Authors: Abel Andres Ramirez Molina, Glenn Tootle, Jiaqi Gong – University of Alabama
Title: Exploring Low-Code AI Techniques for Streamflow Reconstruction in the Po River Basin, Italy
Abstract: This study examines the feasibility of using low-code artificial intelligence (AI) techniques to reconstruct streamflow in the Po River Basin, Italy, over a period of 2000 years using historical Palmer Drought Severity Index (PDSI) data. Traditional methods for streamflow reconstruction are laborious and inefficient, involving the use of regression equations and bias correction techniques. In this study, we explore the effectiveness of nine low-code AI techniques, including Linear Regression (LR), Support Vector Machine (SVM), Deep Learning (DL), Generalized Linear Model (GLM), K-Nearest Neighbors (KNN), Gradient-boosted-trees (GBT), Decision Trees (DT), Random Forests (RF), and Gaussian Process Regression (GPR), to reconstruct streamflow using data from 185 PDSI cells over 92 years. Our results demonstrate that the GLM and RF methods have superior performance. Incorporating a 10-year timing window to extract sequential information from the dataset in the time-based regression task improves the models’ prediction accuracy. The application of low-code AI techniques resulted in improved streamflow reconstruction compared to the conventional approach.