Low (web) Quality - BALTEX
Low (web) Quality - BALTEX
Low (web) Quality - BALTEX
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
43<br />
Settled and remaining issues in regional climate modelling with limited-area<br />
nested models<br />
René Laprise 1,4,5 , Ramón de Elía 2,4,5 , Daniel Caya 2,4,5 , Sébastien Biner 2 , Philippe Lucas-Picher 3 , Emilia<br />
Diaconescu 1,4,5 , Martin Leduc 1,4,5 , Adelina Alexandru 1,4,5 , Leo Separovic 1,4,5 and Alejandro di Luca 1,4,5<br />
1 Université du Québec à Montréal (UQÀM), Montréal (Québec), Canada, laprise.rene@uqam.ca; 2 Consortium Ouranos,<br />
Montréal (Québec), Canada; 3 Danish Meteorological Institute (DMI), Copenhagen, Denmark; 4 Canadian Network for<br />
Regional Climate Modelling and Diagnostics (CRCMD), Montréal (Québec), Canada; 5 Centre pour l’Étude et la Simulation<br />
du Climat à l’Échelle Régionale (ESCER), Montréal (Québec), Canada<br />
1. Introduction<br />
A pragmatic approach to reduce the computing cost of highresolution<br />
global climate models is to apply the high<br />
resolution over only a subset of the globe, a technique<br />
known as nested Regional Climate Modelling. <strong>Low</strong>er<br />
resolution data, either simulated by Coupled Global Climate<br />
Models (CGCM) or from reanalyses of observations, are<br />
interpolated in time and in space on the high-resolution grid<br />
of a limited-area Regional Climate Model (RCM) and serve<br />
to define the lateral (and often the ocean surface) boundary<br />
conditions (LBC). Like all models, RCMs are subject to a<br />
variety of sources of errors and uncertainty, as well as to<br />
scepticism about their potential usefulness. But in addition,<br />
there are specific issues that arise with RCMs relating to<br />
their geographically limited computational domain, the<br />
nesting technique, the resolution jump between the RCM<br />
and the driving data, the update frequency of the LBC, and<br />
the imperfections in LBC data.<br />
This presentation will summarise the work presented in<br />
Laprise (2008), Laprise et al. (2008) and references therein.<br />
2. Dynamical Downscaling Hypothesis<br />
The ansatz behind the “dynamical downscaling” technique<br />
is that an RCM, driven by large-scale fields at its LBC,<br />
generates fine scales that are dynamically consistent with,<br />
and somehow preconditioned by the LBC. Initialised and<br />
driven by data without small-scale information, nested<br />
models develop small-scale variance even in the absence of<br />
strong surface forcing. Hence RCM are expected to act as a<br />
kind of magnifying glass that reveals details that cannot be<br />
resolved on a coarse mesh. The small scales represent the<br />
main potential added value of high-resolution RCM. A more<br />
controversial issue concerns the potential improvement of<br />
the large scales in the case of driving by low-resolution<br />
CGCM data.<br />
The spin-up of fine scales proceeds fairly rapidly. Within a<br />
few days in the simulation, atmospheric variance spectra and<br />
forecast error spectra become independent of the time since<br />
the simulation’s initiation and of the presence or absence of<br />
small scales in the initial and LBC. The hypothesised<br />
development mechanisms include fine-scale surface forcing,<br />
hydrodynamic instabilities, mesoscale processes, and<br />
nonlinear interactions cascading information from the large<br />
to the small scales.<br />
3. Deterministic Forecast Vs Climate Skill<br />
In idealised forecast experiments for mid-latitude domains<br />
with 100 by 100 grid points on a 45-km mesh, scales larger<br />
than about 800 km appear to retain extended deterministic<br />
predictability, while scales smaller than about 400 km loose<br />
predictability in a fashion similar to global models. For the<br />
small scales, the shorter the length scale, the shorter the<br />
predictable time. Hence even when fine scales are present<br />
initially and in the LBC, they do not retain deterministic<br />
temporal coherence (at the right place at the right time)<br />
beyond a day or so. This implies that part of the<br />
downscaling is not deterministic, i.e. not entirely<br />
determined or preconditioned by LBC. A nondeterministic,<br />
“free” component exists (free in the sense of<br />
its stochastic-like character) which is also reflected in the<br />
presence of some level of internal variability (IV) in<br />
nested models.<br />
Hence small scales are generated by nested models, but<br />
not with deterministic skill. Climate statistics of small<br />
scales appear to be skilful though, lending some<br />
confidence in the potential usefulness of RCMs in climate<br />
simulations and climate-change projections. IV puts<br />
however severe limitations on the usefulness of singleseason,<br />
single-simulation experiments. While IV may not<br />
be a major problem for climate applications as far as loworder<br />
statistics are concerned (when the domain size is not<br />
too big; see next section), as long as the user of an RCM is<br />
aware of its presence in the interpretation of results.<br />
4. Regional Domain Size<br />
In idealised experiments, it is seen that fine scales develop<br />
in RCM within a few days, and they have the right<br />
amplitude and the right statistics. It has been shown<br />
however that the full spin-up of small scales within the<br />
regional domain requires rather large domains, particularly<br />
in the upper troposphere in mid-latitudes where the flow is<br />
strong. Due to the continuous transport of low-resolution<br />
information from the lateral boundary, some distance is<br />
required for the spin-up process to proceed. The spin-up<br />
distance on the inflow side of domain is larger for stronger<br />
ventilation flow through the domain. For mid-latitude<br />
domains, this distance is thus larger in winter and at upper<br />
levels due to stronger winds. The physically meaningful<br />
portion of the limited-area domain must exclude the spinup<br />
distance, in addition to any sponge or buffer zone used<br />
as part of the lateral boundary nesting.<br />
The downscaling skill has been documented through a set<br />
of idealised “perfect prog” tests. Over small domains (e.g.<br />
70 linear grid points), the small-scale transient eddies are<br />
amplitude deficient, especially at upper levels. With larger<br />
domain (e.g. of the order of 200 linear grid points), smallscale<br />
transient eddies are simulated with the correct<br />
amplitude, but with very little time correlation with the<br />
reference. Lack of time synchronicity is of secondary<br />
importance for climate applications, but must be taken into<br />
consideration for process studies. Failure to reproduce the<br />
correct monthly or daily anomalies may however have<br />
important consequences for the downscaling skill in<br />
seasonal prediction and even for some climate<br />
applications.<br />
By their nature, nested RCM require externally provided<br />
data to drive them. Over mid-latitude domain, the<br />
application of LBC is usually sufficient to control the<br />
large-scale circulation through the RCM domain. There