Hydrological Model Data Allocation via Discrete Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hydrological models face challenges in maintaining consistency between checking and validation data sets, leading to over-optimistic or pessimistic predictions due to differences in statistical distribution, which affects their credibility and practical application in water conservancy.
Innovation Solution
A novel method combining self-organizing map (SOM) clustering and DUPLEX sampling, employing discrete sampling to ensure consistency between checking and validation data sets, allowing the hydrological process model to run continuously and allocate data discretely for robust performance and improved portability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If observation data are divided into checking and validation data sets using traditional sampling methods (SBSS-N or DUPLEX), then the model checking effectiveness is improved, but the distribution consistency between checking and validation data sets deteriorates, leading to over-optimistic or pessimistic estimation of model performance
Solution Approach 1:
The patent transforms the continuous time-series data into discrete data points by changing the temporal parameter representation. This allows selective sampling of data points while maintaining distribution consistency, resolving the contradiction between improving model checking effectiveness and maintaining distribution stability.
Solution Approach 2:
The patent segments the continuous observation data into discrete data points that can be independently selected and assigned to checking or validation sets. This segmentation enables controlled sampling strategies that preserve distribution characteristics while improving model evaluation reliability.
2Productivity
If continuous time-series data are used for model checking, then the hydrological process model can capture temporal dynamics, but the distribution consistency with validation data deteriorates due to temporal bias
Solution Approach 1:
The patent extracts specific discrete data points from the continuous time-series data for model checking, separating them from the full continuous dataset. This extraction allows the model to capture temporal dynamics from the continuous data while using selectively sampled discrete points for checking, thereby maintaining distribution consistency with validation data.
3Stability of the object's composition
If discrete sampling is used to allocate data to checking and validation sets, then distribution consistency is improved, but the complexity of data allocation methodology increases
Solution Approach 1:
The patent employs a self-service sampling approach where data points are automatically selected and assigned to checking or validation sets based on predefined criteria and randomization. This automated self-service mechanism reduces manual intervention complexity while maintaining distribution consistency, balancing the trade-off between consistency improvement and methodological complexity.
Data Source
AI summary
The present disclosure provides a method based on consistency of a distribution feature of checking-validation data for establishing a hydrological model. The method includes: S1: proposing an idea of discrete checking of data for a hydrological process model according to S11-S12; S2: using an MDUPLEX method to allocate an original runoff data set D to a checking set C and a validation set E according to S21-S28; and S3: checking and validating the model according to S31-S32, determining a model parameter, and establishing the hydrological process model. The present disclosure guarantees consistency of performance of the hydrological process model during checking and validation periods by means of discrete sampling, so as to improve effectiveness of the hydrological process model and stability of engineering application.