302
P. Verma et al.
On the other hand, Ghum is a small hill station near Darjeeling Himalayan
hill region of West Bengal, India. Its latitude and longitude are 27.0008° N and
88.2437° E, respectively, it lies on 2025 m above sea level, and the climate is warm
and temperate. As comparing with winters (November to February), summers (March
to October) have much more rainfall with an average temperature 13.9 °C. As it is
also the hilly region and due to orographic nature of rainfall, it is very important to
study the rainfall feature of this area to prevent the habitat from the adverse effect of
landslide.
2.2 Data Source
Giovani is the Web data source for providing user the National Aeronautics and Space
Administration (NASA) geophysical gridded data from various satellites and surface
observations to analyze or to derive important scientific conclusions. It facilitates
researchers with the capability to analyze or examine data on atmospheric parameters.
Historical satellite-driven monthly rainfall data of thirty-nine years (1980–2018)
having spatial resolution of 1° × 1° for two northeast (NE) Indian hill stations
having very slight climatic variation is considered in this study.
3 Methodology
The historical data of thirty-nine years of the two sites viz Majitar (altitude 200 m)
and Ghum (altitude 2258 m) are pre-processed involving cleaning part to avoid any
spurious value or to fill any missing value. After this step, the data is smoothened to
avoid any sudden spike in the data. The data after pre-processing and smoothening is
taken as the input of the for the ARIMA model. ARIMA model is preferred here for
prediction as it has a fixed structure to suit for the prediction of time series sequential
data. The ARIMA model is used for the forecasting monthly precipitation of the
two stations Majitar and Ghum with a step of ten step ahead. The model coefficients
specifically the model order and model delay are changed to various values, and
at each step, the prediction error is calculated. In that way, the model is tuned for
lowest prediction error. Then, the error is calculated in terms of MSE, RMSE, SI, R
2
value, and MAE between the dataset used as the input for ten step ahead prediction
of rainfall and predicted series of rainfall by the tuned model.
Following equations are used for analysis:
The ARIMA model equations are
A(z)y(t) = C(z)e(t)
P. Verma et al.
On the other hand, Ghum is a small hill station near Darjeeling Himalayan
hill region of West Bengal, India. Its latitude and longitude are 27.0008° N and
88.2437° E, respectively, it lies on 2025 m above sea level, and the climate is warm
and temperate. As comparing with winters (November to February), summers (March
to October) have much more rainfall with an average temperature 13.9 °C. As it is
also the hilly region and due to orographic nature of rainfall, it is very important to
study the rainfall feature of this area to prevent the habitat from the adverse effect of
landslide.
2.2 Data Source
Giovani is the Web data source for providing user the National Aeronautics and Space
Administration (NASA) geophysical gridded data from various satellites and surface
observations to analyze or to derive important scientific conclusions. It facilitates
researchers with the capability to analyze or examine data on atmospheric parameters.
Historical satellite-driven monthly rainfall data of thirty-nine years (1980–2018)
having spatial resolution of 1° × 1° for two northeast (NE) Indian hill stations
having very slight climatic variation is considered in this study.
3 Methodology
The historical data of thirty-nine years of the two sites viz Majitar (altitude 200 m)
and Ghum (altitude 2258 m) are pre-processed involving cleaning part to avoid any
spurious value or to fill any missing value. After this step, the data is smoothened to
avoid any sudden spike in the data. The data after pre-processing and smoothening is
taken as the input of the for the ARIMA model. ARIMA model is preferred here for
prediction as it has a fixed structure to suit for the prediction of time series sequential
data. The ARIMA model is used for the forecasting monthly precipitation of the
two stations Majitar and Ghum with a step of ten step ahead. The model coefficients
specifically the model order and model delay are changed to various values, and
at each step, the prediction error is calculated. In that way, the model is tuned for
lowest prediction error. Then, the error is calculated in terms of MSE, RMSE, SI, R
2
value, and MAE between the dataset used as the input for ten step ahead prediction
of rainfall and predicted series of rainfall by the tuned model.
Following equations are used for analysis:
The ARIMA model equations are
A(z)y(t) = C(z)e(t)