Short-term forecasting of spring freshet peak flow with the Generalized Additive model

Dubos, Véronique; Hani, Ilias; Ouarda, Taha B. M. J.; St‐Hilaire, André

doi:10.1016/j.jhydrol.2022.128089

Cited by 7 publications

(5 citation statements)

References 63 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The absence of surface runoff observed in each sub-basin (SURQ) and water infiltrating into the soil profile (PERC) is due to the same reason. Additional to that proven by Dubos et al (2022), who demonstrated the effectiveness of applying the GAM model to climate data, the significance level obtained in this study for the fit function in the period 2006-2017 (C), mostly p < 0.001, indicates that the line describes very well the monthly mean behavior for the variables ET, GW, WYLD, PET, SW, and SW1. It is evident that their increase separates them from the previous periods A and B (Figure 5).…”

Section: Temporal Analysissupporting

confidence: 73%

“…The Generalized Additive Model (GAM) was used to compare their behavior at the level of monthly mean values. This is a smoothed, non-parametric regression model, which is considered an extension of the Generalized Linear Models (GLM) (Dubos et al, 2022), choosing the Gaussian family for the response variable. The plots and statistical parameters derived from the GAM application were obtained with the help of the ggplot2 and mgcv libraries within the RStudio software (R Core Team, 2022).…”

Section: Hydrological Balancementioning

confidence: 99%

See 1 more Smart Citation

Spatial-Temporal Surface Water Availability in the Tulancingo River Basin, Hidalgo, Mexico

Torres-Suárez,

Bolaños-González,

Ibáñez-Castillo

et al. 2024

agro

View full text Add to dashboard Cite

The Tulancingo River basin is part of the upper Metztitlán River basin and has an aquifer deficiency that covers 87 % of its surface area; however, floods have occurred in the valley region. This disparity encouraged the current work, the goal of which was to estimate the spatio-temporal availability of water in the soil within the basin from the modeling of surface water balance elements. Sensitivity analysis of the model fit parameters was performed with the SWATCUP program and the SUFI2 algorithm. The runoff values were estimated with the SWAT program and were compared with the records of the Venados Hydrometric Station, obtaining values of the statistical indicators R2 = 0.89 and Nash Sutcliffe (NS) = 0.86 for the calibration period (1982–2002); and R2 = 0.77 and NS = 0.62 for the validation period (2003–2013). Of the nine sub-basins that make up the Tulancingo basin, the upper sub-basins of the Río Chico (11), Ventoquipa stream (12), and San Lorenzo River (13) showed values 2.2 times higher than the rainfall depth of the lower sub-basins (5, 6, 7, 8, 9, and 10). Surface runoff and infiltration reported maximum values from July to October and minimum values from April to May. The maximum final water availability in the soil occurred in September. Regarding the analysis by periods, the Generalized Additive Model provided a good fit for most of the variables (0.6 < R2 < 0.94); during the last period, higher values were observed than in the previous periods (p < 0.001), identifying a change in the hydrological state of the basin. This was mainly associated with a significant change in rainfall, which was reflected in the trend of surface runoff.

show abstract

Section: Temporal Analysissupporting

confidence: 73%

Section: Hydrological Balancementioning

confidence: 99%

Spatial-Temporal Surface Water Availability in the Tulancingo River Basin, Hidalgo, Mexico

Torres-Suárez,

Bolaños-González,

Ibáñez-Castillo

et al. 2024

agro

View full text Add to dashboard Cite

show abstract

“…Although these models can effectively indicate the correlation between runoff sediment and its driving factors, they generally cannot effectively eliminate the negative impact of the multicollinearity among variables. Moreover, they cannot accurately explain the nonlinear characteristics between the response and explanatory variables, and it is difficult to have further improvement in model accuracy (Dubos et al, 2022). Accordingly, this study used the generalized additive model to quantitatively determine the complex nonlinear relationship between explanatory and response variables in the Yeji River basin and to quantitatively measure the degree to which the explanatory variables (rainfall, sediment source–sink landscape, and hydrological connectivity) could explain runoff and sediment.…”

Section: Introductionmentioning

confidence: 99%

“…Hydrological processes are complex nonlinear phenomena caused by the combined action of multiple driving factors (Latt & Wittenberg, 2014). Currently, linear and nonlinear models are available for determining factors driving runoff and sediment in watershed areas (Dubos et al, 2022). Common linear models mainly include multiple linear regression (Sharifi et al, 2019) and generalized linear models (Wang et al, 2018).…”

mentioning

confidence: 99%

Effects of rainfall, sediment source–sink landscape, and hydrological connectivity on runoff more than sediment in a karst basin

