Creating synthetic meteorological data for ecosystem modeling

State of art process models of forest ecosystems typically benefit from testing with realistic meteorological data. While it may be possible to show that a particular model produces acceptable results when forced by daily or even monthly climate records, testing with hourly or half-hourly observations provides a baseline for comparison with detailed site level measurements of carbon and water fluxes, such as those obtained from sites in the Fluxnet-Canada Research Network (FCRN). In particular, these data are needed to explore the sensitivity of vegetation responses to natural climate variability and the effects of extreme events that will be lost in a record based on daily or monthly mean values. For long term simulations (i.e., decades or longer) at remote sites, however, genuine high frequency meteorological data are simply unavailable, and we are compelled to replay short-term data sets repeatedly to generate a long time series.

Here, I present a method for combining high frequency data observed at eddy covariance sites for one or more years with long-term monthly climate data interpolated either from climate station records or from general circulation model (GCM) scenarios. The objective is to produce a high frequency data set that captures both the long-term observed or simulated trends and the typical day to day and interannual variability observed at the study site. Several approaches have been explored, but in general, they convert high frequency (half-hourly) data into anomalies from the observed means, and combine these with the long-term time series of monthly data. Problems can occur when the high frequency data are combined sequentially and repetitively: these will be discussed together with possible remedies.