Search

This paper utilized the Auto Regressive Integrated Moving Average (ARIMA) technique to forecast the near future rainfall using the long-term satellite rainfall data. At present, many of the future rainfall and climate forecast was conducted for a long-term period (>10 years) at the macroscale (∼1000 km and above) level. Those forecasts are inappropriate to meet the demand for the short-term action (1–3 years). In addition, the macroscale information is unsuitable and too general for the local administrative council and water resources stakeholders to make the adaptation plan, mitigation action, and preventive measures at smaller scale (10–50 km). Many developing countries were facing hydrological data conflict (scarcity, sparse, etc.) and utilizing the publicly accessed satellite data could be useful alternatives. Therefore, sound evidence is required to validate the hypothesis where the use of ARIMA and the long-term satellite rainfall data could provide near future rainfall forecast. Prior to that, the southern part of Johor in Malaysia is selected as the experimental site for this study. To perform the task, 204 of monthly gridded of satellite rainfall from 1998 to 2014 is utilized. In implementing the ARIMA, the standard seasonal lag based on local monsoon period is chosen. The auto-regression co-efficient is computed for each individual grid based on the time series of monthly rainfall, as well as the forecasted rainfall. To determine the ideal duration of historical rainfall for the best near-future forecasted rainfall, the rainfall datasets are divided into decadal and semi-decadal (epoch is every 5 years). The results showed that the satellite-based forecasted rainfall based on ARIMA (SAFORA) using half-decadal datasets able to forecast the 2015 rainfall at average temporal correlation of 0.703. Measurement biases range from 5 to 70%, depending on monsoon season. The SAFORA tends to overestimate low rainfall months. The output of this experiment proved that SAFORA able to forecast the monthly-scale rainfall at reliable accuracy and could be useful for hydrological data conflict situation catchment or location in the humid tropics. • The satellite forecasted rainfall using ARIMA was able to forecast the monthly rainfall with correlation of 0.703. • The best datasets to forecast near future of monthly rainfall for the study area is half-decade datasets. • Measurement bias range from 5 to 70% depending on monsoon season. • The satellite forecasted rainfall using ARIMA (SAFORA) tends to overestimate ground rainfall during low rainfall months. [ABSTRACT FROM AUTHOR]

This paper utilized the Auto Regressive Integrated Moving Average (ARIMA) technique to forecast the near future rainfall using the long-term satellite rainfall data. At present, many of the future rainfall and climate forecast was conducted for a long-term period (>10 years) at the macroscale (∼1000 km and above) level. Those forecasts are inappropriate to meet the demand for the short-term action (1–3 years). In addition, the macroscale information is unsuitable and too general for the local administrative council and water resources stakeholders to make the adaptation plan, mitigation action, and preventive measures at smaller scale (10–50 km). Many developing countries were facing hydrological data conflict (scarcity, sparse, etc.) and utilizing the publicly accessed satellite data could be useful alternatives. Therefore, sound evidence is required to validate the hypothesis where the use of ARIMA and the long-term satellite rainfall data could provide near future rainfall forecast. Prior to that, the southern part of Johor in Malaysia is selected as the experimental site for this study. To perform the task, 204 of monthly gridded of satellite rainfall from 1998 to 2014 is utilized. In implementing the ARIMA, the standard seasonal lag based on local monsoon period is chosen. The auto-regression co-efficient is computed for each individual grid based on the time series of monthly rainfall, as well as the forecasted rainfall. To determine the ideal duration of historical rainfall for the best near-future forecasted rainfall, the rainfall datasets are divided into decadal and semi-decadal (epoch is every 5 years). The results showed that the satellite-based forecasted rainfall based on ARIMA (SAFORA) using half-decadal datasets able to forecast the 2015 rainfall at average temporal correlation of 0.703. Measurement biases range from 5 to 70%, depending on monsoon season. The SAFORA tends to overestimate low rainfall months. The output of this experiment proved that SAFORA able to forecast the monthly-scale rainfall at reliable accuracy and could be useful for hydrological data conflict situation catchment or location in the humid tropics. • The satellite forecasted rainfall using ARIMA was able to forecast the monthly rainfall with correlation of 0.703. • The best datasets to forecast near future of monthly rainfall for the study area is half-decade datasets. • Measurement bias range from 5 to 70% depending on monsoon season. • The satellite forecasted rainfall using ARIMA (SAFORA) tends to overestimate ground rainfall during low rainfall months. [ABSTRACT FROM AUTHOR]