Climate Sense

[

] 242

daptation

and

itigation

trategies

Comparison between observed and predicted precipitation

Time series of observed and artificial neural network reconstructed

precipitation for 30 per cent of data randomly degraded (top) and scatter

plot of reconstructed against observed precipitation for 30 per cent

degraded data (bottom)

Source: Caribbean Institute for Meteorology and Hydrology (CIMH)

be appropriate to predict climate change, but a combined approach

where regional models are also statistically downscaled would be

valuable. Additionally, islands that are small, irregular land masses

would benefit greatly from high resolution modelling, as processes

such as land/sea breezes, which are a result of the different thermal

characteristics of land and ocean, cannot be captured by a GCM.

The type of climate information required may also influence

whether or not high resolution modelling is desirable. Certain

surface variables, such as surface temperature or precipitation, are

more likely to be significantly improved by using high-resolution

data as opposed to atmosphere variables such as 500 hectopascal

heights. One may have a case where a low resolution model gives

fairly accurate mean rainfall, however capturing extreme precipita-

tion events usually requires high resolution models.

In recent years, there has been an increase in use of RCMs through-

out the Caribbean. These models use dynamical downscaling and

provide more accurate representations of physical processes. There are

a number of Caribbean institutions – Climate Studies Group Mona

(CSGM), University of the West Indies, Instituto de Meteorologia de

la Republic de Cuba and the Caribbean Community Climate Change

Centre –that, through their collaborative efforts, have produced

regional climate change projections that can be used by researchers

and policy makers to assess potential impacts and develop adaptation

plans. Extensive work is being done to ensure these products will

be readily accessible and available to the wider community in addi-

tion to universities and research centres. CSGM also uses statistical

downscaling to downscale global projections to specific observa-

tional stations. Statistical downscaling of temperature

and precipitation from global models to station sites

in Barbados, Jamaica and Trinidad was undertaken in

an Assessments of Impacts and Adaptations to Climate

Change in Multiple Regions and Sector Small Island

States SIS06 project. Information from these climate

scenarios would be useful to policy makers for charting

adaptation and mitigation strategies for the region.

CIMH has investigated the plausibility of using an

artificial neural network (ANN) to reconstruct missing

daily precipitation data for a rainfall station in Guyana.

This functionality was tested by asking ANN to predict

data at a station using incomplete data as well as data

from nearby stations. Comparisons of the predicted

data to actual measured data at the site showed good

correlation, even though the data given to ANN had

30 per cent of its values randomly removed. Further

ongoing studies are being conducted to assess the effec-

tiveness of the ANN technique for longer time series. If

successful, this would support statistical downscaling

of climate models and climate change impact studies in

the Caribbean region.

Future work at CIMH

Future climate-oriented work at CIMH will focus

on investigating long-term climate variability with

the intention of improving predictions and enhanc-

ing understanding of climate change in the region.

Research will be conducted using mesoscale climate

models such as the weather research and forecasting

model to obtain information about the variability of

climate in the region and to determine the medium-

to long-term variability of weather systems affecting

Caribbean climate.

Additional research will use mesoscale climate models

to examine the effects of Saharan dust on Caribbean

climate. Long-term monitoring studies in the Caribbean

region have shown large interannual increases in atmos-

pheric dust concentrations. Atmospheric dust can

influence the Earth’s radiative budget directly, by scat-

tering solar radiation in the atmosphere, and indirectly,

by changing cloud condensation nuclei concentrations.

Dust has also been linked to reduced precipitation

in the Caribbean and to changes in seasonal rainfall

patterns.

It may also be a major source of tropical

hurricane suppression in the Northern Tropical Atlantic

due to factors such as cooling of sea surface tempera-

tures in the main hurricane development region

and

the Saharan Air Layer.

A key regional initiative that would be of great benefit

to any mitigation and adaptation strategy would be an

objective intercomparability and assessment study,

which would enhance understanding of the range and

causes of uncertainties in climate model outputs and the

model biases for the Caribbean region. It would also be

useful in showing how model predictions may change

according to different scenarios. Information from all

these model simulations and intercomparisons would

be useful for adaptation and mitigation initiatives.