A Climate for Change

Human Development Report - Croatia 2008^255

Basic Information about Climate Models

Annex 3

Numerous factors can influence the Earth’s climate.
When discussing climate change and human causes,
the most important factor presenting a new variable
is the increase of greenhouse gases (GHGs) – most im-
portantly CO 2. Through the use of advanced comput-
ing technology, a number of “Climate Models” have
been developed. These models attempt to imitate the
Earth’s climate with changes in greenhouse gas levels
along with many other variables such as sea ice, the
carbon cycle, evaporation rates, etc. Climate models
attempt to imitate the development of natural con-
ditions using a number of climate related variables
and to simulate the possible development of those
variables in the future – such as rain, temperature,
cloud cover, etc. In general, climate models have been
categorised into Regional Climate Models (RCMs) –
which cover a smaller area in more detail – and Global
Climate Models (GCM) – also called General Climate
Models – which cover the entire globe in less detail.

While Global Climate Models give a good represen-
tation of the entire planet, they are not as helpful for
looking in detail at smaller areas – such as the differ-
ent regions of Croatia that have very different land-
scapes and climates. For example, the latest Hadley
CentreI model, HadGEM1,II uses 135 km grid boxes.
This means that the climate model does not estimate
variations within an area 135 km by 135 km. For Croa-
tia, this could mean that the climate in Karlovac could
be estimated as the same as the climate in Rijeka. That
is why Regional Climate Models (RCM) have been
developed with more detailed resolution, usually 50
x 50 km or higher. The RCM models are obtained by
downscaling GCMs.^1 Downscaling methodologies use
two broad and markedly different approaches to re-
solve climate parameters at substantially finer (higher)

resolutions than global-scale GCMs provide. The first

category is dynamic downscaling, sometimes called

mesoscale simulation, which uses a high-resolution grid

(e.g. a 10 km by 10 km grid) and is performed with-

in a GCM but over a chosen local area. A substantial

advantage of dynamic downscaling is that the wide

range of parameters, available within a GCM (e.g.

temperature, precipitation, soil moisture, wind direc-

tion and strength, etc.), are also available within the

finer-scale grid. However, dynamic downscaling re-

quires supercomputer systems to run the simulations.

As supercomputer capabilities increase in resolution

capabilities and availability, dynamic downscaling will

become more available. The second category, called

statistical or empirical downscaling, has become more

fully developed and more widely used. Statistical

downscaling relies on the availability of a multi-de-

cade data set (e.g. 25-30 years) of past climate change

parameters (e.g. weather station data from a number

of stations across the region), and the GCM data sets

for the same parameters for the same past time period.

To project climatic conditions into the future, the GCM

data for the desired future time period is combined

with an existing statistical relationship for each of the

weather station locations for that region. Statistically

downscaled RCMs usually require less computational

power and can be run on Personal Computers.^2

I The Met Office Hadley Centre is the UK’s official centre for climate

change research and one of the world’s leading centres for climate

change research.

II The HadGEM1 model is the Met Office Hadley centre global envi-

ronment model. This version of the model includes a detailed repre-

sentation of the atmosphere, land surface, ocean, and cryosphere.

Basic Information about Climate Models