2014-10-01T06:00:00 and [i] Interpolating PSUM (PSUM_SNOW, 1 station, AVG_LAPSE, 1 station, Computed regression with NoisyLinear model (fixed lapse rate=0.0001665) - r=1) afterwards.
So, there is different behavior whether or not time stamps are present with only -999, or time stamps without data are absent.
Attachment: