Statistical Downscaling Modeling for Temperature Prediction

Ashraf, Zeeshan; Kanwal, Bushra; Hussain, Ijaz; Dashtipour, Kia; Gogate, Mandar; Kanwal, Summrina

Abstract

The application compares the Statistical Downscaling Model (SDSM) and partial least square (PLS) to bridge the gap between (minimum and maximum) daily temperatures of 11 sites in Punjab between 1961 and 2013 with atmospheric variables. The data set was utilized for the first time using the proposed framework, which uses PLS and SDSM in conjunction with several regression models to predict future conditions up to the year 2099 under various scenarios. HadCM3 (Hadley Centre Coupled Model 3) data for 26 variables are applied for calibration and validation. After calibration, a Q-Q plot of observed and modeled data was used to validate the model. HadCM3 daily data for A2 and B2 stories were used to generate future scenarios for the years 2014 to 2099. We generated the prediction after using explained variance and partial correlation to select predictors. Using partial least squares (PLS), we select predictive factors and construct future scenarios through 2099. Finally, we conduct a comparative analysis of models developed utilizing the SDSM and PLS approaches for selecting features. The root mean square error was used to pick meaningful anticipated results from many models. After the data is downscaled, it is evaluated and a substantial correlation with the observed data is discovered. After applying R-square and root mean square error (RMSE), we conclude that the PLS (partial least square) variable selection method is preferable to the SDSM method.

• Online Publication Date: Feb 21, 2024
• Publication Date: 2024
• Deposit Date: May 21, 2024
• Publisher: Springer
• Peer Reviewed: Peer Reviewed
• Pages: 147-169
• Series Title: Advances in Information Security
• Book Title: Decision Making and Security Risk Management for IoT Environments
• ISBN: 978-3-031-47589-4
• DOI: https://doi.org/10.1007/978-3-031-47590-0_8