Joint work with Jean-Michel Poggi and Bruno Portier
Forecast a curve sampled within the day (seconds, minutes, hours...), using past measured curves + past exogenous informations, which could be some aggregated measure on the past curves, the weather... Main starting point: computeForecast().
NOTE: algorithms are not specific to PM10 and could be applied to anything else (in similar contexts). However, seasons are hard-coded to follow pollution events. You may want to change them in pkg/R/utils.R, function .isSameSeason()
NOTE 2: the package works for only one series (one station). Further extensions might consider the multi-series case.
The final report may be found at this location