Abstract:
This study mainly investigates the changes in future climate using a variable resolution
Conformal Cubic Atmospheric Model (CCAM) developed at CSIRO, Australia. Bias corrected
sea surface temperatures (SSTs) and sea ice concentration from several global climate models
(GCMs), which were performed for the Intergovernmental Panel on Climate Change (IPCC) fifth
assessment report, were used as boundary conditions for the high resolution simulations over Sri
Lanka at 8 km resolution.
The first study present statistical and neural network approaches in estimating a serially
complete data set of daily maximum and minimum temperature records of Jafftia meteorology
station situated at northern part of Sri Lanka. The daily maximum and minimum temperature
records from 1966 to 1980 (15 years) were used to develop the models. By calculating the
standard deviation between the difference in estimated and measured values it is shown that error
in estimating the daily maximum and minimum temperature of both approaches are about
±0.7°C and ±0.9°C respectively. The results shows that both approaches can be applied to
estimate the missing daily maximum and minimum temperature data in Jaffna for the period 1981
to 2000 where large gaps in whether observations are reported.
CCAM simulations at 50 km forced by the sea surface temperature and sea ice
concentration from six global climate models (GCMs) from the Coupled Model Inter-comparison
Project Phase 5 (CMLP5) over South Asia, centered on Sri Lanka were evaluated to select few
simulations to carry out downscaling at 8 km resolution. Comparatively, three CCAM
simulations CNRM-CM5, GFDL- CM3 and ACCES1-0 show good agreement over the Sri
Lankan region. Mean biases, root mean square errors (RMSE) and pattern correlations were
calculated by comparing the observations against simulations to assess the performance of the
model. Results show that for temperature, biases vary between -1.2 and -2.2 °C. The pattern
correlations for temperature vary among the seasons between 0.92 and 0.96. The model tends to
underestimate observed temperature values over the selected domain. For rainfall, the variation of
pattern correlation is relatively low for the southwest monsoon season which is 0.67 with RMSE
about 3.8 mm/day. CCAM simulations show small biases during the northeast monsoon season
with a strong pattern correlation of 0.75 with RMSE of 1.7 mm/day.
Projected changes in mean temperature and rainfall are also presented for years 2030,
2050 and 2070 for RCP 4.5 and RCP 8.5. Ensemble mean of area average temperature is
expected to increase between 0.8 - 1.6 °C for RCP 4.5. Temperature is expected to increase
between 0.9 - 2.5 °C for RCP 8.5. By 2030, the rainfall increase about 40% in the northern part
of the country during first inter-monsoon season and about 20% in the south-eastern part of the
country during northeast monsoon season. The pattern remains the same in 2050 and 2070.
The extreme rainfall indices, such as, annual highest one-day rainfall, annual highest fiveday
consecutive rainfall, annual maximum length of wet spells and annual maximum length of
dry spells were also analyzed. The projected ensemble mean changes simulated by CCAM for
2050 under RCP4.5 and RCP8.5 suggest that the one-day rainfall and five-day rainfall are likely
to increase along the southwestern and southeastern parts of the country and a decrease over the
northwestern region. The changes are more prominent for higher emission scenario. The length of
wet spells is projected to decrease between 0% and -45% for RCP8.5 and -5 to -43% for RCP4.5.
A moderate decrease in wet spells over the northwestern region and a large decrease over the
southeastern part of the country have been simulated. For the northeastern region, the model
suggests that all four extreme indices will decrease in the future.