Abstract:This study uses 14 Coupled Model Intercomparison Project Phase 6 (CMIP6)‐GCMs to develop accurate daily precipitation maps for historical and future periods. This study introduces triple coupling machine learning approach (including a pixel‐wise classification and two separate regression algorithms) to correct GCM biases and generate an ensemble of outputs. Comparisons show this triple coupling method outperforms conventional downscaling, revealing potential future impacts like decreased average precipitation, longer dry spells, and fewer wet days in the study area. Climate change is expected to change the frequency, duration, intensity, and pattern of precipitation, underscoring the need for accurate predictive tools. Earth system models (ESMs) serve as invaluable instruments in this endeavour, simulating climate variable variations across temporal and spatial dimensions. This study aims to develop a methodology for generating precise daily precipitation maps by rectifying biases inherent in ESM outputs. The proposed methodology includes downscaling ESM outputs to simulate historical daily grid‐based precipitation, thereby enhancing the fidelity of daily precipitation representation. For this purpose, 14 models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) were employed. Random forest (RF) machine learning method was used to correct biases in these ESM outputs. This study's novelty lies in integrating results of a grid‐based RF classification model, employed to distinguish between rainy and non‐rainy days, with those obtained by two RF regression models, to estimate precipitation amounts for grid cells receiving extreme and non‐extreme precipitation, to generate an ensemble of ESM outputs. The resulting method, termed the triple coupling method (EN‐RF), was validated using precipitation data from the Divandareh‐Bijar Basin in western Iran to simulate historical climate conditions. Furthermore, the accuracy of the developed triple coupling approach was compared with that of a commonly used single machine learning‐based downscaling model (EN‐Single‐RF). Comparative analysis against a commonly used single machine learning‐based downscaling model (EN‐Single‐RF) revealed the superior performance of the EN‐RF approach in replicating the intensity and distribution of daily precipitation. Furthermore, within the triple coupling framework, support vector machine (SVM) was utilized to simulate daily historical precipitation (EN‐SVM), while the quantile mapping (QM) method served as a benchmark. Comparison of the results showed superiority of the EN‐RF to other methods (EN‐Single‐RF, EN‐SVM, and QM) in terms of various accuracy metrics (Kling‐Gupta Efficiency = 0.95, mean square error = 0.22). The findings indicated the capability of the proposed triple coupling framework using the RF algorithm to simulate the spatio‐temporal distribution of precipitation using the ESM precipitation outputs. The developed framework can be used to produce reliable projections to gain deeper insights into the potential impacts of climate change on regional precipitation patterns.

Predicting rainfall response to climate change and uncertainty analysis: introducing a novel downscaling CMIP6 models technique based on the stacking ensemble machine learning

Improving the Spatial Prediction of Soil Organic Carbon Content in Two Contrasting Climatic Regions by Stacking Machine Learning Models and Rescanning Covariate Space

Forecasting actual evapotranspiration without climate data based on stacked integration of DNN and meta-heuristic models across China from 1958 to 2021

Statistical downscaling of precipitation in northwestern Iran using a hybrid model of discrete wavelet transform, artificial neural networks, and quantile mapping

A stacked ensemble learning and non‐homogeneous hidden Markov model for daily precipitation downscaling and projection

Evaluating the downscaling uncertainty of hydrometeorological data in snowmelt runoff simulation

Statistical downscaling of global climate model outputs to monthly precipitation via extreme learning machine: A case study

A New Methodology for Reference Evapotranspiration Prediction and Uncertainty Analysis under Climate Change Conditions Based on Machine Learning, Multi Criteria Decision Making and Monte Carlo Methods

Uncertainty Analysis of Climate Change Impacts on Flood Frequency by Using Hybrid Machine Learning Methods

Enhancing groundwater quality prediction through ensemble machine learning techniques

Machine learning-based downscaling: application of multi-gene genetic programming for downscaling daily temperature at Dogonbadan, Iran, under CMIP6 scenarios

Enhancing spatial resolution of satellite soil moisture data through stacking ensemble learning techniques

Modeling Hydrological Responses of Watershed Under Climate Change Scenarios Using Machine Learning Techniques

Downscaling of Precipitation for Climate Change Projections Using Multiple Machine Learning Techniques: Case Study of Shenzhen City, China

Downscaling long lead time daily rainfall ensemble forecasts through deep learning

Advancing Flood Susceptibility Modeling Using Stacking Ensemble Machine Learning: A Multi-Model Approach

Triple coupling random forest approach for bias correction of ensemble precipitation data derived from Earth system models for Divandareh‐Bijar Basin (Western Iran)

Runoff Simulation Under Future Climate Change Conditions: Performance Comparison of Data-Mining Algorithms and Conceptual Models

Near future flash flood prediction in an arid region under climate change

Application of ensemble machine learning model in downscaling and projecting climate variables over different climate regions in Iran

Ensemble of optimised machine learning algorithms for predicting surface soil moisture content at a global scale