Abstract:In hydrology, modeling streamflow remains a challenging task due to the limited availability of basin characteristics information such as soil geology and geomorphology. These characteristics may be noisy due to measurement errors or may be missing altogether. To overcome this challenge, we propose a knowledge-guided, probabilistic inverse modeling method for recovering physical characteristics from streamflow and weather data, which are more readily available. We compare our framework with state-of-the-art inverse models for estimating river basin characteristics. We also show that these estimates offer improvement in streamflow modeling as opposed to using the original basin characteristic values. Our inverse model offers 3\% improvement in R$^2$ for the inverse model (basin characteristic estimation) and 6\% for the forward model (streamflow prediction). Our framework also offers improved explainability since it can quantify uncertainty in both the inverse and the forward model. Uncertainty quantification plays a pivotal role in improving the explainability of machine learning models by providing additional insights into the reliability and limitations of model predictions. In our analysis, we assess the quality of the uncertainty estimates. Compared to baseline uncertainty quantification methods, our framework offers 10\% improvement in the dispersion of epistemic uncertainty and 13\% improvement in coverage rate. This information can help stakeholders understand the level of uncertainty associated with the predictions and provide a more comprehensive view of the potential outcomes.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in hydrology, due to the limited availability of watershed characteristic information (such as soil geology and geomorphology) and the possible noise or missing of these characteristics, streamflow modelling becomes a challenging task. To solve this problem, the author proposes a knowledge - guided probabilistic inverse modelling method, aiming to recover physical characteristics from more readily available streamflow and meteorological data. Specifically, the main contributions of the paper include: 1. **Proposing a new probabilistic inverse model**: This model can simultaneously learn the representation of static features and quantify the uncertainty in these predictions. By modifying the self - supervised learning auto - encoder architecture, the encoder parameters are estimated using the Bayes by Backprop method to learn the posterior weight distribution and quantify the uncertainty in static feature reconstruction. 2. **Introducing the Uncertainty - Based Learning (UBL) method**: This method reduces epistemic uncertainty (i.e., uncertainty due to imperfect models and data) by adding a regularization term to the loss function. Specifically, higher penalties are imposed on reconstructions with higher uncertainty, which are reflected in the loss. 3. **Improving the performance of streamflow prediction**: Using these robust static feature reconstructions, the paper shows performance improvements in the forward model (streamflow prediction). Compared with the baseline method, the proposed framework improves the $R^2$ value by 3% in the inverse model (static feature estimation) and by 6% in the forward model (streamflow prediction). 4. **Improving the quantification of uncertainty**: Compared with the baseline uncertainty quantification method, this framework improves the dispersion of epistemic uncertainty by 10% and the coverage by 13%. This helps stakeholders better understand the uncertainty level of prediction results and provides a more comprehensive view of potential results. In conclusion, by introducing a new probabilistic inverse model and an uncertainty - based learning method, this paper not only improves the accuracy of streamflow prediction but also enhances the interpretability and reliability of the model, especially when dealing with data containing uncertainty.

Uncertainty Quantification in Inverse Models in Hydrology

A Surrogate Modelling Approach Based on Nonlinear Dimension Reduction for Uncertainty Quantification in Groundwater Flow Models

Uncertainty Analysis in Statistical Modeling of Extreme Hydrological Events

Surrogate-Based Inverse Modeling of the Hydrological System: An Adaptive Approach Considering Surrogate Structural Error.

Uncertainty in High and Low Flows Due to Model Structure and Parameter Errors

Quantification of predictive uncertainty in hydrological modelling by harnessing the wisdom of the crowd: Methodology development and investigation using toy models

Surrogate-Based Bayesian Inverse Modeling of the Hydrological System: An Adaptive Approach Considering Surrogate Approximation Error

Quantile-Based Hydrological Modelling

Total Uncertainty Quantification in Inverse PDE Solutions Obtained with Reduced-Order Deep Learning Surrogate Models

Quantification of predictive uncertainty in hydrological modelling by harnessing the wisdom of the crowd: A large-sample experiment at monthly timescale

Probabilistic and ensemble simulation approaches for input uncertainty quantification of artificial neural network hydrological models

A two-step surrogate method for sequential uncertainty quantification in high-dimensional inverse problems

Nonlinear Expectation Inference for Direct Uncertainty Quantification of Nonlinear Inverse Problems

Uncertainty Quantification of Hydrological Predictions Using a Hierarchical Error Model

Accelerating Uncertainty Quantification of Groundwater Flow Modelling Using a Deep Neural Network Proxy

Streamflow Prediction with Uncertainty Quantification for Water Management: A Constrained Reasoning and Learning Approach

A Comprehensive Survey of Inverse Uncertainty Quantification of Physical Model Parameters in Nuclear System Thermal-Hydraulics Codes

Observational Uncertainty in Hydrological Modelling using Data Depth

Uncertainty quantification in watershed hydrology: Which method to use?

Inverse modeling of hydrologic systems with adaptive multi-fidelity Markov chain Monte Carlo simulations

Parameter estimation and uncertainty analysis in hydrological modeling