Notes on spectral nudging urban climate models - Nudging is wrong, do not do it

Debate erupted about spectrally nudging regional climate models during summer 2016 training on regional climate modeling at National Center for Atmospheric Research (NCAR). The debate was whether nudging is valid for regional climate models. Because of the interest of many participants who were PhDs and postdocs from all over the world, special session was assigned for extended discussion. I was new to the nudging concepts at the time, but thought that it is interesting, especially whether nudging is useful for downscaling reanalysis and GCMs incorporating urban canopy models. I dropped my ear and followed the discussion. Based on the ideas raised from the organizers and participants, I also forwarded some questions of relevance for the urban climate modeling.

As an extended discussion during the afternoon, many questions were raised: If a regional climate model is spectrally nudged, doesn't it lose its added value as a regional climate model because nudging forces the model to forget the mesoscale and microscale atmospheric processes? What difference does it make if regional climate model is nudged towards the reanalysis or GCMs? some people taught that it is difficult to obtain the efficiency or uncertainty of the nudged RCMs because the processes are forced to simulate the large scale processes similar to the GCMs? How can the strength of nudging be determined? some participants believed that so far nudging parameters are determined arbitrarily through trial and error, and it is questionable whether trial and error is applicable in a scientific procedure.

There were many questions raised and the debate gone too far. Some argued that nudging is imposing spurious waves to correct the deviations and others argued that there is nothing wrong in correcting the deviations by imposing certain restrictions on the models. In order to achieve an atmospheric state close to the boundary forcing the large spatial scales of the RCMs is “nudged" towards the reanalysis or GCMs  to reduce the internal deviations. The general idea of forcing the regional model is to capture the large scale atmospheric processes just like the reanalysis and GCM. The reanalysis and GCMs are believed to resolve large scale processes and spectral nudging is imposed on RCMs to behave the same. This is not true for regional scale, because mesoscale processes are only resolved in RCMs and the applicability of nudging for small scale weather systems  (e.g., orographics features and processes due to land-sea contrast) is wrong.

Nudging is creating spurious wave numbers to correct internal wave drift of the regional climate model from reanalysis and GCMs. It prevents the interior of the mesoscale model diverging from the large-scale circulation (Claire et al. 2015). This is the main reason why spectral nudging needs to be avoided in urbanized regional climate models. This is because small scale urban features must be resolved to accurately obtain the urban effects.

Spectral nudging suppresses internal model variability (Rocket et al., 2008) and gives better results for large scales but at the expense of a reduced variability at smaller scales (Davies, 1976). Weaver et al. [2002] reached a similar conclusion using a configuration with weaker nudging and higher resolution, supporting the idea that by not nudging synoptic scale observations/reanalysis at the mesoscale, simulations can be improved.

“Spectral nudging is able to nearly cancel out the differences independent of the domain size, i.e., differences for the simulations with spectral nudging are of the same order for the small and large domain.”

Furthermore, it is fairly known how much nudging is required and the wave number adjustment. Because of the above reasons, in urban climate models, applying spectral nudging is not advisable.