Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning Generative Models for Lumped Rainfall-Runoff Modeling

Yang Yang, Ting Fong May Chui

Abstract

This study presents a novel generative modeling approach to rainfall-runoff modeling, focusing on the synthesis of realistic daily catchment runoff time series in response to catchment-averaged climate forcing. Unlike traditional process-based lumped hydrologic models that depend on predefined sets of variables describing catchment physical properties, our approach uses a small number of latent variables to characterize runoff generation processes. These latent variables encapsulate the intrinsic properties of a catchment and can be inferred from catchment climate forcing and discharge data. By sampling from the latent variable space, the model generates runoff time series that closely resemble real-world observations. In this study, we trained the generative models using neural networks on data from over 3,000 global catchments and achieved prediction accuracies comparable to current deep learning models and various conventional lumped models, both within the catchments from the training set and from other regions worldwide. This suggests that the runoff generation process of catchments can be effectively captured by a low-dimensional latent representation. Yet, challenges such as equifinality and optimal determination of latent variables remain. Future research should focus on refining parameter estimation methods and exploring the physical meaning of these latent dimensions to improve model applicability and robustness. This generative approach offers a promising alternative for hydrological modeling that requires minimal assumptions about the physical processes of the catchment.

Learning Generative Models for Lumped Rainfall-Runoff Modeling

Abstract

This study presents a novel generative modeling approach to rainfall-runoff modeling, focusing on the synthesis of realistic daily catchment runoff time series in response to catchment-averaged climate forcing. Unlike traditional process-based lumped hydrologic models that depend on predefined sets of variables describing catchment physical properties, our approach uses a small number of latent variables to characterize runoff generation processes. These latent variables encapsulate the intrinsic properties of a catchment and can be inferred from catchment climate forcing and discharge data. By sampling from the latent variable space, the model generates runoff time series that closely resemble real-world observations. In this study, we trained the generative models using neural networks on data from over 3,000 global catchments and achieved prediction accuracies comparable to current deep learning models and various conventional lumped models, both within the catchments from the training set and from other regions worldwide. This suggests that the runoff generation process of catchments can be effectively captured by a low-dimensional latent representation. Yet, challenges such as equifinality and optimal determination of latent variables remain. Future research should focus on refining parameter estimation methods and exploring the physical meaning of these latent dimensions to improve model applicability and robustness. This generative approach offers a promising alternative for hydrological modeling that requires minimal assumptions about the physical processes of the catchment.
Paper Structure (29 sections, 2 equations, 12 figures, 1 table)

This paper contains 29 sections, 2 equations, 12 figures, 1 table.

Table of Contents

  1. Introduction
  2. A Generative View on Lumped Hydrological Modeling
  3. Deep Learning Approaches for Generative Modeling in Hydrology
  4. Methods
  5. Assumptions of Generative Hydrological Models
  6. Learning a Generative Hydrological Model
  7. Application of a Generative Model to Arbitrary Catchments
  8. Experimental Setup
  9. Neural Network Architecture of the Generative Model
  10. Training and In-Sample Evaluation Strategy
  11. Out-of-Sample Evaluation Strategy
  12. Data
  13. Catchment Climate Forcing and Discharge Data
  14. Benchmark Data on Lumped Hydrological Models' Performance
  15. Numerical Experiments and Results
  16. ...and 14 more sections

Figures (12)

  • Figure 1: Illustration of the processes of generating discharge time series using a 2-parameter Leaky Bucket model.
  • Figure 2: Illustration of the process of generating discharge time series using a generative hydrological model.
  • Figure 3: Architecture of the generative hydrological model $g_\theta$ used in this study.
  • Figure 4: Roles of datasets and dataset splitting strategy.
  • Figure 5: Non-exceedance probability of the training and test KGE scores for each Caravan catchment. The medians of the training and test KGE scores are also displayed. For clarity, results for KGE values less than -1 are not shown here but can be found in Figure \ref{['fig_caravan_train_vs_test']}. The test KGE scores for two catchments unavailable because the recorded runoffs were constantly 0.
  • ...and 7 more figures