Answers are grounded in this paper and cite their sources.
This study presents a two-phase simulation process that applies GEP, SVM, and RF methodologies. The process involves estimation and prediction and uses a “v-fold” testing data assignment. The definition of “v” is based on seasonal variations. The analysis uses data from six stations located in two rivers with different flow ranges in Florida and Oregon States, U.S.A. During the estimation phase, the influential inputs were identified using both the MIC analysis and literature review, and introduced to the models through seven different input combinations. The results demonstrated that these methods can be used to accurately estimate DO under different flow rates in different seasons. Selection of the best models was carried out using three error statistics and variance values interpretations. Correlation analysis was then applied for identifying the input parameters of the prediction models. For estimation procedure, RF (with the average Nash-Sutcliffe (NS) coefficient of 0.910) showed its superiority over the GEP and SVM models, while GEP (with average NS coefficient of 0.962) produced the most accurate results for prediction procedure. An additional uncertainty analysis confirmed the given statements. Water temperature was identified as the most influential parameter on DO estimation, followed by the specific conductance and river water discharge.
No takes yet. Share an insight, caveat, or question.
Shiri et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: