Abstract:
The meteorological station in Jaffna 80 9202 238 ′𝐸, 9 9241 238 ′𝑁 which
is the main meteorology station to maintain weather records for the northern
peninsula experienced difficulties in collecting data during the period from
1984 to 2000 due to hostilities in the region. Although the weather
observations were resumed in 2001, no estimates of missing observations
have been reported. This paper presents a neural network approach of
reconstructing a serially complete data set of monthly temperature records at
the Jaffna meteorology station based on the data available at 4 neighbouring
stations. The standard departures of monthly temperature values calculated
from stations in Mannar, Anuradhapura, Puttalam and Trincomalee were
used as the input to the neural network model to estimate the standard
departure of monthly temperature at Jaffna which was converted back to
monthly temperature values by using the long-term mean monthly
temperature and standard deviation in Jaffna. The neural network was
trained using the data from 1931 to 1960 (30 years) and the output of the
model was tested using data from 1961 to 1980 (20 years). The accuracy of
reconstruction obtained through the neural network model based on the
standard deviation between the difference in actual and estimated values
was ±0.310C. The neural network was applied to reconstruct the missing
data in Jaffna during the period 1981 to 2000 where large gaps in weather
observations are reported.
