Abstract:As in many fields of dynamic modeling, the long runtime of hydrological models hinders Bayesian inference of model parameters from data. By replacing a model with an approximation of its output as a function of input and/or parameters, emulation allows us to complete this task by trading-off accuracy for speed. We combine (i) the use of a mechanistic emulator, (ii) low-discrepancy sampling of the parameter space, and (iii) iterative refinement of the design data set, to perform Bayesian inference with a very small design data set constructed with 128 model runs in a parameter space of up to eight dimensions. In our didactic example we use a model implemented with the hydrological simulator SWMM that allows us to compare our inference results against those derived with the full model. This comparison demonstrates that iterative improvements lead to reasonable results with a very small design data set.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the computational efficiency problem encountered during Bayesian inference in hydrological modeling. Specifically, since hydrological models are usually based on complex systems of ordinary differential equations or partial differential equations, these models have a long running time, resulting in a very high computational cost when using traditional methods for Bayesian parameter estimation, and even being infeasible in some cases. To overcome this challenge, the paper proposes a method using a mechanistic emulator, combined with low - discrepancy sampling and iterative improvement of the design data set, to achieve efficient and accurate Bayesian inference. This method allows researchers to significantly reduce the number of required model runs while maintaining high accuracy, thereby accelerating the Bayesian inference process. ### Main problems solved 1. **Computational efficiency problem**: Traditional Bayesian inference methods require a large number of model runs to evaluate different input and parameter values, which is very time - consuming when dealing with complex hydrological models. 2. **Limit on the number of model runs**: By using a mechanistic emulator, the paper aims to approximate the model output with a small number of design data points (for example, 128 model runs), thus greatly reducing the number of model runs. 3. **Trade - off between accuracy and speed**: The paper explores how to achieve rapid Bayesian inference through the mechanistic emulator and the iterative improvement method while ensuring a certain level of accuracy. ### Method overview - **Mechanistic emulator**: Use the Gaussian Process as a mechanistic emulator, which can provide the probability distribution of the model output, thereby quantifying uncertainty. - **Low - discrepancy sampling**: Use the Halton sequence to generate design parameter vectors to ensure uniform coverage of the parameter space. - **Iterative improvement**: By gradually adding new design data points, continuously optimize the performance of the emulator to ensure high accuracy within the relevant parameter regions. ### Application case The paper demonstrates the effectiveness of this method through a practical case. Researchers used the Storm Water Management Model (SWMM) to model a medium - sized urban watershed in Adliswil, Switzerland, and calibrated it for only one rainfall event. By comparing with the calibration results of the full model, it is verified that this method can still obtain reasonable results when using a small number of design data points. ### Key contributions - **Application of the mechanistic emulator**: The introduction of the mechanistic emulator can significantly reduce the number of model runs while maintaining high accuracy. - **Iterative improvement strategy**: Proposed an iterative improvement method for the design data set to ensure high accuracy within the relevant parameter regions. - **Verification of practical applications**: Verified the effectiveness of the method through practical cases, showing its potential application value in hydrological modeling. In conclusion, through the introduction of the mechanistic emulator and the iterative improvement method, this paper effectively solves the computational efficiency problem of Bayesian inference in hydrological modeling, providing new ideas and tools for research in related fields.

Accelerating Bayesian inference in hydrological modeling with a mechanistic emulator

A Robust Strategy to Account for Data Sampling Variability in the Development of Hydrological Models

Appraisal of data-driven and mechanistic emulators of nonlinear hydrodynamic urban drainage simulators

Parametric emulation and inference in computationally expensive integrated urban water quality simulators

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

Improving Simulation Efficiency of MCMC for Inverse Modeling of Hydrologic Systems With a Kalman‐Inspired Proposal Distribution

Simulation-based inference for parameter estimation of complex watershed simulators

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

Evolving Bayesian Emulators for Structured Chaotic Time Series, with Application to Large Climate Models

Calibration of a Distributed Flood Forecasting Model with Input Uncertainty Using a Bayesian Framework

A comparison of emulation methods for Approximate Bayesian Computation

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

A Bayesian hierarchical framework for post-processing daily streamflow simulations across a river network

Evaluation of Uncertainties in Input Data and Parameters of a Hydrological Model Using a Bayesian Framework: A Case Study of a Snowmelt–Precipitation-Driven Watershed

An improved Approximate Bayesian Computation approach for high-dimensional posterior exploration of hydrological models

Bayesian inference for geophysical fluid dynamics using generative models

A Hybrid Surrogate Modelling Strategy for Simplification of Detailed Urban Drainage Simulators

A Bayesian Surveillance Model to Track Variable Rainfall-Runoff Responses for Small Watersheds

Multifidelity multilevel Monte Carlo to accelerate approximate Bayesian parameter inference for partially observed stochastic processes

Emulation of Multivariate Simulators Using Thin-Plate Splines with Application to Atmospheric Dispersion

An Adaptive Kriging Surrogate Method for Efficient Joint Estimation of Hydraulic and Biochemical Parameters in Reactive Transport Modeling.