In fact, early biogeochemical models relied on nudging (then also referred to as restoring) of model nutrients to climatological nutrient distributions in order to infer net community production and other biogeochemical processes (Najjar et al., 1992 and Marchal et al., 1998). In PARP inhibition the more recent, mechanistically detailed biogeochemical

models nudging is frequently used for the reduction of biases resulting from imperfect boundary conditions; for instance, in nested 3D applications variables are nudged to physical and biogeochemical distributions (either from lower-resolution, larger-scale models or climatological observations) in buffer zones along their open boundaries (e.g. Fennel et al., 2008). Nudging is also used in 1D models

to drive variables toward either direct observations (e.g. Bagniewski et al., 2011) or climatologies (e.g. Fennel et al., 2003) in order to account for unresolved 3D processes. Advantages of conventional nudging are that it is easy to implement, robust and can force the model arbitrarily close to the observations. Unfortunately, there are serious limitations as well if the technique is used to nudge a model towards a climatology: high-frequency variability (e.g., eddies in ocean circulation models) are suppressed and artificial phase lags are introduced, especially when nudging is strong (Woodgate and Killworth, 1997 and Thompson et al., 2006). As a solution to this problem, selleck kinase inhibitor Thompson et al. (2006) proposed limiting the nudging to prescribed frequency bands, leaving the model to evolve freely outside of these bands. We will refer to this modified method as frequency dependent nudging. (In the original paper by Thompson et al. (2006) the nudges were filtered in both space and time and, for this reason, the

original method was called spectral nudging. In the present application the nudges are only filtered in time (i.e., in the frequency domain) and so we will refer to the method as frequency dependent nudging.) In ocean and atmosphere models the chosen frequency bands are often Metformin purchase centered on the mean and annual cycle, which tend to be well characterized in climatologies. Frequency dependent nudging has been demonstrated to be effective in reducing bias errors in eddy resolving ocean circulation models (e.g. Thompson et al., 2006, Thompson et al., 2007, Stacey et al., 2006 and Zhu et al., 2010). Here we perform an exploratory study to assess the utility of frequency dependent nudging in reducing seasonal biases in biogeochemical ocean models without suppressing higher frequency variations (e.g., blooms with typical scales of a week). To our knowledge, frequency dependent nudging has not yet been applied to such models. We use a framework where a simulation from a complete model is sampled and these samples are treated as observations. Although these “observations” are synthetic we henceforth refer to them simply as observations. A climatology, defined to consist only of the mean and annual cycle (i.e.