GLUE uncertainty analysis of hybrid models for predicting hourly soil temperature and application wavelet coherence analysis for correlation with meteorological variables

Accurate prediction of soil temperature (Ts) is critical for efficient soil, water and field crop management. In this study, hourly Ts variations at 5, 10, 30 cm depth were predicted an arid site (Sirjan) a semi-humid (Sanandaj) in Iran. Existing machine learning models have high performance, but suffer from uncertainty instability prediction. Therefore, GLUE approach was implemented to quantify model uncertainty, while wavelet coherence used assess interactions between meteorological parameters. Standalone (adaptive neuron fuzzy interface system (ANFIS), support vector (SVM), radial basis function neural network (RBFNN), multilayer perceptron (MLP)) hybridized with four optimization algorithms (sunflower (SFO), firefly algorithm (FFA), salp swarm (SSA), particle (PSO)) improve accuracy reduce uncertainty. For both sites, ANFIS-SFO produced the most accurate performance studied depths. At best, hybridization SFO (ANFIS-SFO, MLP-SFO, RBFNN-SFO, SVM-SFO) decreased RMSE by 5.6%, 18%, 18.3%, 18.2% 5 cm, 11.8%, 10.4%, 10.6%, 12.5% 10 9.1%, 12.1%, 13.9%, 14.2% compared respective standalone models. analysis confirmed superiority hybrid over models, predictions. covered 95%, 94%, 96% observation data soli depths, respectively. Wavelet demonstrated that air temperature, relative humidity, solar radiation, not wind speed, had different depths parameters mostly influenced upper layers. conclusion, necessary powerful technique obtain realistic Ts. contrast, useful tool investigate effective variables strongly affect predictions.Graphic abstract