66
S. Kiran and D. R. Micheal
effects. Sikkim is divided into four quadrants based on the amount of rainfall it
receives. Rainfall is more in the southeast and southwest quadrants when compared
to the northeast and northwest quadrants. The narrow zone between the southeast and
southwest quadrants also experiences lesser rainfall [1]. The high rainfall in Sikkim
is controlled by the north–south oscillations of monsoon trough [2].
Variation in rainfall pattern, influenced by climate change, has cascading impacts
on various sectors. The impact due to the variations in seasonal rainfall was elucidated
through community observations in Sikkim. It also had a considerable influence on
drying up of water sources and drastic reduction in the lean period discharge of springs
bringing about crop yield instability [3]. The excess runoffs from high intensity
rainfall have had a destabilizing impact on the sensitive topography leading to soil
erosion and landslides especially in North Sikkim and inflicting damages on irrigation
and urban infrastructure, shunting even their optimal utilization [4]. Hydropower
generation in enhanced stream flows is susceptible to rockfalls, debris, and river
bank erosion [5], knowing the distribution of rainfall assists in the interpretation
of rainfall pattern and all associated hydrological parameters, particularly runoff.
Identifying the best-fit probability distribution of seasonal and annual rainfall for
stations in Sikkim will provide vital information required in planning, designing,
managing water resource systems and mitigating the impact of extreme rainfall in
hydrologically vulnerable regions.
Quantitative analyses involving rainfall or precipitation have predominantly
focused on extreme event forecasting, frequency analysis, and trend analysis using
time series. Frequency analysis of rainfall without adequate information about the
probability distribution function that best fits the data leads to erroneous interpretation
of recurrence interval which is an important input parameter in various hydrological designs. For example, rainfall data rarely follows normal distribution. However,
it is a common practice to apply normal statistics rules for rainfall data analysis
resulting in an inaccurate prediction. Therefore, precise knowledge about the probability distribution of rainfall at a station is important and crucial for rainfall data
analysis, interpretation, anomaly identification, and events’ prediction. Fitting probability distributions to sample hydrological data involves estimation of associated
parameters representative of the population. This is accomplished using various
methods of which the method of L-moments and the method of maximum likelihood estimation are widely used. The maximum likelihood method is consistent
and theoretically generates the most efficient estimates of parameter in a probability
distribution [6, 7]. L-moments are more robust, unbiased, and are less sensitive to
the sample size and outliers [8, 9]. In this study, an effort is made to determine the
best-fit probability distribution(s) of annual and seasonal rainfall from the monthly
rainfall data recorded across four observation stations in Sikkim, India.
S. Kiran and D. R. Micheal
effects. Sikkim is divided into four quadrants based on the amount of rainfall it
receives. Rainfall is more in the southeast and southwest quadrants when compared
to the northeast and northwest quadrants. The narrow zone between the southeast and
southwest quadrants also experiences lesser rainfall [1]. The high rainfall in Sikkim
is controlled by the north–south oscillations of monsoon trough [2].
Variation in rainfall pattern, influenced by climate change, has cascading impacts
on various sectors. The impact due to the variations in seasonal rainfall was elucidated
through community observations in Sikkim. It also had a considerable influence on
drying up of water sources and drastic reduction in the lean period discharge of springs
bringing about crop yield instability [3]. The excess runoffs from high intensity
rainfall have had a destabilizing impact on the sensitive topography leading to soil
erosion and landslides especially in North Sikkim and inflicting damages on irrigation
and urban infrastructure, shunting even their optimal utilization [4]. Hydropower
generation in enhanced stream flows is susceptible to rockfalls, debris, and river
bank erosion [5], knowing the distribution of rainfall assists in the interpretation
of rainfall pattern and all associated hydrological parameters, particularly runoff.
Identifying the best-fit probability distribution of seasonal and annual rainfall for
stations in Sikkim will provide vital information required in planning, designing,
managing water resource systems and mitigating the impact of extreme rainfall in
hydrologically vulnerable regions.
Quantitative analyses involving rainfall or precipitation have predominantly
focused on extreme event forecasting, frequency analysis, and trend analysis using
time series. Frequency analysis of rainfall without adequate information about the
probability distribution function that best fits the data leads to erroneous interpretation
of recurrence interval which is an important input parameter in various hydrological designs. For example, rainfall data rarely follows normal distribution. However,
it is a common practice to apply normal statistics rules for rainfall data analysis
resulting in an inaccurate prediction. Therefore, precise knowledge about the probability distribution of rainfall at a station is important and crucial for rainfall data
analysis, interpretation, anomaly identification, and events’ prediction. Fitting probability distributions to sample hydrological data involves estimation of associated
parameters representative of the population. This is accomplished using various
methods of which the method of L-moments and the method of maximum likelihood estimation are widely used. The maximum likelihood method is consistent
and theoretically generates the most efficient estimates of parameter in a probability
distribution [6, 7]. L-moments are more robust, unbiased, and are less sensitive to
the sample size and outliers [8, 9]. In this study, an effort is made to determine the
best-fit probability distribution(s) of annual and seasonal rainfall from the monthly
rainfall data recorded across four observation stations in Sikkim, India.