Figure 3 : GCM grid system component ( Adapted from Lukas et al ., 2014 )
potential trajectories . With increased realism , expanded scope , and diverse scenarios , CMIP6 equips us with the essential knowledge to navigate the evolving climate responsibly and forge a sustainable path toward a resilient future ( Eyring et al ., 2016 ).
Coupling of Machine Learning and Global Climate Models
ML coupled with GCMs for modelling future irrigation water availability from river basins and rivers for rice granaries has made the process much easier , relying solely on climate data . The only key data needed to successfully model future rice irrigation water availability using ML are precipitation , temperature , relative humidity , and wind speed for observed and GCMs . Figure 4 shows the framework of the ML technique as a streamflow predictive model considering future climate change .
The observed streamflow ( river flow rate ) data is also needed for training and testing purposes under ML . These climate data will be trained and tested under the selected ML technique to study and capture the pattern of observed streamflow . Adib and Harun ( 2023 ) have evaluated the performance of two MLs , named support vector regression ( SVR ) and random forest ( RF ), to predict streamflow patterns of the Kurau River from the Kurau River Basin solely based on climate parameters . The models successfully achieved good performance results , as categorised by Moriasi et al .( 2015 ). The validated ML model can then be used for future river flow projection to model the availability of irrigation supply for rice granaries ( Adib & Harun , 2022 ). At this point , future climate data extracted from GCMs that have been downscaled to the required study area are injected into the validated ML model . The validated ML will then use the GCM data to generate future streamflow values from river basins and rivers to assess the availability of water resources for rice irrigation . The GCM can provide future projections up until 2100 , while certain scenarios can reach 2300 .
ML algorithms can expedite the process of model calibration , parameter optimisation , and uncertainty quantification , reducing the computational burden associated with traditional hydrological modelling techniques and enabling easy access to water resource availability statuses . This efficiency enables researchers and practitioners to perform ensemble simulations , sensitivity analyses , and scenario assessments more quickly and cost-effectively .
Future projections of irrigation availability for rice are crucial for several reasons :
● Food Security : Rice is a staple food for a large portion of the world ’ s population , particularly in Asia . Reliable irrigation is essential for ensuring stable rice production . Future projections help policymakers and farmers anticipate potential water shortages or changes in
29