|
Scientific Research Challenges
GENIEfy will develop the technology to tackle three key scientific priorities:
- The need for a traceable spectrum of models. It is essential to the scientific process that we develop a
full range of understanding, with modelling capability ranging from conceptual to predictive. Currently
there are sharp discontinuities between the modelling tools used for basic understanding, those used for
long-term and paleo climate studies, and those used for complex predictive studies. We need to engineer
a much smoother, and so far as possible, traceable, transition between models, thus enabling us to
progress rapidly from a qualitative recognition of the potential importance of a process through to
quantitative assessment and the implementation of parameterisation. Ultimately we need a rigorous
methodology to determine the spatial resolution and process complexity that actually need to be included
in an ESM to tackle any particular scientific problem. Progress in this direction can be made by designing
a model spectrum using a modular and traceable set of components and by using extensive model
inter-comparison to determine the relative importance of various model attributes. We have made progress
in this area, but we now propose to extend our framework to be able to couple higher resolution and full
complexity components, and to assimilate output from full complexity ESMs in faster ESMs. This will enable
users to examine the variability of integral properties of the Earth system, including the oceanic
thermohaline circulation (THC), climate sensitivity, and carbon cycle, across a range of model resolution
and complexity, for present day, future and paleo climates.
- Improving models through data assimilation and uncertainty analysis. Adequate testing of model components
demands an ability to undertake repeated and/or long integrations, and hence a model spectrum which includes
fast models with moderate spatial resolution. Data assimilation techniques (e.g. the EnKF) and Bayesian
approaches now exist that provide a generic methodology for objectively fitting and comparing models to data.
In addition, the quantification of parameter uncertainty allows attribution of remaining model-data mismatch
to errors in the model structure (and perhaps systematic errors in the data). In addition to the assimilation
of data from the recent past, we must also look to the use of paleo data to constrain Earth system models,
an area in which we have recently been successful. It is no longer sufficient to produce a single prediction
of climate change, even if it is with the highest resolution, most complex model available. Any prediction of
the future (or hindcast of the past) should include an estimate of its uncertainty. Moreover, this estimate
should include a quantification of as many sources of error as possible. Lack of resolution is one source of
error, but there are many others, some of which may be considerably larger. Uncertainty analysis must
ultimately be done using the most complete ESM possible and should be performed in an intelligent, economical
and easily reproducible fashion. The data assimilation techniques we have applied in GENIE phase 1, together
with theoretical techniques developed in Tyndall phase 1, allow this to be done in an efficient and rigorous
way.
- Coupling global change and human activities. Future scenarios that treat human activities as a black box and
include only a one-way interaction from prescribed emissions and land-use change to their consequences are
incomplete. They miss a host of potentially important feedbacks on activities within the 'anthroposphere',
which need to be elucidated and quantified. Better integrated future scenarios are also needed to inform the
policy process. Current integrated assessment models tend to use extremely simplified representations of the
natural Earth system. They lack an adequate representation of climate variability and extreme events, and
have not yet fully grappled with feedback between the human and natural systems. A fast ESM has recently
been developed in GENIE phase 1, in collaboration with the Tyndall Centre, for use in their Community
Integrated Assessment System (CIAS), and it is also being used for integrated assessment by our collaborator
Neil Edwards in Bern. However, a tighter integration of the GENIE and CIAS modelling frameworks is required
to properly represent ecosphere-anthroposphere feedbacks, and the interfacing needs to be developed to
couple models of human land and ocean resource use processes and terrestrial and marine ecosystem services.
|
Overview