Gao,

Li,

Shu

et al. 2023

Land Degrad Dev

View full text Add to dashboard Cite

Understanding the mechanisms underlying soil and water loss in karst basins is beneficial for water resources management and soil and water loss conservation. However, the effects of rainfall, sediment source–sink landscape, and hydrological connectivity on runoff sediment characteristics in karst basins are unclear. This study used sediment source–sink landscape and hydrological connectivity theories and a generalized additive model to analyze the changes and combined effects of sediment source–sink landscape patterns, hydrological connectivity, and rainfall on runoff sediment from 2000 to 2020 in the Yeji River basin in the karst region of southwest China. The results indicated that the average sediment transport modulus and runoff modulus in the Yeji River basin during 2000–2020 were 144 t·km−2·a−1 and 15 L·s−1·km−2, respectively. The karst surface and subsurface double‐layer hydrological structure resulted in a lower sediment transport modulus than in non‐karst areas. The sediment‐source landscape area decreased, sediment‐sink landscape area increased, and sediment source–sink landscape load ratio index slightly increased. Hydrological connectivity exhibited a decreasing trend over time, but the overall change was small. Changes in source–sink landscape patterns significantly affected hydrological connectivity. As the main affecting factor, rainfall and its combined effects with sediment source–sink landscape patterns and hydrological connectivity explained 85% and 95% of the variances in runoff, and 40% and 63% of the variances in sediment, respectively. These results suggest that a rational layout of the sediment source–sink landscape can effectively improve hydrological connectivity and regulate the soil and water loss process. Our findings have important implications for karst basin ecological restoration and protection.

show abstract

“…This is problematic for our application as typical hydrologic data sets used in regional analyses contain a large number of candidate variables, many of which are highly correlated. For these reasons, current applications of GAMs to flood quantiles commonly rely on backwards stepwise selection (Chebana et al, 2014;Rahman et al, 2018;Noor et al, 2022;Msilini et al, 2022) sometimes coupled with a pre-selection step as in Dubos et al (2022). Backward selection approach used for GAMs has the potential to select appropriate covariates at the same rate as shrinkage approaches, but only when the information content of the data is high (Marra and Wood, 2011).…”

Section: Introductionmentioning

confidence: 99%

Regional index flood estimation at multiple durations with generalized additive models

Barna,

Engeland,

Kneib

et al. 2023

Preprint

View full text Add to dashboard Cite

Abstract. Estimation of flood quantiles at ungauged basins is often achieved through regression based methods. In situations where flood retention is important, e.g. floodplain management and reservoir design, flood quantile estimates are often needed at multiple durations. This poses a problem for regression-based models as the form of the functional relationship between catchment descriptors and the response may not be constant across different durations. A particular type of regression model that is well-suited to this situation is a generalized additive model (GAM), which allows for flexible, semi-parametric modeling and visualization of the relationship between predictors and the response. However, in practice, selecting predictors for such a flexible model can be challenging, particularly given the characteristics of available catchment descriptor datasets. We employ a machine learning-based variable pre-selection tool which, when combined with domain knowledge, enhances the practicality of constructing GAMs. In this study, we develop a GAM for index (median) flood estimation with the primary objective of investigating duration-specific differences in how catchment descriptors influence the median flood. As the accuracy of this explainable approach is dependent on the fitted GAM being adequate, the secondary objective of our study is prediction of the median flood at ungauged locations and multiple durations, where predictive performance and reliability at ungauged locations are used as proxies for adequacy of the GAM. Predictive performance of the GAM is compared to two benchmark models: the existing log-linear model for median flood estimation in Norway and a fully data-driven machine learning model (an extreme gradient boosting tree ensemble, XGBoost). We find that the predictive accuracy and reliability of the GAM matched or exceeded that of the benchmark models at both durations studied. Within the predictor set selected for this study, we observe duration-specific differences in the relationship between the median flood and the two catchment descriptors effective lake percentage and catchment shape. Ignoring these differences results in a statistically significant decline in predictive performance. This suggests that models developed and estimated for prediction of the index flood at one duration may have reduced performance when applied directly to situations outside of that specific duration.

show abstract

Short-term forecasting of spring freshet peak flow with the Generalized Additive model

Cited by 7 publications

References 63 publications

Spatial-Temporal Surface Water Availability in the Tulancingo River Basin, Hidalgo, Mexico

Spatial-Temporal Surface Water Availability in the Tulancingo River Basin, Hidalgo, Mexico

Effects of rainfall, sediment source–sink landscape, and hydrological connectivity on runoff more than sediment in a karst basin

Regional index flood estimation at multiple durations with generalized additive models

Contact Info

Product

Resources

About